INTRODUCTION: Introduction, Rationale, System Descriptions, etc. INTEGER TEST 1: Software Installation of IRIX 6.5 (Default Configuration). INTEGER TEST 2: System Bootup Time for IRIX 6.5 and IRIX 6.2. INTEGER TEST 3: Installing near-identical selections from all other 6.5 CDs. INTEGER TEST 4: Application Initiation Times for IRIX 6.5 and IRIX 6.2. INTEGER TEST 5: Movie Data Conversion Test 1. INTEGER TEST 6: Movie Data Conversion Test 2. INTEGER TEST 7: Software Installation of IRIX 6.2 (Default Configuration). INTEGER TEST 8: Compiling Lynx V2.7.1 INTEGER TEST 9: RC5-64 and DES Encryption-cracking Performance Benchmark GRAPHICS TEST 1: Inventor Objects: rotation speed (single-buffered). GRAPHICS TEST 2: Buttonfly Powerflip Models: rotation speed (double-buffered). IMAGING TEST 1: GIMP V1.0.4: Filters -> Distorts -> Ripple IMAGING TEST 2: GIMP V1.0.4: Script-Fu -> Decor -> Lava IMAGING TEST 3: GIMP V1.0.4: Script-Fu -> Decor -> Circuit MEMORY USAGE: Default 6.2/6.3/6.5 memory usage given varying RAM resources. CDROM TEST 1: Analysing the 6.5 OS installation times to compare CDROM performance. CDROM TEST 2: Copying a large file straight to disk. CDROM TEST 3: Copying many smaller files straight to disk. DAT TEST 1: Archiving a single large file to DAT (DDS1 vs. DDS3). DAT TEST 2: Extracting a single large file from DAT (DDS1 vs. DDS3). DAT TEST 3: Archiving a large number of small files to DAT (DDS1 vs. DDS3). DAT TEST 4: Extracting a large number of small files from DAT (DDS1 vs. DDS3). DAT TEST 5: Archiving the default 6.5 filesystem to DAT. AUDIO TEST 1: Parametric Equalization and Normalization of a Music CD Track. FUTURE TESTS: A summary of other tests I intend to conduct in the future.
This page presents the results of an extensive series of performance tests on various SGI systems, focusing on CPU, graphics, CDROM and DAT performance, and on memory usage with different RAM configurations.
Also, instead of studying synthetic benchmarks such as SPEC95 (I have many other pages which cover SPEC comparisons), I present an analysis of the OS installation procedure itself to compare integer processing and CDROM performance. Some aspects of software installation do not involve the CDROM - these elements probably involve integer computations that are similar to compiling programs.
Also included are the results of various simple, real-world application tests such as:
The possibly unexpected answer in this case is, "No, you're far better off upgrading the main CPU first." - my analysis explains why.
After several years of helping people with buying decisions, I'm now convinced that this kind of real-world, practical, every-day performance information is what many people making purchasing or upgrade decisions actually want to know, especially those with privately owned systems. Benchmarks such as Viewperf and SPEC95 can be helpful, but only if one's intended tasks are similar to the test, and on their own these benchmarks can only convey relative differences between systems, ie. they cannot convey absolute performance of a single system. Worse, these synthetic tests are often misused, eg. quoting final averages instead of comparing individual test results.
My tests are far more practical in nature; nobody has ever collated this kind of information before for SGIs. If you're wondering what I mean by practical, consider any of the SPEC95 or Viewperf tests: the results are meaningful (frame rates, time durations, etc.), but only people with personal experience of the tests concerned will intuitively understand the complexity of each test, ie. most people cannot use Viewperf or SPEC95 scores to guage how their own system might perform.
Over time, I hope to create the most extensive SGI performance resource anywhere on the web. Buyers will be able to convert their desire for a particular 'level' of performance and abilities into specific system suggestions, eg. if a 2nd-hand buyer wants Netscape to activate in less than 10 seconds, then an R4600PC Indy is not appropriate; if a buyer wants ISDN, then Indy is a good choice.
Further performance data will be added to this page as and when I can perform the necessary tests. They take alot of time to do, and I have to carry out most of them at home, though as time goes by I aquire more and more results from people around the world who have taken the time to run the same tests on their own system(s). Since I do not have access to many types of SGI system (eg. Onyx2, Octane), please feel free to submit your own results based on the test descriptions given here. The results will be added with credit fully given.
My help-page on installing IRIX 6.2 has proved very useful to a great number of people; thus, I decided to record my experiences of dealing with 6.5 as well. Timing the individual installation steps, this page presents analyses of the results to compare integer and CDROM performance.
System Descriptions
Some test results were submitted by other SGI users; for such results, detailed system descriptions (ie. hinv) are given where possible. Here is a description of the systems which I personally tested:
Indy with 256MB RAM, 2GB disk, 24bit XL gfx, 17" monitor, using R4600PC/100, R4600PC/133 and R4400SC/200 CPUs. O2 with R5000SC 200MHz CPU (1MB L2), 256MB RAM, 2x 4.5GB internal disks, 9GB external disk, CRM gfx (rev2, CRMrevC, GBErevB), 20" monitor and 12X CDROM (TOSHIBA XM-5701TA 0167). Indigo2 with R4400SC 250MHz CPU (2MB L2), 256MB RAM, 4.5GB disk, GR3-Elan gfx (4 GEs, 1 RE, 24 bitplanes, GR2rev4, HQ2.1revA, GE7revB, RE3.1revA, VC1revB, MCrevC), 20" monitor. Indigo2 with R4400SC 200MHz CPU (1MB L2), 256MB RAM, 2GB disk, 24bit XL graphics, 17" monitor.The following additional components were used:
The installation test times given in INT tests 1 and 7 do not include the 'human' element of the installation procedure, eg. switching CDROMs, typing in commands, etc. When using my stopwatch to time the installations, I halted the stopwatch whenever a prompt appeared and restarted the stopwatch once the required input was entered and the procedure again underway.
All of the systems I tested had 7200rpm disks, but the types are slightly different:
O2: UltraSCSI (IBM), with an UltraSCSI controller Indigo2: SCSI-2 (Quantum Atlas), ordinary SCSI controller Indy: UltraSCSI (Seagate), ordinary SCSI controller
This means average disk seek times, etc. will vary slightly between the systems.
Performance Tests
I have conducted a large number of tests which can be divided into a number of categories:
Feel free to browse through the results below, or use the
quick-access index at the start of this page.
INTEGER TEST 1: Software Installation of IRIX 6.5 (Default Configuration)
These results came out of my study of installing IRIX 6.5, which you should read if you want to know exactly how the data was obtained. Table 1 includes timing details of all intermediate stages of the installation process. You may need to widen your browser window to view the complete table.
******* Sub-Task Completion Times *******
(16%) (51%) (86%) (91%) (94%) (100%)
CD Read Pre- Inst End of End of
ROM Inst inst F 1 Apps F 2 Tools Exit rqsall
CPU Type / L2 Tools Ends Ends Ends Ends Ends Coms ELF libs
O2 R5000SC/200 1MB: 32 0:42 0:54 06:31 15:07 16:46 17:45 22:52 0:34:06
O2 R5000SC/200 1MB: 12 0:46 0:54 06:58 16:04 18:05 19:20 24:08 0:35:28
I2 R4400SC/250 2MB: 32 0:47 1:11 06:45 15:17 16:57 17:55 23:12 0:36:05
I2 R4400SC/250 2MB: 2 1:16 1:11 09:18 21:06 23:28 24:38 29:57 0:42:31
Indy R4400SC/200 1MB: 32 0:42 1:26 08:32 19:42 21:40 22:50 29:41 0:45:24
Indy R4400SC/200 1MB: 2 1:16 1:26 11:11 25:49 28:29 29:58 36:53 0:52:35
Indy R5000SC/180 512K: 32 0:49 1:00 09:04 20:40 22:37 23:49 31:11 0:46:55
Indy R5000PC/150: 2 1:19 1:17 12:06 27:47 32:05 ? 41:58 ? [1]
Indy R4600PC/133: 32 0:44 1:45 11:48 28:26 31:28 33:07 46:11 1:08:45
Indy R4600PC/133: 2 1:16 1:45 14:10 33:37 37:12 39:00 52:05 1:14:39
Indy R4600PC/100: 32 0:35 2:04 12:36 30:18 33:36 35:20 48:28 1:12:11
Indy R4600PC/100: 2 1:16 2:04 15:01 35:33 39:32 41:28 54:54 1:18:36
Table 1: Detailed Timings for a Default IRIX 6.5 OS Installation
[1] Test run by Euan Harris of Glasgow University.
Rationale
Some parts of the software installation process do not involve the CDROM or source device at all. This includes:
Those aspects of the installation which do use the CDROM can be analysed to study how CDROM performance varies with CPU strength, and whether or not a fast CDROM is worth having given a particular CPU type (these results are discussed later).
Table 2 shows the time taken for each individual step in the installation process, ie. the times obtained by subtracting one column from the next.
CD PRE- INST- EXIT RQSALL
ROM INST F1 APPS F2 TOOLS COMS ELF
O2 R5000SC/200: 32 0:54 05:37 08:36 01:39 00:59 05:07 11:14
O2 R5000SC/200: 12 0:54 06:04 09:06 02:01 01:15 04:48 11:20
I2 R4400SC/250: 32 1:11 05:34 08:32 01:40 00:58 05:17 12:53
I2 R4400SC/250: 2 1:11 08:07 11:48 02:22 01:10 05:19 12:34
Indy R4400SC/200: 32 1:26 07:06 11:10 01:58 01:10 06:51 15:43
Indy R4400SC/200: 2 1:26 09:45 14:38 02:40 01:29 06:55 15:42
Indy R5000SC/180: 32 1:00 08:04 11:36 01:57 01:12 07:22 15:44
Indy R5000SC/180: 2
Indy R4600PC/133: 32 1:45 10:03 16:38 03:02 01:39 13:04 22:34
Indy R4600PC/133: 2 1:45 12:25 19:27 03:35 01:48 13:05 22:34
Indy R4600PC/100: 32 2:04 10:32 17:42 03:18 01:44 13:08 23:43
Indy R4600PC/100: 2 2:04 12:57 20:32 03:59 01:56 13:26 23:42
Table 2: Individual times for each installation step (IRIX 6.5)
Here are the percentage differences for the paired results, ie. highest divided by lowest (these figures should be low in order to be usable):
PRE- EXIT RQSALL
INST COMS ELF
O2 R5000SC/200: 0% 6.6% 0.9%
I2 R4400SC/250: 0% 1.0% 2.5%
Indy R4400SC/200: 0% 1.0% 0.0%
Indy R4600PC/133: 0% 0.0% 0.0%
Indy R4600PC/100: 0% 2.0% 0.0%
Table 3: Paired result differences.
These are satisfactory, indeed they're perfect in many cases. However, I may run the O2 tests again to make sure I didn't make any mistakes during the timings.
Now for the all-important comparison: the average performance for each system compared to the average performance of the slowest system (Indy R4600PC/100) for each stage (divide the target time by the slowest system's time). Thus, the numbers denote how much faster each stage was executed compared to the slowest system (ie. smaller = better).
PRE- EXIT RQSALL AVERAGE
INST COMS ELF
O2 R5000SC/200: 0.44 0.37 0.48 | 0.44
I2 R4400SC/250: 0.57 0.40 0.54 | 0.49
Indy R4400SC/200: 0.69 0.52 0.66 | 0.62
Indy R4600PC/133: 0.85 0.98 0.95 | 0.96
Indy R4600PC/100: 1.00 1.00 1.00 | 1.00
Table 4: Comparison to Indy R4600PC/100
performance (factor differences)
The variances between tests are probably due to the different degrees to which the disk is being accessed, the benefit of having any L2 cache to aid the computations, differences in disk speeds and latencies, different amounts of data being processed, etc.
It's very noticable that not having any L2 cache hinders the processing of the EXIT COMS and RQSALL ELF stages.
Note: to calculate percentage improvements, subtract each number from 1 and multiply by 100, eg. the RQSALL step for Indy R4600SC/200 is 34% faster than Indy R4600PC/100, while (overall) Indigo2 R4400SC/250 is 51% faster than Indy R4600PC/100 at performing these particular integer/disk tasks.
INTEGER TEST 2: System Bootup Time for IRIX 6.5 and IRIX 6.2
This test combines integer computations and disk accesses. All systems had 128MB RAM for this test, as they did for the OS installation tests. Remember that these times are for a default installation only.
Time to
boot up the
system (mm:ss)
Indigo2 R10000SC/195 1MB: 1:17
Indy R4400SC/200 1MB: 1:23
Indigo2 R4400SC/250 2MB: 1:35
Indy R5000SC/180 512K: 1:38
Indy R5000PC/150: 1:48
Indy R4600PC/133: 1:48
Indy R4600PC/100: 1:48
O2 R5000SC/200 1MB: 2:04
Table 5: IRIX 6.5 Bootup Times
For some reason, O2 has always taken longer to bootup than older systems. Does anyone know why? It's probably due to the more complex I/O arhcitecture, and thus likely more startup initialisation signals, etc.
Note that because the above results are for a default installation, the bootup sequence includes a deliberate 10 second delay while the message concerning 'standalone network mode' is displayed. Hence, if one edited the /etc/init.d/network file so that MSGTIME was set to zero, then all the above results could be decreased by 10 seconds.
Here is the table for IRIX 6.2:
Time to
boot up the
system (mm:ss)
Indy R4400SC/200 1MB: 1:07
Indy R4600PC/133: 1:22
Indy R4600PC/100: 1:22
Indigo2 R4400SC/250 2MB: 1:24
Table 5b: IRIX 6.2 Bootup Times
Notice that in both cases the Indigo2 R4400SC/250 takes slightly longer than the R4400SC/200 Indy - I expect this is because the Indigo2 has a larger number of internal checks and initialisations to do since it has more internal hardware systems than Indy, etc.
INTEGER TEST 3: Installing near-identical selections from all
other 6.5 CDs
Table 6 shows the time taken to install everything else, which includes more-or-less identical selections from Foundation 1, Foundation 2, Applications, Development Foundation, Development Libraries, NFS, Varsity (Aug98), MIPS Pro Compiler Patches and Cosmo Suite. CDROMs used were 12X for O2, 32X CDROM for Indy and Indigo2 (one can extrapolate times for a 2X CDROM from Table 1 and Table 2). Times are in hours, minutes and seconds (note that these times do not include the human element of swapping the CDs, selecting the products to install with the mouse, etc.)
CD Total
ROM Time
(h:mm:ss)
O2 R5000SC/200: 12 1:16:50
I2 R4400SC/250: 32 1:16:00
Indy R4400SC/200: 32 1:26:20
Indy R4600PC/133: 32 2:06:10
Indy R4600PC/100: 32 2:26:14
Table 6: Installing All Other CDs
INTEGER TEST 4: Application Initiation Times for IRIX 6.5 and IRIX 6.2
These tests measure how long it takes to fully bring up an application after one has double-clicked on its icon, to the nearest half-second.
IRIX 6.5:
Manage
Installed
Activating Software Activating
Netscape (swmgr) CosmoCreate
Fuel R16000SC/700 4MB L2: 1.0 N/A 0.5 [hinv]
O2 R12000SC/400 2MB L2: 1.5 0.5 1.5
Indigo2 R10000/195 1MB L2: 3.5 2.0
Indigo2 R10000/175 1MB L2: 4.0 2.5
O2 R5200SC/300 1MB L2: 4.5 3.0 [hinv]
O2 R5000SC/200 1MB L2: 5.0 3.5 5.0
Indigo2 R4400SC/250: 5.0 4.0 5.0
Indy R5000SC/180: 6.5 6.5 [hinv]
Indy R4400SC/200: 8.0 7.0
Indy R4600SC/133: 9.0 7.5
Indy R5000PC/150: 9.5 8.5 [hinv]
Indy R4600PC/133: 14.5 13.5
Indy R4600PC/100: 15.0 14.0
Table 7: Application Initiation Times for IRIX 6.5
Here is the same information for IRIX 6.2. Note the differences
(obviously, one cannot compare to O2):
Manage
Installed
Activating Software Activating Activating
Netscape (swmgr) CosmoCode CosmoCreate
Indigo2 R4400SC/250: 4.0 3.5
Indy R4400SC/200: 5.0 5.5
Indy R4600PC/133: 9.5 11.0 11.0 12.0
Indy R4600PC/100: 11.0 12.0
Table 7b: Application Initiation Times for IRIX 6.2
There is a noticable decrease in speed when moving to IRIX 6.5 for
older systems such as Indy. Newer systems are less affected.
Now for applications that also include a degree of
floating-point/graphics processing:
Activating Activating
CosmoWorlds CosmoWorlds
with tunnels.wrl
Fuel R16000SC/700 4MB L2: 1.5 1.5 [hinv]
Indigo2 R10000SC 175MHz 1MB L2: 5.0 7.0
O2 R5000SC 200MHz 1MB L2: 7.0 8.0
Indigo2 R4400SC 250MHz 2MB L2: 9.0 12.0
Indigo2 R4400SC 200MHz 1MB L2: 11.0 13.0
Indy R4400SC 200MHz 1MB L2:
Indy R4600SC 133MHz 512K L2:
Indy R4600PC 133MHz: 19.0 23.5
Indy R4600PC 100MHz:
Table 7c: Further Application Initiation Times for IRIX 6.5
tunnels.wrl is a very simple VRML model I made long ago. It consists
of a number of brick-textured tubes, twisted around each other.
INTEGER TEST 5: Movie Data Converstion Test 1
Time to convert 11.4 seconds of half-size-PAL MJPEG-A video (22MB file) to different compression schemes with mediaconvert (minutes and seconds). Note that for the conversion to uncompressed video, the type of disk used will make a difference. The table is arranged in order of MPEG performance. Notice the better results for certain systems when hardware acceleration is available, eg. Photo JPEG on O2.
Photo
NONE RLE JPEG Indeo Cinepak MPEG
6.5 Fuel R16000SC 700MHz 4MB L2: 0:07 0:12 0:17 0:20 00:47 01:26
6.5 Fuel R14000SC 500MHz 2MB L2: 0:11 0:17 0:25 0:30 01:09 02:08
6.5 Octane R12000SC 400MHz 2MB L2: 0:14 0:21 0:29 0:36 01:24 02:32
6.5 O2 R12000SC 400MHz 2MB L2: 0:17 0:24 0:17 0:42 01:36 02:44
6.5 Octane R12000SC 300MHz 2MB L2: 0:19 0:29 0:41 0:50 02:08 03:31
6.5 O2 R12000SC 300MHz 1MB L2: 0:23 0:33 0:19 0:54 01:57 03:42
6.5 Onyx2 R10000SC 250MHz 4MB L2: 0:20 0:31 0:45 0:57 02:12 03:53
6.5 Octane R10000SC 250MHz 1MB L2: 0:22 0:33 0:48 0:58 02:15 04:03
6.5 O2 R10000SC 195MHz 1MB L2: 0:22 0:37 0:19 1:10 02:41 05:10
6.5 Indigo2 R10000SC 195MHz 1MB L2: 0:34 0:48 1:07 1:22 02:55 05:18
6.5 Octane R10000SC 195MHz 1MB L2: 0:30 0:45 1:05 1:19 02:55 05:23
6.5 Origin200 R10000SC 180MHz 1MB L2: 0:30 0:46 1:04 1:20 03:01 05:27
6.5 Indigo2 R10000SC 175MHz 1MB L2: 0:37 0:54 1:15 1:32 03:53 05:57
6.5 O2 R5200SC 300MHz 1MB L2: 0:21 0:36 0:18 1:20 03:11 07:01
6.5 Indigo2 R4400SC 250MHz 2MB L2: 0:58 1:21 2:18 2:45 06:16 09:56
6.5 O2 R5000SC 200MHz 1MB L2: 0:30 0:46 0:18 2:01 05:41 10:14
6.3 O2 R5000SC 200MHz 1MB L2: 0:27 0:56 0:16 2:47 08:15 11:41
6.5 O2 R5000SC 180MHz 512K L2: 0:25 0:49 0:21 2:04 05:10 11:29
6.5 Indy R5000SC 180MHz 512K L2: 1:11 1:45 2:11 2:50 05:56 12:11
6.5 Indigo2 R4400SC 200MHz 1MB L2: 1:09 1:45 2:27 3:20 06:28 12:26
6.5 Indy R4400SC 200MHz 1MB L2: 1:12 1:54 3:00 3:28 07:55 12:29
6.5 Indy R5000SC 150MHz 512K L2: 1:16 1:45 2:30 3:13 07:33 14:22
6.5 Indy R5000PC 150MHz: 1:23 1:51 2:39 3:25 07:20 15:14
6.5 Indigo2 R4400SC 150MHz 1MB L2: 1:30 2:13 3:20 4:24 08:38 16:42
6.5 Indy R4600SC 133MHz 512K L2: 1:32 2:09 3:07 4:09 08:05 17:33
6.5 Indy R4600PC 133MHz: 1:47 2:19 4:03 4:49 10:52 20:03
6.5 Indy R4600PC 100MHz: 2:04 2:53 4:57 5:56 13:35 24:54
6.5 Indigo R4000SC 100MHz 1MB L2: 2:21 3:28 5:08 6:49 14:18 25:32
Table 8
Note that tasks which use less agressive compression methods are probably more
disk-bound, eg. 'NONE' in Table 8. Also, this test is carried out using
mediaconvert; thus, systems which have no graphics head (Origin200, etc.) are
accessed remotely from another SGI (in the above table, I used an O2 via
100Mbit switch) - I noticed that remote tasks displayed elsewhere suffered a
speed hit if the connection speed was only 10Mbit (I originally tested using an
Indigo2 for remote display, but changed to O2 when it was obvious the speed was
lower than expected), ie. the overhead of displaying a remote window (X
communications, etc.) can wipe out the benefit of a faster CPU if the link is
too slow. This would be less important if the task was a command line task such
as dmconvert. Some of the above tests could be done using command-line
dmconvert instead of mediaconvert; thus, Origin200 can actually be faster than
the above table conveys.
INTEGER TEST 6: Movie Data Conversion Test 2
Time to convert 11.4 seconds of half-size-PAL MJPEG-A to quarter-size MJPEG-A with mediaconvert (minutes and seconds). Note that some results may seem oddly out of order. This is usually due to the different OS versions involved, ie. changes made to mediaconvert can affect performance. Disk speed may also be relevant.
6.5 Fuel R16000SC 700MHz 4MB L2: 0:12
6.5 O2 R12000SC 400MHz 2MB L2: 0:15
6.5 Fuel R14000SC 500MHz 2MB L2: 0:19
6.5 Octane R12000SC 400MHz 2MB L2: 0:22
6.5 O2 R12000SC 300MHz 1MB L2: 0:35
6.5 Octane R10000SC 250MHz 1MB L2: 0:38
6.5 Origin200 R10000SC 180MHz 1MB L2: 0:46
6.5 O2 R10000SC 195MHz 1MB L2: 0:46
6.5 Indigo2 R10000SC 175MHz 1MB L2: 0:47
6.5 Octane R10000SC 195MHz 1MB L2: 0:50
6.5 O2 R5200SC 300MHz 1MB L2: 0:53
6.5 O2 R5000SC 200MHz 1MB L2: 1:13
6.5 O2 R5000SC 180MHz 512K L2: 1:17
6.3 O2 R5000SC 200MHz 1MB L2: 1:25
6.5 Indigo2 R4400SC 250MHz 2MB L2: 1:41
6.5 Indigo2 R4400SC 200MHz 1MB L2: 2:00
6.5 Indy R5000SC 180MHz 512K L2: 2:01
6.5 Indy R4400SC 200MHz 1MB L2: 2:11
6.5 Indy R5000SC 150MHz 512K L2: 2:13
6.5 Indy R5000PC 150MHz: 2:29
6.5 Indigo2 R4400SC 150MHz 1MB L2: 2:45
6.5 Indy R4600SC 133MHz 512K L2: 2:54
6.5 Indy R4600PC 133MHz: 3:33
6.5 Indy R4600PC 100MHz: 4:12
6.5 Indigo R4000SC 100MHz 1MB L2: 4:35
Table 9
The R12K/400 O2 result is strange. I'll rerun it again sometime to be sure it's correct (perhaps it's using ICE in some way).
INTEGER TEST 7: Software Installation of IRIX 6.2 (Default
Configuration)
This analysis is identical to that given for INTEGER TEST 1, except the discussion which follows is not as detailed, ie. I concentrate on the results. For an explanation of how this analysis is performed and the rationale behind it, see the explanations given for INTEGER TEST 1.
Table A summarises the time to read the installation tools from the CDROM, the total time taken for the OS installation, and how the results compare to the slowest system (R4600PC/100 with 2X CDROM).
Time Time to
CDROM to Read the install the FACTOR
SPEED Installation 2 base CDs COMPARED
Tools (m:ss) (mm:ss) TO SLOWEST
Indigo2 R10000SC/175: 32 0:32
Indigo2 R4400SC/250: 32 1:07 18:02 0.44
Indigo2 R4400SC/250: 2 1:54 20:40 0.51
Indy R4400SC/200: 32 0:46 21:48 0.53
Indy R4400SC/200: 2 1:51 25:11 0.62
Indy R4600PC/133: 32 0:50 35:58 0.88
Indy R4600PC/133: 2 1:51 39:07 0.96
Indy R4600PC/100: 32 0:50 38:00 0.93
Indy R4600PC/100: 2 1:51 40:48 1.00
Table A: IRIX 6.2 Software Installation Timing Summary
It's rather strange that my Indigo2 R4400SC/250 with a 32X CDROM takes so long to read the installation tools (17 seconds longer than the R4600PC/100 Indy). Installing 6.5 on my Indigo2 system took much less time to read the installation tools (47 seconds). I ran the tests again, but saw the same results - something is slowing down the I2's ability to access the CDROM that first time (it doesn't seem to be happening during the main installation); anyone have any ideas?
Even so, it's clear that reading data from a 2X CDROM leaves the main CPU starved for data, ie. the bottleneck is the CDROM's ability to get data off the CD and into the system. The proof of this is that the time taken to read the installation tools does not correlate with CPU power.
In contrast to the 2X results, a faster CPU helps when reading data from a 32X CDROM to such an extent that the time taken to carry out a brief action (such as reading the installation tools) probably depends more on how a system's I/O hardware handles the data once it has left the CDROM. Since O2, Indigo2 and Indy have different I/O systems (very much so in the case of O2), it's easy for differences not related to CPU power to creep into the timing results. Thus, to properly evaluate how different CPUs utilise a 32X CDROM, one needs more demanding tests. Relevant tests of this type are present on this page in the form of a study of how different CDROMs are exploited during the installation of IRIX 6.5, and how different systems cope when copying different types of file(s) from a CDROM straight to disk [one large file | lots of small files].
Remember that the OS installation process for 6.2 is not the same as 6.5. Two CDs form the basic 6.2 OS set: each CD is dealt with separately, using a 'delay_conflicts' flag to avoid installation conflicts. As a result, the columns in Table B below are not the same events as those shown in Table 1 for installing IRIX 6.5.
************* Sub-task Completion Times ***************
------ 1st CD ------ ------ 2nd CD ------ -Post-
CD Pre- Exit 1ST Pre- Exit 2ND rqsall
ROM inst Coms CD inst Coms CD ELF
Ends Begin Done Ends Begin Done libs
I2 R10000SC/175: 32 00:14 05:53 07:47 08:02 12:25 14:02 17:30
I2 R4400SC/250: 32 00:15 05:12 07:45 08:05 13:08 15:16 18:02
I2 R4400SC/250: 2 00:15 06:20 08:52 09:11 15:34 17:47 20:40
Indy R4400SC/200: 32 00:20 06:18 09:26 09:53 16:01 18:39 21:48
Indy R4400SC/200: 2 00:20 07:40 10:47 11:14 19:00 22:00 25:11
Indy R4600PC/133: 32 00:32 09:01 15:54 16:35 26:08 31:45 35:58
Indy R4600PC/133: 2 00:32 10:33 17:30 18:09 29:14 34:54 39:07
Indy R4600PC/100: 32 00:34 09:43 16:55 17:38 27:51 33:35 38:00
Indy R4600PC/100: 2 00:34 11:06 18:14 18:57 30:40 36:24 40:48
Table B: Detailed Timings for a Default IRIX 6.2 OS Installation
Table C shows the time taken for each individual step in the
installation process, ie. the times obtained by subtracting one
column from the next.
CD PRE- EXIT PRE- EXIT RQSALL
ROM INST CD1 COMS INST CD2 COMS ELF
I2 R4400SC/250: 32 00:15 04:57 02:33 00:20 05:03 02:08 02:46
I2 R4400SC/250: 2 00:15 06:05 02:32 00:19 06:23 02:13 02:53
Indy R4400SC/200: 32 00:20 05:58 03:08 00:27 06:08 02:38 03:09
Indy R4400SC/200: 2 00:20 07:20 03:07 00:27 07:46 03:00 03:11
Indy R4600PC/133: 32 00:32 08:29 06:53 00:41 09:33 05:37 04:13
Indy R4600PC/133: 2 00:32 10:01 06:57 00:39 11:05 05:40 04:13
Indy R4600PC/100: 32 00:34 09:09 07:12 00:43 10:13 05:44 04:25
Indy R4600PC/100: 2 00:34 10:32 07:08 00:43 11:43 05:44 04:24
Table C: Individual times for each installation step (IRIX 6.2)
Table D gives the percentage differences for the paired results, ie.
2X vs. 32X for each system. The times for all steps that don't
involve the CDROM are added together, then the highest is divided by
the lowest for each system pair. The results should be low in order
to be usable:
TOTAL TIMES PERCENTAGE
2X vs. 32X VARIATION
I2 R4400SC/250: 08:12 / 08:02 2.1%
Indy R4400SC/200: 10:05 / 09:42 4.0%
Indy R4600PC/133: 18:01 / 17:56 0.5%
Indy R4600PC/100: 18:33 / 18:38 0.5%
Table D: Paired result differences.
These are satisfactory. In fact, if you look at the individual stage times in Table C, many relevant paired results are identical.
Now for the all-important comparison: the average performance for each system compared to the average performance of the slowest system (Indy R4600PC/100). Thus, the numbers denote how much faster each stage was executed compared to the slowest system (ie. smaller = better). Table 4 showed these figures for each installation stage, but 6.2 produces times that are 50% smaller than 6.5, so errors have a greater effect; thus, the stage times have been added together to reduce the effect of error, the results of which are in Table E.
FACTOR
DIFF
I2 R4400SC/250: 0.44
Indy R4400SC/200: 0.53
Indy R4600PC/133: 0.97
Indy R4600PC/100: 1.00
Table E: Comparison to Indy R4600PC/100
performance (factor differences)
As with installing 6.5, there is little benefit overall in having an R4600PC/133 as opposed to an R4600PC/100, but when one moves to a higher clocked CPU which has some L2 cache, the performance improvement is considerable.
Also, the figures in Table E have a similar profile to the final averages given in Table 4. This shows that the OS installation tasks are:
Remember: these tests involved alot of disk accesses. An integer task which did not access the disk so much would show an even greater improvement in speed for faster systems, eg. ImgLab tests. These results obviously prove a real-world integer task will be faster on better systems, but what's important is that one shouldn't rely on synthetic benchmarks alone when attempting to estimate system performance - the expected improvement may not be as great as one eventually sees if one's task involves disk operations.
NB: the analysis of the tasks which involve accessing the CDROM, as shown in Table C (columns CD1 and CD2), I leave as an exercise for the reader. This is because a complete analysis of CDROM performance during the installation of 6.5 is already available in CDROM TEST 1. The main reason why I have done INTEGER TEST 7 is to show that the analysis given in INTEGER TEST 1 is valid.
INTEGER TEST 8: Compiling Lynx V2.7.1
This test shows compilation times for compiling Lynx V2.7.1, the text-only web browser.
UPDATE: 18/Oct/2005. For a long time I've wanted to redo these tests, using a common compiler version in order to eliminate differences which had nothing to do with system hardware. Finally, I've been able to obtain 6.5.22, so I can run this and other tests on older systems with a common execution library setup. I'll also run the tests with 6.5.26, just to see if there are any significant differences. Thus, at some point soon, the older tables currently shown below will be removed. There will be new tables for MIPS Pro, and separate tables for GCC. In the old tables, adding results compiled with GCC caused confusion because, by default, GCC does little optimisation (as far as I can tell), while MIPS Pro does a lot. Thus, without changing the default Makefiles, the GCC compiler completes this test twice as fast as the MIPS Pro compilers, ie. GCC isn't doing any optimisation calculations whereas MIPS Pro is.
The Makefiles contain the -O option for the SGI compile lines; I'm not sure what this means for GNU, but for MIPS Pro it activates heavy optimisation (-O is the same as -O2 by default). Thus, if one replaces -O with -O0 (no optimisation) in the two Makefiles (the other is in WWW/Library/sgi), the MIPS Pro compilers end up much faster than GNU.
To put it another way, without splitting the results apart, one is comparing GNU compilations using little optimisation to MIPS Pro compilations using heavy optimisation, which isn't a fair comparison. For the same system, the difference can be 100%, eg. for O2 R5K/200, the default -O gives 5:32 (minutes and seconds), compared to 2:14 when -O0 is used. One can't include such results in the same table. So, treat the two tables as non-optimised MIPS Pro vs. heavy-optimised MIPS Pro vs. GNU.
If you want to perform this test yourself, download the archive and then follow these instructions:
If you're using the GNU C compiler, just extract the archive, cd into the lynx-2.7.1 directory and enter:
timex make sgi
The 'real' time is the one to submit. Note that if some elements of SGI's compilers are installed on your system, but no license is present, then the presence of gcc might not be detected due to the use of just "cc" in the Makefiles (the compiler wrapper uses /usr/bin/cc which won't know about gcc in /usr/freeware/bin). If this happens, change "cc" to "gcc" in the relevant Makefiles (see below for details on which Makefiles to alter).
If you're using SGI's MIPS Pro compilers, then you should run the test twice (optimised vs. non-optimised):
Test 1: do exactly the same as described above for GNU: just extract the archive and run the test. The 'real' time is the one to submit for inclusion into Table 32a (heavy optimisation).
Test 2: there are two Makefiles to modify. One is in the lynx-2.7.1 directory, the other is in WWW/Library/sgi. For the first Makefile, load the file into jot, press CTRL-F (for Find), type 'sgi' and press Return. Press CTRL-F once more. You will now see the SGI-related compilation definition. Change the '-O' flag to '-O0' to turn off optimisation. Save the file and exit jot. Now load the second Makefile from WWW/Library/sgi; the 'CFLAGS' line also has a '-O' option, so change it to '-O0', save the file and exit jot. From back within the lynx-2.7.1 directory, enter the same command as above to run the test (ie. 'timex make sgi') and note the 'real' time.
Thus, Table 32 contains non-optimised compilation using MIPS Pro and/or GNU, while Table 32a shows heavy-optimised results with MIPS Pro only. If you do submit a result, please state which compiler was used and, if it was MIPS Pro, whether or not the Makefile -O setting was the default, or changed to -O0. Note that, as explained above, by changing the compiler definition from cc to gcc in the Makefiles, an SGI which normally uses the MIPS Pro compilers can be forced to use the GNU compiler instead (assuming the GNU compiler has been installed).
NB: Newer systems are finishing this test so fast now that the results are likely to be more about disk speed than CPU speed, or at least are affected too much by disk speed. Thus, I'm looking for an alternative test, which will probably consist of compiling stage 1 of GCC. It will be a more carefully controlled test though, with the same disk used each time (at least that's the plan).
Time Credit
(mm:ss) Compiler (if any)
Octane R120000SC 300MHz 2MB L2: 00:55 MIPS Pro 7.3.1.3m I.M. [hinv]
O2 R12000SC 400MHz 4MB L2: 01:04 MIPS Pro 7.3.1.2m S.E. [hinv]
O2 R12000SC 300MHz 1MB L2: 01:23 MIPS Pro 7.3.1.2m I.M. [hinv]
Indy R5000SC 180MHz 512K L2: 03:25 MIPS Pro 7.3.1.2m I.M. [hinv]
Indy R5000SC 150MHz 512K L2: 03:42 MIPS Pro 7.3.1.2m I.M. [hinv]
Indigo R4400SC 150MHz 1MB L2: 05:02 MIPS Pro 7.3.1.2m I.M. [hinv]
Indy R5000PC 150MHz: 05:16 MIPS Pro 7.3.1.2m I.M. [hinv]
Indigo R4000SC 100MHz 1MB L2: 09:48 MIPS Pro 7.3.1.2m I.M. [hinv]
Older results with earlier compiler versions which are not comparable:
Octane R12000SC 400MHz 2MB L2: 00:36 MIPS Pro 7.3 I.M. [hinv]
Onyx2 R10000SC 250MHz 4MB L2: 00:47 MIPS Pro 7.2.1 I.M.
Origin200QC R10000SC 180MHz 2MB L2: 01:06 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Onyx R10000SC 195MHz 1MB L2: 01:16 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Indigo2 R10000SC 195MHz 1MB L2: 01:21 MIPS Pro 7.2.1 INT9/Credit 2
Indigo2 R10000SC 175MHz 1MB L2: ? I.M. [hinv]
Onyx R4400SC 250MHz 4MB L2: 02:06 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
O2 200MHz R5000SC 1MB L2: ? I.M.
Onyx 200MHz R4400SC 4MB L2: 02:30 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Indy R4400SC 200MHz 1MB L2: ?
O2 R10000SC 150MHz 1MB L2: ? I.M. [hinv]
POWER Onyx R8000SC 90MHz 4MB L2: 02:50 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Indigo2 250MHz R4400SC 2MB L2: ?
Indigo2 200MHz R4400SC 1MB L2: ?
O2 R5000SC 180MHz 512K L2: ? I.M. [hinv]
POWER Onyx R8000SC 75MHz 4MB L2: 03:16 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Onyx 150MHz R4400SC 1MB L2: 03:27 MIPS Pro 7.2.1 INT9/Credit 2
POWER Series 100MHz R4400SC 1MB L2: 03:53 MIPS Pro 7.2.1.2m INT9/Credit 2
Indy R4600SC 133MHz 512K L2: 04:01 MIPS Pro 7.2.1.2m INT9/Credit 2
POWER Series 100MHz R4000SC 1MB L2: 07:22 MIPS Pro 7.2.1.2m INT9/Credit 2
Indy R4400SC 200MHz 1MB L2: ?
Table 32: Time to compile Lynx V2.7.1 using MIPS Pro with -O0 in the Makefiles.
Time Credit
(mm:ss) Compiler (if any)
Octane R120000SC 300MHz 2MB L2: 02:12 MIPS Pro 7.3.1.3m I.M. [hinv]
O2 R12000SC 400MHz 4MB L2: 02:30 MIPS Pro 7.3.1.2m S.E. [hinv]
O2 R12000SC 300MHz 1MB L2: 03:36 MIPS Pro 7.3.1.2m I.M. [hinv]
Indigo2 R10000SC 195MHz 1MB L2: 03:43 MIPS Pro 7.3.1.2m I.M. [hinv]
Indy R5000SC 180MHz 512K L2: 08:18 MIPS Pro 7.3.1.2m I.M [hinv]
Indy R5000SC 150MHz 512K L2: 08:44 MIPS Pro 7.3.1.2m I.M [hinv]
Indigo R4400SC 150MHz 1MB L2: 11:29 MIPS Pro 7.3.1.2m I.M. [hinv]
Indy R5000PC 150MHz: 12:07 MIPS Pro 7.3.1.2m I.M [hinv]
Indigo R3000 33MHz: 15:57 MIPS Pro 5.3 I.M. [3] [hinv]
Indigo R4000SC 100MHz 1MB L2: 18:10 MIPS Pro 7.3.1.2m I.M. [hinv]
Older results with earlier compiler versions which are not comparable:
Octane R12000SC 400MHz 2MB L2: 01:30 MIPS Pro 7.3 I.M. [hinv]
Onyx2 R10000SC 250MHz 4MB L2: 01:55 MIPS Pro 7.2.1 [2] I.M.
Origin2000 300MHz R12000SC 8MB L2: 02:02 MIPS Pro 7.3b [1] John McCalpin
Onyx2 195MHz R10000SC 4MB L2: 02:32 MIPS Pro ? [hinv]
Origin200QC R10000SC 180MHz 2MB L2: 02:39 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Onyx R10000SC 195MHz 1MB L2: 03:04 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Origin200 R10000SC 180MHz 1MB L2: 03:15 MIPS Pro 7.2.1.2m INT9/Credit 2 [hinv]
Indigo2 R10000SC 195MHz 1MB L2: 03:17 MIPS Pro 7.2.1 INT9/Credit 2
Octane R10000SC 175MHz 1MB L2: 03:28 MIPS Pro 7.2.1.2m INT9/Credit 2 [hinv]
Indigo2 R100000SC 175MHz 1MB L2: 04:04 MIPS Pro 7.3 I.M. [hinv]
Onyx R4400SC 250MHz 4MB L2: 04:53 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
O2 R10000SC 175MHz 1MB L2: 04:59 MIPS Pro 7.2.1.2m INT9/Credit 2 [hinv]
O2 200MHz R5000SC 1MB L2: 05:23 MIPS Pro 7.2.1 I.M.
Onyx 200MHz R4400SC 4MB L2: 05:41 MIPS Pro 7.2.1.2m INT9/Credit 2 [hinv]
Indigo2 250MHz R4400SC 2MB L2: 05:42 MIPS Pro 7.2.1 I.M.
O2 R10000SC 150MHz 1MB L2: 06:15 MIPS Pro 7.3 I.M. [hinv]
POWER Onyx R8000SC 90MHz 4MB L2: 06:35 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
Indigo2 250MHz R4400SC 2MB L2: 06:46 MIPS Pro 7.3
Indigo2 200MHz R4400SC 1MB L2: 07:19 MIPS Pro 7.2.1 I.M.
O2 R5000SC 180MHz 512K L2: 07:27 MIPS Pro 7.3 I.M. [hinv]
Indy 200MHz R4400SC 1MB L2: 07:31 MIPS Pro 7.2.1 I.M.
POWER Onyx R8000SC 75MHz 4MB L2: 07:52 MIPS Pro 7.2.1 INT9/Credit 2 [hinv]
POWER Series 150MHz R4400SC 1MB L2: 08:10 MIPS Pro 7.2.1.2m INT9/Credit 2
Onyx 150MHz R4400SC 1MB L2: 08:13 MIPS Pro 7.2.1 INT9/Credit 2
Indy R4600SC 133MHz 512K L2: 09:42 MIPS Pro 7.2.1.2m INT9/Credit 2
Indy R4600SC 133MHz 512K L2: 10:22 MIPS Pro 7.2.1 I.M.
POWER Series 100MHz R4400SC 1MB L2: 11:04 MIPS Pro 7.2.1.2m INT9/Credit 2
POWER Series 100MHz R4000SC 1MB L2: 12:10 MIPS Pro 7.2.1.2m INT9/Credit 2
Indy 133MHz R4600PC: 15:59 MIPS Pro 7.2.1 I.M.
Indy 100MHz R4600PC: 16:30 MIPS Pro 7.2.1 I.M.
Table 32a: Time to compile Lynx V2.7.1 using MIPS Pro Compilers (-O2).
Time Credit
(mm:ss) (if any)
Indigo2 R10000SC 175MHz 1MB L2: 02:02 I.M. [hinv]
O2 200MHz R5000SC 1MB L2: 02:50 I.M.
Indigo2 250MHz R4400SC 2MB L2: 03:07 I.M.
Indy R5000SC 180MHz 512K L2: 03:45 I.M.
Indigo2 R4400SC 150MHz 1MB L2: 05:25 I.M. [hinv]
Indy R5000PC 150MHz: 05:26 I.M.
Indy R4600SC 133MHz 512K L2: 05:32 I.M.
Indigo R4000SC 100MHz 1MB L2: 08:48 I.M. [hinv]
Table 32b: Time to compile Lynx V2.7.1 using GNU Compiler (GCC V2.8.1).
[1] System had an old slow disk, using Beta 7.3 compilers.
[2] System used a modern 18GB USCSI disk.
[3] A number of modules were compiled without optimisation due to certain
values exceeding limits for the older 5.3 compiler.
This discussion applies to results from a particular table, but
don't compare times between two different tables.
Many other tests I've done show how the lack of L2 cache can hurt the performance of R4600 Indys, but this example reveals the effect very clearly. For example, if the 200MHz Indy had no L2 cache, its time would be more like 10 or 11 minutes, so the 1MB L2 gives an extra 30% or 40% speed increase for this test. Thus, if you ever have to use or are considering buying an R4600 Indy, I strongly recommend you go for the SC version rather than the PC version if you can (the R4600SC 133MHz CPU has 512K L2 cache).
Another observation worthy of note is the better speed from the R10K O2 compared to the R5K O2; even at a lower clock speed, the R10K O2 is faster. R10K O2s are good for integer processing. In fact, probably by the vaguaries of compiler optimising (chance variance), the SPECint95 data shows an R10K/250 O2 actually beating an R10K/250 Origin2000 for the m88ksim test. In other words, for those cases where the bottleneck is not memory bandwidth, memory latency or L2/RAM cache access issues, an R10K/R12K O2 is a good choice (this applies to integer processing only; fp is a different story). Note that this is partly why R10K O2s are faster for 3D graphics than an R5K O2 at the same [similar] clock speed: as an SGI engineer explained, 3D graphics involves integer operations such as array accesses and pointer chasing - the R10K will be faster than R5K for these operations. Thus, at the same clock speed, R10K O2s offer around 20% better performance than R5K O2s for 3D graphics.
The run times for the Lynx test are getting quite low now with newer systems (eg. the Origin2000 R12K/300 is probably being limited by disk speed, not CPU speed), so I'm on the lookout for a tougher test. Any suggestions? However, it's not easy choosing what package to use as a compilation test; sometimes a compilation involves a configure script which can hide the possibility that different systems are not compiling in the same way (eg. modules or plugins installed on one system but not on another). This is why I'm not going to use GIMP as a compilation test. One possibility is GNU, which definitely takes a long time to finish.
INTEGER TEST 9: RC5-64 and DES Encryption-cracking Performance
Benchmark
There is an ongoing project to develop a distributed computer system for cracking the RC5 and DES encryption algorithms. The purpose of the project is to prove that brute-force computation can break the encryption, so the project organisers have enlisted the help of anyone on the Internet who has some compute cycles to spare. The combined computational power of the entire collection of machines in enormous; as a result, DES can be cracked in a matter of hours.
Rather usefully for benchmarking purposes, the software can be downloaded for any platform and run in a 'benchmark' mode, in which 10 million RC5-64 tests are executed along with 20 million DES tests (or 'keys') - the numbers of keys processed per second for RC5-64 and DES are the benchmark results (NB: RC5-64 is a more complex problem and so gives lower numbers). Running the benchmark test is simple: download the pre-compiled software for the target system, unpack the archive and run the main program in benchmark mode. For example, on an Indy, the command is:
rc5des-mips3-32bit -benchmark
A typical output, for example from a 200MHz R4400SC Indy, looks like this:
Benchmarking RC5 with 10000000 tests: .....10%.....20%.....30%.....40%.....50%.....60%.....70%.....80%.....90%.... Completed in 0.00:01:34.63 [105674.20 keys/sec] Benchmarking DES with 20000000 tests: .....10%.....20%.....30%.....40%.....50%.....60%.....70%.....80%.....90%.... Completed in 0.00:00:31.07 [674831.50 keys/sec]
To keep the results intuitively meaningful, one often divides the results by 1000, to give kilokeys/sec. Note: at present, I am not entirely sure that these tests perform integer-only calculations. I shall check on this, but in the meantime I'll include RC5/DES as an integer test - more details below.
Table 34 shows the results (note carefully how you form your initial judgements based on the numbers shown in this table). These results are not multi-threaded, ie. a multi-CPU system only uses one CPU for this test (any multi-CPU system in the table includes a note in curly brackets {} showing the number of CPUs in the system, even though only one is used, just for reference).
Clock RC5-64 DES
System CPU (MHz) L2 (Kkeys/sec) (Kkeys/sec)
Fuel R16000SC 700 4MB 1133.78 10933.22 I.M. [hinv]
Fuel R14000SC 500 2MB 782.20 7482.29 I.M. [hinv]
Octane R12000SC 400 2MB 645.91 6227.00 I.M. [hinv]
Origin200 R12000SC 360 4MB 585.44 5641.78 I.M. [hinv]
O2 R12000SC 400 4MB 581.94 2973.82 Stefan E. [hinv]
O2 R12000SC 400 2MB 557.70 2929.73 Colin Anderson [hinv]
Octane R12000SC 300 2MB 467.59 4489.21 I.M. [hinv]
Origin200 R12000SC 270 4MB 438.92 4229.23 I.M. [hinv]
O2 R12000SC 300 1MB 434.37 2213.72 I.M. [hinv]
O2 R7000SC 300 256K 395.80 1787.09 Stefan E. [hinv]
Onyx2 R10000SC 250 4MB 395.76 3827.72 I.M. {14} [hinv]
Octane R10000SC 250 1MB 384.17 3720.62 {2}
Onyx R10000SC 195 1MB 312.95 2971.11 [Credit 2] [hinv]
Onyx2 R10000SC 195 4MB 307.63 3006.16 {12} [hinv]
Indigo2 R10000SC 195 1MB 306.93 2979.51 [Credit 2]
Octane R10000SC 195 1MB 288.26 2803.58 I.M.
Origin200QC R10000SC 180 2MB 286.59 2787.29 [Credit 2] {2} [hinv]
O2 R10000SC 195 1MB 276.56 1443.21 I.M. [hinv]
Indigo2 R10000SC 175 1MB 271.45 2652.31 I.M. [hinv]
Origin200 R10000SC 180 1MB 257.73 1250.00 [Credit 2] {2} [hinv]
O2 R10000SC 175 1MB 244.32 1209.19 [Credit 2] [hinv]
Octane R10000SC 175 1MB 234.47 1160.77 [Credit 2] [hinv]
O2 R5200SC 300 1MB 226.92 1045.00 Stefan E. [hinv]
O2 R5000SC 250 1MB 188.64 869.36 Stefan E. [hinv]
O2 R5000SC 200 1MB 141.82 652.33 I.M.
Onyx R4400SC 250 4MB 138.88 1609.56 [Credit 2] {4} [hinv]
Indy R5000SC 180 512K 134.40 617.85 [Credit 1]
Indigo2 R4400SC 250 2MB 133.03 851.04 I.M.
O2 R5000SC 180 512K 131.43 605.61 I.M. [hinv]
POWER Onyx R8000SC 90 4MB 129.79 1101.54 [credit 2] {2} [hinv]
Indy R5000SC 150 512K 113.07 518.02 I.M. [hinv]
Onyx R4400SC 200 4MB 110.87 1289.13 [Credit 2] {4} [hinv]
Indy R5000PC 150 - 110.32 506.13 I.M.
Indigo2 R4400SC 200 1MB 106.45 672.71 I.M.
POWER Onyx R8000SC 75 4MB 107.96 906.43 [credit 2] {2} [hinv]
Indy R4400SC 200 1MB 105.67 674.83 I.M.
Indy R4600SC 133 512K 97.66 442.76 [Credit 2]
Indy R4600PC 133 - 92.71 408.47 I.M.
Onyx R4400SC 150 1MB 83.12 968.46 [Credit 2]
Crimson R4400SC 150 1MB 81.66 522.75 [Credit 2] [hinv]
Indigo R4400SC 150 1MB 79.17 503.75 I.M. [hinv]
Indigo2 R4400SC 150 1MB 78.46 501.14 I.M. [hinv]
Indy R4600PC 100 - 69.17 305.48 I.M.
POWER Series R4400SC 100 1MB 54.05 345.92 [Credit 2] [hinv]
POWER Series R4000SC 100 1MB 32.67 282.19 [Credit 2] [hinv]
Indigo R4000SC 100 1MB 31.10 267.65 I.M. [hinv]
Indigo2 R4000SC 100 1MB 30.89 265.06 I.M. [hinv]
POWER Series R3000SC 33 256K 27.67 118.10 [Credit 2] {6} [hinv]
POWER Series R3000SC 25 256K 20.50 89.13 [Credit 2] {2} [hinv]
Table 34: RC5-64/DES v2.7100.415 Benchmark Test Results (Kkeys/sec)
Larger L2 cache sizes seems to aid the DES test, eg. the R4400SC/250 Indigo2 DES result. Another example: Onyx2 R10K/195 (4MB) is a bit faster than Octane R10K/195 (1MB).
Either way, here is the same information as Table 34, but this time the results are shown as elapsed times for each system, rounded to the nearest second, in order of best RC5-64 scores:
RC5-64 DES TOTAL
(mm:ss) (mm:ss) (mm:ss)
Fuel R16000SC 700MHz 4MB L2: 00:09 00:02 00:11 [hinv]
Fuel R14000SC 500MHz 2MB L2: 00:13 00:03 00:16 [hinv]
Octane R12000SC 400MHz 2MB L2: 00:16 00:03 00:19 [hinv]
O2 R12000SC 400MHz 4MB L2: 00:17 00:07 00:24 [hinv]
O2 R12000SC 400MHz 2MB L2: 00:17 00:07 00:24 [hinv]
Origin200 R12000SC 360MHz 4MB L2: 00:17 00:04 00:21 [hinv]
Octane R12000SC 300MHz 2MB L2: 00:21 00:05 00:26 [hinv]
Origin200 R12000SC 270MHz 4MB L2: 00:23 00:05 00:28 [hinv]
O2 R12000SC 300MHz 1MB L2: 00:23 00:09 00:32 [hinv]
Onyx2 R10000SC 250MHz 4MB L2: 00:25 00:05 00:30 [hinv]
O2 R7000SC 300MHz 256K L2: 00:25 00:12 00:37 [hinv]
Octane R10Kx2 250MHz 1MB L2: 00:26 00:06 00:32
Onyx R10000SC 195Mhz 1MB L2: 00:32 00:07 00:39 [hinv]
Indigo2 R10000SC 195MHz 1MB L2: 00:33 00:07 00:40
Octane R10000SC 195MHz 1MB L2: 00:35 00:07 00:42
Origin200QC R10000SC 180MHz 2MB L2: 00:35 00:08 00:43 [hinv]
O2 R10000SC 195 1MB L2: 00:36 00:15 00:51 [hinv]
Indigo2 R10000SC 175MHz 1MB L2: 00:37 00:08 00:45 [hinv]
Origin200 R10000SC 180MHz 1MB L2: 00:39 00:16 00:55 [hinv]
O2 R10000SC 175MHz 1MB L2: 00:41 00:17 00:58 [hinv]
Octane R10000SC 175MHz 1MB L2: 00:43 00:17 01:00 [hinv]
O2 R5200SC 300MHz 1MB L2: 00:44 00:20 01:04 [hinv]
O2 R5000SC 250MHz 1MB L2: 00:53 00:24 01:17 [hinv]
O2 R5000SC 200MHz 1MB L2: 01:11 00:32 01:43
Onyx2 R4400SC 250MHz 4MB L2: 01:12 00:13 01:25 [hinv]
I2 R4400SC 250MHz 2MB L2: 01:15 00:25 01:40
Indy R5000SC 180MHz 512K L2: 01:15 00:34 01:49
O2 R5000SC 180MHz 512K L2: 01:16 00:35 01:51 [hinv]
POWER Onyx R8000SC 90MHz 4MB L2: 01:17 00:19 01:36 [hinv]
Indy R5000SC 150MHz 512LK L2: 01:28 00:40 02:08 [hinv]
Onyx R4400SC 200MHz 4MB L2: 01:30 00:16 01:46 [hinv]
Indy R5000PC 150MHz: 01:31 00:41 02:12
Indigo2 R4400SC 200MHz 1MB L2: 01:34 00:31 02:05
Indy R4400SC 200MHz 1MB L2: 01:34 00:31 02:05
Indy R4600SC 133MHz 512K L2: 01:43 00:47 02:30
Indy R4600PC 133MHz: 01:48 00:51 02:49
Crimson R4400SC 150MHz 1MB L2: 02:02 00:40 02:42 [hinv]
Indigo R4400SC 150MHz 1MB L2: 02:06 00:42 02:48 [hinv]
Indigo2 R4400SC 150MHz 1MB L2: 02:07 00:42 02:49 [hinv]
Indy R4600PC 100MHz: 02:25 01:09 03:34
POWER Series R4400SC 100MHz 1MB L2: 03:05 01:00 04:05 [hinv]
POWER Series R4000SC 100MHz 1MB L2: 05:07 01:14 06:21 [hinv]
Indigo R4000SC 100MHz 1MB L2: 05:22 01:18 06:40 [hinv]
Indigo2 R4000SC 100MHz 1MB L2: 05:24 01:19 06:43 [hinv]
POWER Series R3000SC 33MHz 256K: 06:01 02:58 08:59 [hinv]
POWER Series R3000SC 25MHz 256K: 08:08 03:55 12:03 [hinv]
Table 35: RC5-64/DES Benchmark Test Results (Times)
Another two items to consider, points which in a way contradict each other:
timex ./rc5des-mips4-32bit -benchmark
and this command for Indigo2:
timex ./rc5des-mips3-32bit -benchmark
Here are the results, plus data for other systems, in order of real (total) time:
user real sys
(mm:ss.ss) (mm:ss.ss) (s.ss)
Fuel R16000SC 700MHz 4MB L2: 00:10.70 00:10.95 0.02 [hinv]
Fuel R14000SC 500MHz 2MB L2: 00:15.00 00:15.63 0.09 [hinv]
Octane R12000SC 400MHz 2MB L2: 00:18.67 00:18.92 0.02 [hinv]
Origin200 R12000SC 360MHz 4MB L2: 00:20.77 00:21.05 0.02 [hinv]
O2 R12000SC 400MHz 2MB L2: 00:23.73 00:24.50 0.03 [hinv]
Octane R12000SC 3000MHz 2MB L2: 00:25.04 00:26.09 0.17 [hinv]
Origin200 R12000SC 270MHz 4MB L2: 00:27.69 00:27.90 0.02 [hinv]
Onyx2 R10000SC 250MHz 4MB L2: 00:30.63 00:30.76 0.03 [hinv]
Octane R10000SC 250MHz 1MB L2: 00:30.76 00:32.22 0.14
O2 R12000SC 300MHz 1MB L2: 00:31.83 00:32.53 0.08 [hinv]
Onyx R10000SC 195MHz 1MB: 00:38.93 00:39.06 0.03 [hinv]
Indigo2 R10000SC 195MHz 1MB L2: 00:39.16 00:39.68 0.06
Octane R10000SC 195MHz 1MB L2: 00:39.38 00:42.36 0.70
Origin200QC R10000SC 180MHz 2MB L2: 00:42.32 00:42.44 0.03 [hinv]
Indigo2 R10000SC 175MHz 1MB L2: 00:43.69 00:45.65 0.08 [hinv]
O2 R10000SC 195 1MB L2: 00:49.58 00:50.74 0.14 [hinv]
Origin200 R10000SC 180MHz 1MB L2: 00:53.87 00:55.64 0.15 [hinv]
O2 R10000SC 175MHz 1MB L2: 00:55.58 00:57.56 0.14 [hinv]
Octane R10000SC 175MHz 1MB L2: 00:55.57 00:59.90 0.58 [hinv]
Onyx R4400SC 250MHz 4MB L2: 01:24.73 01:25.11 0.05 [hinv]
I2 R4400SC 250MHz 2MB L2: 01.37.77 01:39.87 0.17
O2 R5000SC 200MHz 1MB L2: 01:34.65 01:42.77 0.28
POWER Onyx R8000SC 90MHz 4MB L2: 01:35.49 01:36.25 0.10 [hinv]
Onyx R4400SC 200MHz 4MB L2: 01:45.94 01:46.63 0.08 [hinv]
Indy R5000SC 180MHz 512K L2: 01:45:30 01:48.83 0.30
O2 R5000SC 180MHz 512K L2: 01:45.20 01:50.75 0.31 [hinv]
POWER Onyx R8000SC 75MHz 4MB L2: 01:54.66 01:55.84 0.14 [hinv]
Indigo2 R4400SC 200MHz 1MB L2: 02:02.41 02:05.16 0.26
Indy R4400SC 200MHz 1MB L2: 02:02.97 02:06.35 0.41
Indy R5000SC 150MHz 512K L2: 02:06.16 02:09.06 0.25 [hinv]
Indy R5000PC 150MHz: 02:12.15 02:07.52 0.62
Indy R4600SC 133MHz 512K L2: 02:26.03 02:30.44 0.50
Indy R4600PC 133MHz: 02:24.83 02:32.36 0.62
Indigo R4400SC 150MHz 1MB 02:43.43 02:48.09 0.44 [hinv]
Indigo2 R4400SC 150MHz 1MB L2: 02:43.65 02:49.41 0.42 [hinv]
Crimson R4400 150MHz 1MB: 02:51.54 02:52.66 0.19 [hinv]
Indy R4600PC 100MHz: 03:20.84 03:33.14 1.53
POWER Series R4400SC 100MHz 1MB L2: 04:02.42 04:05.81 0.46 [hinv]
POWER Series R4000SC 100MHz 1MB L2: 06:15.53 06:20.84 0.68 [hinv]
Indigo R4000SC 100MHz 1MB L2: 06:23.73 06:40.09 1.56 [hinv]
Indigo2 R4000SC 100MHz 1MB L2: 06:23.68 06:42.95 1.74 [hinv]
POWER Series R3000 33MHz 256K L2: 09:04.15 09:10.20 2.93 [hinv]
POWER Series R3000 25MHz 256K L2: 11:49.01 12:03.47 4.98 [hinv]
Table 36: Timex output for running RC5-64/DES.
Credits:
[1] - Martin Doll, Ph. D. (mkhd@into.ch), Institute of Organic Chemistry Univerisity of Zurich Wintherthurerstr. 157 CH-8057 Zurich.
[2] - Simon Pigot (simon@dpiwe.tas.gov.au), Parks and Wildlife GIS Unit, Tasmania. Supplied hinv of the POWER Series, Octane, O2, Origin200, Origin200QC and Onyxs:
GRAPHICS TEST 1: Inventor Objects: rotation speed
(single-buffered).
These tests give the frames-per-second rates whilst spinning various 3D Inventor objects. Some of the models are my own, created over the years; others are from various SGI demos, eg. the 'Underwater Camera' and 'Huge Engine Model'.
NOTE: The frame rate figures given here are for single-buffered tests, ie. no attempt is made to synchronise the drawing of the 3D image with the display refresh of the monitor. Thus, a system's graphics hardware renders a model as quickly as possible, which means observed frame rates will usually be an obscure fraction of the monitor refresh rate, but can also be many times the refresh rate if a particular graphics system is sufficiently powerful (eg. Onyx2) or if the model is sufficiently simple (ag. the arm2.iv model data shown below).
By contrast, double-buffered tests (see GRAPHICS TEST 2 below) involve waiting for a monitor refresh event to synchronise the drawing of the object with the frame-by-frame display of pixel data on the screen. Thus, frame rates for double-buffered tests can never be more than the monitor refresh rate, are usually an integer divisor of the monitor rate, and frequently do not represent the maximum capabilities of a graphics system. I present results of both types of test here in order to offer as broad a range of information as possible, and also to show very clearly how different these two types of test can be. For full details on this issue, and I strongly recommend you read this information before looking at the results below, see my page on The Effects of Double-Buffering on Animation Frame Rates. I dedicate most effort to single-buffered tests though since they are far more meaningful.
The objects I've used for these tests vary in their nature, but they do share one aspect in common: they're not 'perfect' objects, ie. they're often inefficient in their definition. This is entirely deliberate, because in the real world that is exactly how many 3D object files are constructed, eg. a landscape data set may have a huge number of triangles wastefully used to represent a flat area, but someone working with the model might not be allowed to alter it, or may not know how even if they could. Thus, although I always use the command 'ivview -p' to run these tests, I never use ivview's object-optimisation options. Sometimes of course, it's possible to use software such as Optimizer to improve performance, but I don't deal with such issues here.
For each object, a brief description, the number of polygons it contains and a link to a screenshot are provided. Feel free to browse the screenshots page, which has most of the object models available for download.
Caveats:
MWin (MW) = Maximum Window (click top-right button)
SWin (SW) = Small Window (default size of ivview)
/xx = xx is the rendering mode; if none is given then the mode is
'default', ie. texturing is ON if the model contains texture.
Otherwise, see SH, W, T or HL below.
SH = shaded mode
W = wireframe mode
T = transparent
HL = hidden-line mode
Test A: Underwater Camera (42650 polygons). This model [screenshot] is from SGI's own
demonstration collection. High-quality transparency was turned on for
this test.
MWin MWin/SH MWin/W SWin SWin/SH SWin/W
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 133.47 134.06 82.75 133.33 133.80 82.75 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 97.00 183.14 76.81 103.47 215.25 86.42 [hinv]
Fuel R14K/500 2MB L2 V10: 98.62 194.61 76.13 98.46 194.50 75.17 [hinv]
Fuel R16K/700 4MB L2 V10: 96.24 182.05 73.55 109.95 263.64 86.04 [hinv]
Octane R12000SC/300 2MB L2 V6: 82.30 123.64 50.90 91.45 146.37 67.81 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 77.46 77.87 61.15 77.48 78.10 62.17 [hinv]
Octane2 R12000 400MHz 2MB L2 SSE: 5.93 57.25 34.15 14.30 65.60 36.52 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 47.96 55.71 27.27 53.68 63.53 28.56 [hinv]
Octane R12000SC/300 2MB L2 SSE: 3.98 53.97 32.56 10.33 63.20 35.20 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 49.75 53.74 27.55 51.24 54.88 28.33 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 47.66 49.15 26.14 48.66 51.36 27.91 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 46.16 47.01 26.95 46.91 47.98 28.09 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 40.66 40.63 26.14 40.24 41.21 27.45 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 34.77 34.86 24.90 35.07 35.21 27.17 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 34.69 35.10 27.01 35.07 35.18 28.26 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 33.50 38.00 23.30 35.50 40.43 24.11 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 32.42 36.41 23.07 34.42 39.00 23.99 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 28.88 35.53 20.71 31.59 36.07 22.09 [hinv]
POWER Series R4400SC/150 RE 2RM: 31.44 32.54 21.29 31.63 32.65 21.30 [hinv]
POWER Series R4400SC/150 RE 4RM: 31.46 32.37 20.80 31.30 32.37 21.27 [hinv]
POWER Series R4400SC/150 RE 1RM: 31.27 32.39 20.33 31.82 32.65 21.65 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 30.44 34.45 22.28 32.70 36.79 23.08
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 27.67 29.23 22.25 29.06 30.82 23.09 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 24.94 30.13 13.94 28.04 34.95 14.62
Octane R12000SC/400 2MB L2 SE+Texture: 24.24 29.31 16.21 26.93 33.93 17.17
Octane R10000SC/195 1MB L2 SE+Texture: 22.89 27.53 13.59 25.66 31.78 12.72
Indigo2 R4400SC/200 2MB HighIMPACT-AA 1MB: 21.98 24.98 15.90 24.35 28.36 16.10 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT: 21.18 24.35 15.53 23.76 28.05 16.60
Indigo2 R10K/175 1MB L2 HighIMPACT 4MB: 23.64 23.80 13.33 24.11 28.13 14.33
Octane R10000 175MHz 1MB L2 SI+Texture: 20.28 23.66 13.47 22.37 26.60 14.27
O2 R12000SC 300MHz 1MB: 19.70 23.03 19.08 21.57 25.71 21.75 [hinv]
POWER Series R4000SC/100 RE 1RM: 19.53 19.62 19.75 19.52 19.64 20.51
POWER Series R4000SC/100 RE 2RM: 19.33 19.50 19.60 19.45 19.54 20.50
O2 R12000SC 270MHz 1MB: 17.59 20.54 18.04 18.61 23.15 20.00 [hinv]
O2 R5200SC 300MHz 1MB: 15.57 18.53 20.28 16.58 20.03 23.76 [hinv]
Octane R12000SC/300 2MB L2 SE: 3.84 31.67 18.82 10.17 39.53 20.64 [hinv]
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 14.64 14.90 18.79 15.03 15.19 19.35 [hinv]
O2 R10000SC 195MHz 1MB: 13.66 16.38 18.05 14.33 17.47 20.69 [hinv]
O2 R5000SC 200MHz 1MB: 11.98 13.70 15.84 11.33 14.47 17.08
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 2.54 27.39 15.21 6.56 30.68 16.05 [hinv]
O2 R10000SC 150MHz 1MB: 10.67 12.74 14.47 11.02 13.29 15.87 [hinv]
Indigo2 R10000 175MHz 1MB L2 SolidIMPACT: 2.26 26.70 15.14 5.50 29.93 15.81 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 1.48 23.92 14.39 2.80 28.55 15.70 [hinv]
O2 R5000SC 180MHz 512K: 9.28 11.35 13.83 9.40 11.87 15.04
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 2.11 11.25 12.43 4.06 11.63 13.67 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 1.68 10.74 12.04 3.55 11.15 13.30 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 1.45 10.75 12.02 2.84 11.04 13.11 [hinv]
POWER Series R3000SC/33 RE 1RM: 5.09 5.13 7.80 5.10 5.13 8.11
Indy R5000SC 180MHz 512K XL24: 3.02 4.80 3.80 4.83 6.54 4.98
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 1.80 5.78 6.03 3.25 5.94 6.43
Indy R5000SC 180MHz 512K XZ (4GE): 1.96 5.73 6.00 3.68 5.90 6.36 [hinv]
Indy R4400SC 200MHz 1MB XZ (4GE): 1.53 5.72 5.92 2.60 5.90 6.40
Indy R4600SC 133MHz 512K L2 XZ (4GE): 1.32 5.67 5.94 2.27 5.84 6.31 [hinv]
Indy R5000PC 150MHz XZ (4GE): 1.34 5.58 5.87 2.85 5.78 6.24 [hinv]
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 1.29 5.75 5.84 2.13 5.98 6.38
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 0.79 5.54 5.82 1.54 5.66 6.17 [hinv]
Indy R4600PC 133MHz XZ (4GE): 0.90 5.36 5.65 1.63 5.56 5.96 [hinv]
Indigo2 R4400SC/250 2MB XL24: 2.66 3.70 3.79 3.69 5.40 4.83
Indy R5000PC 150MHz XL24: 2.37 3.67 3.08 3.79 5.01 4.06
Indigo R3000 33MHz GR2-Elan (4GE): 0.35 4.91 5.60 0.62 5.00 5.89 [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 1.39 3.85 4.14 2.47 3.95 4.43
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 1.22 3.80 4.21 2.08 3.87 4.37 [hinv]
Indigo R4400SC/150 1MB GR2-XS24Z (1GE): 0.98 1.85 2.05 1.58 1.84 2.08 [hinv]
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 0.61 1.83 2.04 1.10 1.84 2.06 [hinv]
Indy R4400SC/200 1MB XL24: 2.12 3.40 3.13 3.17 4.50 3.93
Indy R4600SC/133 512K L2 XL8: 1.45 2.39 2.49 2.00 2.94 3.09
Indy R4600PC/133 XL24: 1.18 1.97 1.89 1.41 2.37 2.34
Indy R4600PC/100 XL24: 0.88 1.59 1.70 1.34 2.06 2.15
Indigo R3000 33MHz XS24Z (1GE): 0.32 1.80 2.01 0.56 1.81 2.02 [hinv]
Indigo R3000 33MHz Entry: 0.52 0.76 0.69 0.64 0.86 0.81 [hinv]
Test B: Huge Engine Model (111638 polygons). This model [screenshot] is also from SGI's own
demonstration collection. High-quality transparency has a different
effect on this model and so is shown as a separate metric.
MW MW/SH MW/W MW/T SW SW/SH SW/W SW/T
Fuel R16K/700 4MB L2 V10: 21.04 40.93 14.82 ? 21.48 49.72 16.59 ? [hinv]
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 22.19 23.55 7.01 41.67 23.00 23.63 7.63 42.18 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 13.53 13.75 7.53 24.01 13.51 13.76 7.35 24.90 [hinv]
Fuel R14K/500 2MB L2 V10: 10.75 40.00 15.42 20.71 10.96 42.80 16.63 20.89 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 10.74 32.37 15.40 20.93 10.95 35.01 16.64 21.00 [hinv]
Octane R12000SC/300 2MB L2 V6: 10.46 23.37 9.03 19.31 10.88 25.95 11.93 19.81
Octane2 R12000SC/400 2MB L2 MXE: 7.62 11.68 4.03 13.90 7.94 7.94 12.48 15.30 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 6.62 8.91 3.01 12.59 6.70 9.30 3.12 12.69 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 6.61 8.51 3.04 12.43 6.69 8.69 3.11 12.57 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 6.57 8.85 3.04 12.53 6.65 9.15 3.11 12.57 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 6.55 7.31 3.04 12.18 6.59 7.22 3.11 12.18 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 6.19 6.34 3.05 11.18 6.20 6.34 3.11 11.21 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 6.06 6.23 2.86 11.02 6.13 6.27 3.08 11.09 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 5.37 7.40 2.94 10.13 5.50 7.63 2.98 10.47 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 5.34 7.33 2.94 10.10 5.46 7.51 2.97 10.46 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 5.10 6.34 2.94 9.53 5.22 6.49 2.99 9.75 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 4.75 5.46 2.94 8.75 4.84 5.57 2.97 8.88 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 4.48 6.25 2.94 8.55 4.70 6.39 2.98 8.78
POWER Series R4400SC/150 RE 4RM: 4.53 5.93 2.39 8.38 4.54 5.96 2.41 8.42 [hinv]
POWER Series R4400SC/150 RE 2RM: 4.52 5.93 2.39 8.38 4.54 5.95 2.41 8.42 [hinv]
POWER Series R4400SC/150 RE 1RM: 4.46 5.73 2.34 8.30 4.54 5.90 2.41 8.42 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 3.58 6.05 2.00 6.80 3.71 6.55 2.05 7.22
Octane R12000SC/400 2MB L2 SE+Texture: 3.55 5.90 2.29 6.78 3.67 6.28 2.33 7.14
Indigo2 R4400SC/200 2MB HighIMPACT-AA 1MB: 3.30 4.31 2.50 6.42 3.39 4.45 2.54 6.42 [hinv]
Octane R10000SC/195 1MB L2 SE+Texture: 3.28 5.50 1.76 6.20 3.39 5.86 1.81 6.54
Indigo2 R4400SC/250 2MB L2 HighIMPACT: 3.02 4.83 1.99 5.73 3.08 5.06 2.02 5.90
POWER Series R4000SC/100 RE 2RM: 3.47 3.49 2.38 6.29 3.46 3.52 2.39 6.22
POWER Series R4000SC/100 RE 1RM: 3.45 3.53 2.34 6.29 3.48 3.55 2.39 6.29
Octane R10000/175 1MB L2 SI+Tex: 2.88 4.47 1.80 5.55 2.96 4.77 1.84 5.68
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 2.55 2.70 2.81 4.71 2.59 2.70 2.81 4.72 [hinv]
O2 R12000SC 400Mhz 4MB: 2.48 [hinv]
O2 R12000SC 300MHz 1MB: 2.14 3.59 2.72 4.18 2.21 3.75 2.93 4.32 [hinv]
O2 R12000SC 270MHz 1MB: 1.92 3.18 2.50 3.70 1.92 3.30 2.67 3.80 [hinv]
O2 R7000SC 300MHz 256K: 1.67 [hinv]
O2 R5200SC 300MHz 1MB: 1.60 2.87 3.16 3.24 1.63 2.86 3.37 3.16 [hinv]
O2 R10000SC 195MHz 1MB: 1.34 2.33 2.73 2.61 1.35 2.39 2.95 2.65 [hinv]
POWER Series R3000SC/33 RE 1RM: 1.67 1.77 2.25 2.92 1.67 1.77 2.31 2.92
O2 R5000SC 250MHz 1MB: 1.18 [hinv]
O2 R10000SC 150MHz 1MB: 1.07 1.84 2.06 2.09 1.08 1.86 2.15 2.09 [hinv]
O2 R5000SC 200MHz: 1.05 2.01 2.37 2.05 1.06 2.05 2.46 2.07
O2 R5000SC 180MHz: 0.90 1.64 2.06 1.76 0.91 1.66 2.16 1.78
Octane2 R12000 400MHz 2MB L2 SSE: 0.44 11.95 4.85 0.85 0.93 12.73 5.04 1.61 [hinv]
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 0.26 5.35 1.80 0.41 0.47 5.66 1.83 0.85 [hinv]
Indigo2 R10000 175MHz 1MB L2 SolidIMPACT: 0.20 5.27 1.79 0.35 0.43 5.56 1.81 0.80 [hinv]
Indy R5000SC 180MHz 512K XZ (4GE): 0.19 0.90 0.93 0.40 0.38 0.91 0.96 0.74
Indy R5000SC 180MHz 512K XL24: 0.26 0.65 0.44 0.35 0.67 0.88 0.55 0.69
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 0.16 2.03 2.14 0.33 0.29 2.05 2.25 0.60 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 0.14 2.01 2.12 0.30 0.25 2.02 2.21 0.49 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 0.11 4.88 1.76 0.22 0.20 5.29 1.82 0.39 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 0.10 1.97 2.06 0.21 0.17 1.99 2.15 0.35 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 0.17 0.90 0.92 0.30 0.28 0.91 0.96 0.55
Indy R5000PC 150MHz XZ (4GE): 0.15 0.88 0.92 0.31 0.26 0.89 0.95 0.54 [hinv]
Indy R5000PC 150MHz XL24: 0.23 0.50 0.39 0.29 0.38 0.69 0.49 0.57
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 0.14 0.90 0.94 0.28 0.25 0.91 0.96 0.49
Indigo2 R4400SC/250 2MB XL24: 0.22 0.61 0.41 0.29 0.33 0.74 0.57 0.55
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 0.14 0.64 0.71 0.28 0.25 0.65 0.70 0.49
Indy R4400SC 200MHz 1MB XZ (4GE): 0.13 0.90 0.93 0.26 0.22 0.91 0.95 0.44
Indy R4600SC 133MHz 512K L2 XZ (4GE): 0.11 0.89 0.92 0.24 0.17 0.90 0.95 0.35 [hinv]
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 0.10 0.63 0.69 0.22 0.18 0.63 0.70 0.38 [hinv]
Indy R4600PC 133MHz XZ (4GE): 0.07 0.86 0.90 0.15 0.11 0.87 0.92 0.25 [hinv]
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 0.06 0.87 0.91 0.12 0.10 0.89 0.94 0.18 [hinv]
Indy R4400SC/200 XL24: 0.17 0.44 0.33 0.12 0.25 0.51 0.40 0.42
Indy R4600SC/133 512K L2 XL8: 0.12 0.34 0.33 0.18 0.17 0.42 0.41 0.34
Indy R4600PC/133 XL24: 0.08 0.26 0.26 0.06 0.11 0.34 0.35 0.11
Indy R4600PC/100 XL24: 0.07 0.20 0.20 0.05 0.10 0.25 0.26 0.09
Test C: chromeskins (7156 polygons). ChromeSkins [screenshot] is a reflective model of the
O2 chassis. O2 users will find the full demo in
/usr/demos/General_Demos/chromeskins.
MW MW/SH MW/W SW SW/SH SW/W
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 321.26 323.04 216.53 322.11 322.72 217.58 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 111.48 199.78 140.61 240.04 240.26 166.67 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 93.37 187.20 126.05 137.54 448.82 212.00 [hinv]
Fuel R14K/500 2MB L2 V10: 92.13 183.66 123.81 124.00 442.26 210.92 [hinv]
Fuel R16K/700 4MB L2 V10: 91.55 174.20 107.60 136.05 440.30 215.21 [hinv]
Octane R12000SC/300 2MB L2 V6: 81.35 126.79 72.79 131.56 298.29 128.06 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 94.38 95.95 61.96 116.82 139.65 79.26 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 94.09 92.01 62.72 113.85 120.09 80.48 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 88.92 89.14 60.34 112.89 128.12 80.03 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 92.02 92.51 60.29 98.59 98.65 78.31 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 87.21 87.59 69.56 87.89 87.61 77.55 [hinv]
POWER Series R4400SC/150 RE 4RM: 74.98 81.38 57.83 76.19 83.27 60.75 [hinv]
Octane2 R12000 400MHz 2MB L2 SSE: 1.47 80.35 68.65 5.32 146.30 91.90 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 60.89 78.28 56.73 97.89 141.05 70.14 [hinv]
Octane R12000SC/300 2MB L2 SSE: 1.33 79.63 64.10 4.15 140.81 86.89 [hinv]
POWER Series R4400SC/150 RE 2RM: 70.85 76.60 52.59 76.05 83.59 59.79 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 59.40 58.70 45.64 92.73 94.30 70.04 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 52.92 70.79 52.96 72.75 101.75 62.63 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 50.30 67.51 51.02 71.92 101.52 61.10 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 48.65 62.54 50.18 68.33 90.24 59.70 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 47.55 62.75 50.15 66.14 87.62 59.22
POWER Series R4400SC/150 RE 1RM: 60.86 59.53 43.39 75.45 80.80 57.66 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 47.03 55.72 49.02 64.48 75.45 57.83 [hinv]
POWER Series R4000SC/100 RE 2RM: 49.09 49.12 50.74 49.37 49.22 56.33
POWER Series R4000SC/100 RE 1RM: 48.10 48.31 41.33 48.31 48.12 52.82
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 30.32 32.44 42.10 35.50 37.05 48.50 [hinv]
Octane R12000SC/300 2MB L2 SE: 1.30 44.20 35.91 4.21 83.13 50.95 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 29.64 39.45 29.12 48.71 73.29 36.02
Octane R12000SC/400 2MB L2 SE+Texture: 28.40 39.63 32.78 44.55 72.62 42.40
Octane R10000SC/195 1MB L2 SE+Texture: 28.41 37.62 25.76 45.08 67.52 31.31
Indigo2 R4400SC/200 2MB HighIMPACT-AA 1MB: 27.99 36.48 34.36 43.59 58.52 45.73 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 27.53 36.50 32.00 43.44 62.20 41.42
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 0.62 38.46 32.15 2.32 67.53 40.55 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 27.90 37.60 29.40 41.29 60.20 36.90
O2 R12000SC 300MHz 1MB: 22.60 33.81 34.85 39.51 61.02 54.51 [hinv]
O2 R12000SC 270MHz 1MB: 21.97 32.42 30.50 36.33 53.86 47.40 [hinv]
O2 R5200SC 300MHz 1MB: 19.55 29.97 34.29 30.65 48.27 54.68 [hinv]
O2 R10000SC 195MHz 1MB: 18.56 27.91 31.86 27.33 42.87 47.99 [hinv]
POWER Series R3000SC/33 RE 1RM: 23.58 23.38 33.38 23.59 23.38 36.37
O2 R5000SC 200MHz 1MB: 17.50 25.28 30.22 22.44 36.60 42.50
O2 R10000SC 150MHz 1MB: 16.12 23.67 27.49 22.27 33.12 38.08 [hinv]
O2 R5000SC 180MHz 512K: 14.46 21.78 27.16 18.50 29.17 36.40
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 0.40 29.19 27.07 1.33 57.58 37.93 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 0.66 23.25 26.10 2.01 28.63 32.76 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 0.59 22.50 25.60 1.71 27.74 31.76 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 0.42 22.02 24.86 1.33 26.72 30.95 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 0.53 11.76 12.89 1.75 14.56 15.77
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 0.54 11.82 13.08 1.65 14.47 15.72
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indy R4400SC 200MHz 1MB XZ (4GE): 0.55 11.66 13.04 1.69 14.40 15.58
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 0.24 11.36 12.55 0.83 13.82 14.91 [hinv]
Indy R4600PC 133MHz XZ (4GE): 0.28 10.94 12.04 1.14 13.20 14.29 [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 0.08 10.00 11.64 0.23 11.82 13.63 [hinv]
Indigo2 R4400SC/250 2MB XL24: 1.21 7.50 7.88 3.20 13.62 11.60
Indy R5000SC 180MHz 512K XZ (4GE): 2.14 5.71 6.03 3.62 5.93 6.33
Indy R5000SC 180MHz 512K XL24: 1.22 6.13 6.77 3.56 12.61 10.57
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 0.55 8.75 10.09 1.69 9.90 11.40
Indy R4400SC/200 1MB XL24: 1.02 5.89 5.95 2.78 10.07 9.12
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 0.42 8.54 9.77 1.34 9.66 11.11 [hinv]
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indy R5000PC 150MHz XL24: 1.02 5.21 5.62 2.91 9.51 8.46
Indy R4600SC/133 512K L2 XL8: 0.72 4.15 4.72 1.91 7.11 7.22
Indy R4600PC/133 XL24: 0.63 3.32 3.94 1.45 5.39 5.89
Indigo R3000 33MHz XS24Z (1GE): 0.12 4.50 5.11 0.31 4.70 5.34 [hinv]
Indy R4600PC/100 XL24: 0.52 2.96 3.22 1.18 4.55 4.75
Indigo R3000 33MHz Entry: 0.29 1.68 1.98 0.46 2.11 2.36 [hinv]
Test D: angus.iv (3459 polygons). Angus [screenshot] is a non-textured model of an
unmanned underwater vehicle.
MW MW/W MW/HL SW SW/W SW/HL
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 281.95 276.35 434.85 282.75 255.10 419.11 [hinv]
Fuel R16K/700 4MB L2 V10: 269.33 268.41 207.80 499.63 491.76 479.28 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 146.33 109.41 170.11 246.25 268.87 373.20 [hinv]
Fuel R14K/500 2MB L2 V10: 145.00 108.30 159.99 251.59 260.04 365.66 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 150.31 109.79 136.02 208.99 201.65 281.55 [hinv]
Octane R12000SC/300 2MB l2 V6: 101.41 67.26 93.49 200.94 128.86 174.70 [hinv]
Onyx2 IR R10K/195 4MB L2 1RM: 155.00 ? ? 228.00 ? ? [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 80.26 62.02 87.62 128.71 105.21 158.76 [hinv]
Octane R12000SC/300 2MB L2 SSE: 71.85 62.04 83.04 125.38 103.41 157.23 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 65.89 53.67 77.11 97.28 81.23 125.48 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 62.85 56.02 77.00 94.33 80.90 120.59 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 67.11 52.77 76.39 92.92 78.42 120.72 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 65.87 51.62 74.55 91.95 77.35 115.55 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 58.55 52.73 71.36 88.34 76.30 112.28
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 58.95 48.19 68.97 80.10 68.75 102.96 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 59.18 59.88 65.89 60.76 67.07 92.02 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 59.19 59.47 49.91 60.70 60.78 66.35 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 58.97 57.65 79.10 59.72 70.35 96.52 [hinv]
O2 R5200SC 300MHz 1MB: 53.57 56.17 53.85 84.58 96.90 93.22 [hinv]
O2 R12000SC 270MHz 1MB: 54.33 50.52 51.90 99.75 89.65 91.30 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 59.58 49.07 63.57 61.07 67.20 92.41 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 59.97 48.51 65.54 60.92 66.62 91.52 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 44.75 36.19 46.37 59.51 59.98 83.25 [hinv]
POWER Series R4400SC/150 RE 4RM: 43.36 48.33 68.18 43.63 53.78 75.81 [hinv]
POWER Series R4400SC/150 RE 2RM: 43.05 44.23 61.12 43.69 52.37 73.81 [hinv]
POWER Series R4000SC/100 RE 2RM: 41.17 42.05 59.46 42.34 50.72 71.63
POWER Series R4000SC/100 RE 1RM: 40.53 36.56 47.48 42.15 47.01 66.50
Octane R12000SC/400 2MB L2 SE+Texture: 40.79 33.70 47.91 72.60 60.71 89.18
Octane R10000SC/250 1MB L2 SE+Texture: 40.88 33.80 47.46 68.50 58.63 85.24
Octane R12000SC/300 2MB L2 SE: 36.79 33.72 45.20 69.07 60.16 86.79 [hinv]
Octane R10000SC/195 1MB L2 SE+Texture: 35.86 33.83 46.60 63.56 58.12 77.86
POWER Series R4400SC/150 RE 1RM: 38.63 33.81 40.98 43.69 49.61 68.98 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 37.70 31.41 46.50 58.95 52.04 78.20
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 38.31 30.40 43.66 59.97 48.08 75.11 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 34.14 32.35 51.49 57.35 50.50 73.67
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 37.51 30.04 42.90 58.02 47.15 72.40 [hinv]
POWER Series R3000SC/33 RE 1RM: 34.32 33.70 45.02 35.38 39.86 55.30
O2 R10000SC 195MHz 1MB: 24.84 28.02 38.22 41.40 51.74 72.15 [hinv]
O2 R5000SC 200MHz 1MB: 21.05 28.90 38.27 32.40 50.30 66.98
O2 R10000SC 150MHz 1MB: 22.14 25.40 [hinv]
O2 R5000SC 180MHz 512K: 17.68 26.80 35.60 26.23 45.25 57.00 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 17.44 21.47 26.80 19.60 35.01 44.20 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 16.93 21.01 26.92 19.01 34.14 43.32 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 16.81 20.50 25.96 18.21 33.02 41.41 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 9.16 11.85 14.93 10.42 19.83 25.27
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 9.12 11.89 14.70 10.22 19.62 25.13
Indy R5000SC 180MHz 512K XZ (4GE): 9.06 11.75 14.71 10.27 19.58 25.02
Indy R4400SC 200MHz 1MB XZ (4GE): 8.95 11.76 14.61 10.19 19.38 24.76
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 8.39 11.27 13.80 9.35 18.05 22.58 [hinv]
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indy R4600PC 133MHz XZ (4GE): 7.85 10.81 13.35 8.69 28.19 28.01 [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 7.00 10.64 13.46 7.63 17.33 21.48
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 6.81 10.70 13.18 7.45 16.70 20.83 [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 7.26 9.33 11.51 8.04 14.14 17.85 [hinv]
Indigo2 R4400SC/250 2MB XL24: 5.86 6.79 6.95 9.58 10.79 12.80
Indy R5000SC 180MHz 512K XL24: 5.48 5.86 6.55 10.51 10.95 12.50 [hinv]
Indigo R4400SC/150 1MB GR2-XS24Z (1GE): 3.17 7.67 9.16 3.27 10.27 11.72 [hinv]
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 3.09 7.44 8.70 3.19 9.68 10.82 [hinv]
Indigo R3000 33MHz XS24Z (1GE): 2.93 6.77 8.03 3.00 8.62 9.93 [hinv]
Indy R4400SC/200 XL24: 4.60 4.91 5.56 7.55 8.59 9.83
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indy R5000PC 150MHz XL24: 4.47 4.42 5.02 7.92 8.22 9.58
Indy R4600SC/133 512K L2 XL8: 3.57 3.56 4.41 5.59 7.26 8.25
Indy R4600PC/133 XL24: 2.38 3.18 3.43 3.90 5.74 6.27
Indy R4600PC/100 XL24: 2.24 2.64 3.00 3.57 4.76 5.59
Indigo R3000 33MHz Entry: 1.40 1.81 1.98 1.73 2.52 2.80 [hinv]
Test E: e.iv (1568 polygons, complexity value 0.7). e.iv [screenshot] is a simple model of Earth
using a single small Earth texture on an Inventor sphere node.
MW MW/SH MW/W MW/HL SW SW/SH SW/W SW/HL
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 717.55 876.27 687.30 919.70 1320.02 3153.05 621.07 1142.86 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 248.22 260.11 263.60 346.97 910.39 1075.10 778.11 1076.10 [hinv]
Fuel R16K/700 4MB L2 V10: 239.06 247.72 250.12 905.18 1072.99 767.58 [hinv]
Fuel R14K/500 2MB L2 V10: 242.80 253.60 257.78 340.89 878.78 1040.50 762.43 1045.58 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 174.08 256.98 245.84 291.92 597.16 888.19 504.98 700.65 [hinv]
Octane R12000SC/300 2MB L2 V6: 180.12 185.25 137.66 192.62 558.95 599.47 369.02 494.81 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 207.14 207.07 133.80 199.33 349.78 362.99 178.78 270.10 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 108.12 119.07 137.07 166.30 330.50 408.42 286.13 389.48 [hinv]
Octane R12000SC/300 2MB L2 SSE: 2.32 119.20 132.23 164.71 7.52 403.59 254.69 385.05 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 136.82 136.79 115.56 158.58 357.21 381.31 169.61 253.58 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 137.25 136.59 115.52 158.09 356.63 377.02 169.79 258.14 [hinv]
POWER Series R4400SC/150 1MB L2 RE 4RM: 197.68 201.27 119.35 178.04 262.68 278.63 133.88 206.04 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 137.12 136.74 115.60 158.96 356.12 377.29 170.03 257.38 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 137.06 137.14 115.56 158.92 330.38 351.13 169.33 257.16 [hinv]
POWER Series R4400SC/150 1MB L2 RE 2RM: 137.85 137.76 114.46 158.56 268.03 274.64 134.50 202.56 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 103.12 115.85 118.56 150.74 300.20 352.21 190.19 272.28 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 100.49 114.62 119.01 150.70 312.38 369.50 186.79 278.54 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 97.47 108.74 113.00 143.48 269.07 315.78 171.27 251.24
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 95.78 108.17 111.59 141.90 257.84 305.81 168.75 246.67 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 95.02 106.55 110.31 140.79 251.34 294.38 166.29 242.76 [hinv]
POWER Series R4000SC/100 1MB L2 RE 2RM: 137.62 138.01 105.51 154.01 211.80 222.36 117.99 181.17
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 85.59 95.25 98.24 124.14 180.67 213.43 138.41 195.86 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 76.04 75.82 79.82 101.95 264.22 268.01 154.56 201.48 [hinv]
POWER Series R4000SC/100 1MB L2 RE 1RM: 77.83 77.68 78.42 103.69 179.33 192.48 114.48 171.79
Octane R12000SC/400 2MB L2 SE+Texture: 53.85 59.78 69.72 86.01 190.46 225.90 147.98 206.40
Octane R10000SC/250 1MB L2 SE+Texture: 52.91 59.88 70.22 85.57 179.50 227.18 149.04 204.95
Octane R10000SC/195 1MB L2 SE+Texture: 52.18 59.75 69.68 85.55 168.07 225.02 147.62 197.90
POWER Series R4400SC/150 1MB L2 RE 1RM: 54.94 54.91 71.36 75.51 213.78 228.28 127.76 189.89
Octane R12000SC/300 2MB L2 SE: 2.28 60.17 70.91 86.71 7.50 211.53 135.23 199.94 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 52.97 60.16 68.68 83.55 171.91 222.86 124.66 178.46
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 54.53 57.75 65.60 81.46 174.04 191.81 117.60 168.80
POWER Series R3000SC/33 RE 1RM: 52.98 78.01 71.75 93.08 100.38 121.15 81.10 117.94
O2 R12000SC 300MHz 1MB: 35.80 54.23 60.42 76.75 111.19 169.84 135.02 182.16 [hinv]
O2 R12000SC 270MHz 1MB: 36.10 54.03 59.44 76.45 110.25 162.05 128.99 178.30 [hinv]
O2 R5200SC 300MHz 1MB: 33.56 52.15 57.84 73.19 98.85 147.58 131.45 179.55 [hinv]
O2 R10000SC 195MHz 1MB: 34.93 52.33 58.44 72.69 96.27 142.33 122.40 169.66 [hinv]
O2 R5000SC 200MHz 1MB: 35.65 51.77 57.48 72.70 88.10 125.31 123.23 160.84
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 1.21 59.08 64.26 81.33 4.19 205.94 106.42 159.69 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 0.72 57.52 62.37 79.01 2.35 178.29 99.30 148.70 [hinv]
O2 R5000SC 180MHz 512K: 33.16 48.68 54.22 69.01 75.50 105.63 110.72 142.44 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 0.32 58.26 57.68 61.68 1.36 140.61 114.02 140.49 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 0.31 56.28 55.86 60.26 1.32 130.60 108.88 132.63 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 0.29 55.00 55.25 58.21 1.22 124.75 103.50 127.75 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 0.35 31.45 32.95 33.75 1.47 70.64 66.83 77.84
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 0.34 32.10 33.55 34.39 1.44 70.00 68.23 78.34
Indy R5000SC 180MHz 512K XZ (4GE): 0.36 30.70 32.63 33.44 1.52 68.72 65.06 75.46
Indy R4400SC 200MHz 1MB XZ (4GE): 0.34 30.85 32.46 33.38 1.41 69.21 65.64 75.37
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 0.25 30.84 31.80 32.90 0.98 63.01 60.80 69.59 [hinv]
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 0.41 30.44 32.13 33.08 1.69 49.33 51.16 58.50
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 0.37 28.92 30.60 31.63 1.54 48.08 49.30 56.30 [hinv]
Indy R4600PC 133MHz XZ (4GE): 0.26 28.73 30.03 30.95 1.01 58.08 55.48 64.07 [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 0.13 26.89 27.94 29.27 0.40 50.38 48.27 57.29 [hinv]
Indigo2 R4400SC/250 2MB XL24: 1.21 12.83 16.28 15.13 3.96 25.55 25.25 27.33
Indy R5000SC 180MHz 512K XL24: 1.20 10.11 13.66 12.73 3.99 26.56 24.34 27.50
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 0.36 19.35 20.61 22.66 1.14 22.77 24.33 27.96 [hinv]
Indigo R3000 33MHz XS24Z (1GE): 0.14 18.01 19.22 21.02 0.44 20.71 22.02 25.51 [hinv]
Indy R4400SC/200 1MB XL24: 1.00 10.03 12.19 11.54 3.12 20.50 20.44 22.10
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indy R5000PC 150MHz XL24: 0.98 8.22 11.10 9.86 3.12 18.78 17.82 18.95
Indy R4600SC/133 512K L2 XL8: 0.62 7.82 9.75 9.26 1.81 16.45 16.32 18.04
Indy R4600PC/133 XL24: 0.52 5.79 7.89 7.34 1.57 11.88 12.94 13.75
Indy R4600PC/100 XL24: 0.40 5.15 6.47 6.37 1.30 10.59 10.78 11.95
Indigo R3000 33MHz Entry: 0.30 3.56 4.50 4.32 0.63 5.06 5.54 5.85 [hinv]
Test F: stars3.iv (0 polygons, 16666 points). This model [screenshot] is a point data set made to
resemble a simple galaxy. The model file actually contains several
sub-models and as such is deliberately inefficient.
MW SW
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 156.43 175.89 [hinv]
Onyx2 IR R10K/195 1-CPU 1RM: 150.00 130.00 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 81.18 81.66 [hinv]
Fuel R16K/700 4MB L2 V10: 75.39 96.51 [hinv]
Fuel R14K/500 2MB L2 V10: 71.54 84.27 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 74.76 84.85 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 47.66 54.92 [hinv]
Octane R12000SC/300 2MB L2 V6: 46.79 53.79 [hinv]
Octane R12000SC/300 2MB L2 SSE: 45.84 52.42 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 42.26 48.48 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 39.95 45.30
Octane R12000SC/300 2MB L2 SE: 38.69 50.80 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 38.55 50.80
Octane R12000SC/400 2MB L2 SE+Texture: 38.25 50.37
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 37.67 40.84 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 37.67 40.61 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 37.60 40.02 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 37.18 40.80 [hinv]
Octane R10000SC/195 1MB L2 SE+Texture: 35.58 45.69
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 34.96 39.13 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 34.53 38.28 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 34.08 37.97 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 33.52 42.10 [hinv]
Indigo2 R4400SC/200 2MB HighIMPACT-AA 1MB: 33.28 43.54 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 33.07 41.60
Onyx R4400SC/150 1MB L2 RE2 4RM4: 32.72 33.85 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 32.24 39.70
Onyx R4400SC/150 1MB L2 RE2 1RM4: 31.13 36.48 [hinv]
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 31.12 39.51 [hinv]
POWER Series R4400SC/150 RE 4RM: 29.48 30.62 [hinv]
POWER Series R4400SC/150 RE 2RM: 28.60 30.50 [hinv]
POWER Series R4000SC/100 RE 2RM: 28.00 29.21
POWER Series R4400SC/150 RE 1RM: 27.39 30.49 [hinv]
POWER Series R4000SC/100 RE 1RM: 26.51 29.13
O2 R12000SC 300MHz 1MB: 25.19 29.09 [hinv]
POWER Series R3000SC/33 RE 1RM: 23.40 24.79
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 23.34 25.14 [hinv]
O2 R12000SC 270MHz 1MB: 22.70 24.95 [hinv]
O2 R5200SC 300MHz 1MB: 20.30 22.60 [hinv]
O2 R10000SC 195MHz 1MB: 19.62 22.01 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 17.40 18.58 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 17.01 18.04 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 16.70 17.57 [hinv]
O2 R5000SC 200MHz 1MB: 15.05 16.48
O2 R5000SC 180MHz 512K: 13.20 14.25 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 11.08 12.05
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 10.96 11.91
Indy R5000SC 180MHz 512K XZ (4GE): 10.91 11.87
Indy R4400SC 200MHz 1MB XZ (4GE): 10.86 11.82
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 10.43 11.27 [hinv]
Indy R4600PC 133MHz XZ (4GE): 10.23 11.02 [hinv]
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 7.29 7.63 [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 6.97 7.35
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 6.82 7.20 [hinv]
Indigo2 R4400SC/250 2MB XL24: 5.32 6.03
Indy R5000SC 180MHz 512K XL24: 5.03 5.87
Indy R4400SC/200 1MB XL24: 4.49 4.91
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indy R5000PC 150MHz XL24: 3.68 4.16
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 3.00 3.08 [hinv]
Indigo R3000 33MHz XS24Z (1GE): 2.96 3.01 [hinv]
Indy R4600SC/133 512K L2 XL8: 2.84 3.15
Indy R4600PC/133 XL24: 2.02 2.24
Indy R4600PC/100 XL24: 1.84 2.01
Indigo R3000 33MHz Entry: 0.91 0.96 [hinv]
Test G: ss78ne_d30.iv (1682 polygons). This is an inefficient
landscape model [screenshot]
with a different colour value for each vertex (30x30 grid). The full
version of this model is a 500x500 grid - I wrote my own program to
'extract' simpler versions from the complete model. The colouring
algorithm is also my own. The dataset is actually part of England,
but I don't know which part. Data supplied by Ordnance Survey.
MW MW/W SW SW/W
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 716.60 634.16 1219.54 603.26 [hinv]
Octane2 R12000SC/400 2MB L2 V12: 297.19 317.42 1076.87 781.36 [hinv]
Fuel R16K/700 4MB L2 V10: 322.38 304.90 1057.00 778.46 [hinv]
Fuel R14K/500 2MB L2 V10: 290.52 308.31 1066.00 771.20 [hinv]
Onyx2 IR R10K/195 1-CPU 1RM: 321.00 ? 875.00 ? [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 295.18 267.11 805.81 482.39 [hinv]
Octane R12000SC/300 2MB L2 V6: 212.85 163.44 780.04 464.69 [hinv]
Octane R12000SC/300 2MB l2 SSE: 141.37 139.66 382.04 238.14 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 128.17 131.25 377.65 209.67 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 190.88 122.75 211.33 157.51 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 126.75 117.58 294.46 180.65 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 124.12 117.78 282.40 176.24 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 155.21 108.24 251.70 151.17 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 154.31 108.02 245.04 148.63 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 153.63 108.23 237.41 151.01 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 153.42 108.19 246.22 150.77 [hinv]
POWER Series R4400SC/150 RE 4RM: 146.87 113.89 152.99 132.38 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 119.20 112.75 256.01 169.08
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 116.70 111.78 256.74 167.12 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 115.97 109.80 239.04 159.25 [hinv]
POWER Series R4400SC/150 RE 2RM: 138.51 103.83 154.17 132.36 [hinv]
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 101.74 96.88 188.51 134.62 [hinv]
POWER Series R4000SC/100 RE 2RM: 121.86 95.41 131.89 115.09
POWER Series R4000SC/100 RE 1RM: 92.00 81.70 133.54 116.12
POWER Series R4400SC/150 RE 1RM: 86.11 82.60 152.73 127.42 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 84.84 81.53 204.86 135.74 [hinv]
Octane R12000SC/300 2MB L2 SE: 70.12 72.31 204.68 136.74 [hinv]
Octane R12000SC/400 2MB L2 SE+Texture: 64.58 70.71 202.80 137.03
Octane R10000SC/250 1MB L2 SE+Texture: 63.88 67.81 185.52 116.81
Octane R10000SC/195 1MB L2 SE+Texture: 62.84 64.17 181.72 105.92
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 64.20 63.76 165.53 100.96 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT: 62.72 66.77 162.70 115.57
Octane R10000/175 1MB L2 SI+Tex: 63.84 65.78 158.64 107.56
POWER Series R3000SC/33 RE 1RM: 75.32 65.30 75.68 79.56
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 61.95 61.88 163.07 99.95 [hinv]
O2 R12000SC 270MHz 1MB: 61.40 63.88 134.36 134.07 [hinv]
O2 R5200SC 300MHz 1MB: 56.86 60.59 120.66 131.41 [hinv]
O2 R12000SC 300MHz 1MB: 55.32 61.18 145.75 128.31 [hinv]
O2 R10000SC 195MHz 1MB: 55.28 58.35 106.12 119.43 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 59.07 60.20 76.61 115.42 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 57.34 59.07 72.16 109.65 [hinv]
O2 R5000SC 200MHz 1MB: 54.41 58.72 91.69 114.50
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 56.09 57.31 70.23 103.67 [hinv]
O2 R5000SC 180MHz 512K: 51.16 56.20 75.50 95.20 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 28.35 37.06 35.38 72.10
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 28.14 36.56 34.95 70.96
Indy R4400SC 200MHz 1MB XZ (4GE): 28.04 34.31 35.00 71.56
Indy R5000SC 180MHz 512K XZ (4GE): 27.92 34.48 34.78 68.68
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 26.82 33.19 32.72 62.55 [hinv]
Indy R4600PC 133MHz XZ (4GE): 25.64 31.27 30.70 56.47 [hinv]
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 24.30 29.02 28.14 48.33 [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 19.66 34.28 22.72 51.30
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 19.12 32.00 22.22 49.28 [hinv]
Indigo2 R4400SC/250 2MB XL24: 11.54 15.74 22.49 25.81
Indy R5000SC 180MHz 512K XL24: 9.04 13.90 21.87 23.81
Indy R4400SC/200 1MB XL24: 9.48 11.76 18.08 17.13
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 9.01 19.73 9.63 22.86 [hinv]
Indigo R3000 33MHz XS24Z (1GE): 8.71 18.38 9.21 20.61 [hinv]
Indy R5000PC 150MHz XL24: 7.94 11.02 18.75 18.43
Indy R4600SC/133 512K L2 XL8: 6.81 9.44 14.40 16.04
Indy R4600PC/133 XL24: 5.11 7.20 9.51 11.08
Indy R4600PC/100 XL24: 4.70 6.43 8.47 10.02
Indigo R3000 33MHz Entry: 2.92 4.38 3.91 5.42 [hinv]
Test H: ss78ne_d100.iv (19602 polygons). This is a more complex
100x100 grid version of the landscape model [screenshot].
MW MW/W SW SW/W
Octane2 R12000SC/400 2MB L2 V12: 130.75 85.12 160.26 97.16 [hinv]
Fuel R14K/500 2MB L2 V10: 130.03 84.77 159.80 97.50 [hinv]
Fuel R16K/700 4MB L2 V10: 128.24 84.21 157.23 93.00 [hinv]
Octane R12000SC/300 2MB L2 V6: 125.64 56.31 159.52 73.93 [hinv]
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 148.44 41.90 148.61 41.15 [hinv]
Onyx2 IR R10K/195 1-CPU 1RM: 116.00 ? 116.00 ? [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 113.39 42.78 113.98 43.56 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 45.65 21.22 52.86 22.67 [hinv]
Octane R12000SC/300 2MB L2 SSE: 44.33 24.05 50.85 26.00 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 33.09 18.00 36.54 19.01 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 32.93 18.08 36.42 19.16 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 32.73 18.10 36.10 19.09 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 32.41 17.95 35.59 18.94 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 32.22 18.09 35.35 19.01
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 29.39 17.09 32.29 17.96 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 25.94 13.59 26.63 14.05 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 25.89 13.58 26.22 14.05 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 25.78 13.56 26.46 14.02 [hinv]
Octane R12000SC/300 2MB L2 SE: 22.74 13.76 26.31 15.01 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 25.76 13.56 26.28 14.05 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 25.62 13.76 25.62 14.12 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 24.85 13.06 25.73 13.94 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 22.40 11.25 26.15 12.18
Octane R12000SC/400 2MB L2 SE+Texture: 22.34 13.89 26.01 15.23
Octane R10000SC/195 1MB L2 SE+Texture: 20.98 10.38 24.19 11.16
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 18.66 12.09 21.06 13.10
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 18.71 10.24 20.99 10.89 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 18.46 10.14 20.95 10.81 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 18.11 11.32 20.25 12.12
O2 R12000SC 300MHz 1MB: 16.90 15.53 18.61 17.44 [hinv]
POWER Series R4400SC/150 RE 4RM: 17.56 13.51 ? ? [hinv]
POWER Series R4400SC/150 RE 2RM: 17.33 12.77 17.61 13.44 [hinv]
POWER Series R4400SC/150 RE 1RM: 17.25 12.95 17.55 13.59 [hinv]
POWER Series R4400SC/100 RE 2RM: 17.12 13.16 17.15 13.16
POWER Series R4400SC/100 RE 1RM: 16.88 12.79 17.15 13.18
O2 R12000SC 270MHz 1MB: 15.03 17.53 16.73 21.72 [hinv]
O2 R5200SC 300MHz 1MB: 14.05 16.83 15.22 18.52 [hinv]
POWER Series R3000SC/33 RE 1RM: 12.10 12.28 12.07 12.36
O2 R10000SC 195MHz 1MB: 11.07 14.80 11.80 16.87 [hinv]
O2 R5000SC 200MHz 1MB: 10.13 13.51 10.65 14.61
O2 R5000SC 180MHz 512K: 9.66 11.37 10.24 12.18 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 7.63 15.07 7.82 18.28 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 7.57 14.75 7.71 17.95 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 7.43 14.26 7.59 17.56 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 3.34 8.09 3.43 8.56
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 3.32 8.05 3.41 8.52
Indy R5000SC 180MHz 512K XZ (4GE): 3.33 7.99 3.42 8.49
Indy R4400SC 200MHz 1MB XZ (4GE): 3.33 8.01 3.41 8.48
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 3.27 7.81 3.32 8.24 [hinv]
Indy R4600PC 133MHz XZ (4GE): 3.25 7.62 3.31 8.06 [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 3.21 7.53 3.25 7.86 [hinv]
Indy R5000SC 180MHz 512K XL24: 2.81 2.64 3.76 3.42
Indigo2 R4400SC/250 2MB XL24: 2.45 2.41 2.97 2.73
Indy R5000PC 150MHz XL24: 2.16 2.19 2.98 2.77
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 2.09 5.05 2.13 5.26
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 2.06 5.01 2.08 5.20 [hinv]
Indy R4400SC/200 1MB XL24: 2.03 2.15 2.77 2.54
Indy R4600PC/133 XL24: 1.11 1.21 1.37 1.43
Indy R4600PC/100 XL24: 0.95 1.14 1.21 1.37
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 0.88 2.17 0.88 2.20 [hinv]
Indigo R3000 33MHz XS24Z (1GE): 0.87 2.12 0.88 2.16 [hinv]
Indigo R3000 33MHz Entry: 0.49 0.66 0.53 0.70 [hinv]
Test I: ss78ne_d300.iv (178802 polygons). More complex again [screenshot], now a 300x300 grid. Notice how
the more powerful systems become very much CPU-bound since this test
only uses one CPU.
MW MW/W SW SW/W
Fuel R16K/700 4MB L2 V10: 18.26 10.92 18.26 10.94 [hinv]
Fuel R14K/500 2MB L2 V10: 18.24 10.91 18.74 11.09 [hinv]
Octane R12000SC/300 2MB L2 V6: 18.12 8.70 18.57 9.25 [hinv]
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 16.93 4.79 16.93 4.79 [hinv]
Onyx2 IR R10K/195 1-CPU 1RM: 13.60 ? 13.50 ? [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 13.02 4.40 13.01 4.43 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 5.99 2.47 6.11 2.47 [hinv]
Octane R12000SC/300 2MB L2 SSE: 5.81 2.91 5.90 2.93 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 4.22 2.15 4.27 2.16 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 4.22 2.13 4.28 2.14 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 4.21 2.14 4.26 2.16 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 4.15 2.11 4.22 2.13 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 4.06 2.13 4.13 2.16
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 3.51 2.08 3.54 2.08 [hinv]
Octane R12000SC/300 2MB L2 SE: 3.03 1.68 3.04 1.69 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 3.02 1.48 3.08 1.49
Octane R12000SC/400 2MB L2 SE+Texture: 2.94 1.65 3.00 1.68
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 3.01 1.53 3.05 1.54 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 2.97 1.57 2.88 1.54 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 2.94 1.53 2.98 1.54 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 2.92 1.53 2.95 1.54 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 2.91 1.53 2.95 1.54 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 2.76 1.52 2.95 1.54 [hinv]
Octane R10000SC/195 1MB L2 SE+Texture: 2.77 1.26 2.81 1.27
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 2.42 1.21 2.46 1.22 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 2.39 1.47 2.44 1.48
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 2.41 1.21 2.44 1.22 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 2.31 1.35 2.36 1.38
POWER Series R4400SC/150 RE 2RM: 2.04 1.51 2.01 1.51 [hinv]
POWER Series R4400SC/150 RE 4RM: 2.02 1.52 2.01 1.51 [hinv]
POWER Series R4400SC/150 RE 1RM: 2.00 1.50 2.01 1.51 [hinv]
POWER Series R4000SC/100 RE 1RM: 2.01 1.48 2.01 1.48
POWER Series R4000SC/100 RE 1RM: 2.01 1.48 1.98 1.50
O2 R12000SC 300MHz 1MB: 1.92 2.26 1.95 2.35 [hinv]
O2 R12000SC 270MHz 1MB: 1.82 2.51 1.83 2.60 [hinv]
O2 R5200SC 300MHz 1MB: 1.68 1.97 1.70 2.19 [hinv]
POWER Series R3000SC/33 RE 1RM: 1.45 1.48 1.45 1.48
O2 R10000SC 195MHz 1MB: 1.29 2.01 1.30 2.05 [hinv]
O2 R5000SC 200MHz 1MB: 1.34 1.65 1.21 1.67
O2 R5000SC 180MHz 512K: 1.16 1.40 1.17 1.41 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 0.88 2.14 0.88 2.21 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 0.87 2.09 0.87 2.20 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 0.86 2.09 0.87 2.17 [hinv]
Indy R5000SC 180MHz 512K XL24: 0.41 0.36 0.53 0.40
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 0.38 0.96 0.38 0.97
Indy R5000SC 180MHz 512K XZ (4GE): 0.38 0.96 0.38 0.97
Indy R4400SC 200MHz 1MB XZ (4GE): 0.38 0.96 0.38 0.97
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 0.38 0.95 0.38 0.97
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 0.38 0.94 0.38 0.95 [hinv]
Indy R4600PC 133MHz XZ (4GE): 0.37 0.94 0.37 0.94 [hinv]
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 0.37 0.93 0.37 0.94 [hinv]
Indigo2 R4400SC/250 2MB XL24: 0.36 0.32 0.41 0.35
Indy R5000PC 150MHz XL24: 0.35 0.26 0.39 0.31
Indy R4400SC/200 1MB XL24: 0.29 0.27 0.33 0.31
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 0.24 0.59 0.24 0.59
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 0.23 0.57 0.23 0.57 [hinv]
Indy R4600PC/133 XL24: 0.15 0.15 0.16 0.17
Indy R4600PC/100 XL24: 0.13 0.15 0.13 0.17
Indigo R3000 33MHz XS24Z (1GE): 0.10 0.24 0.10 0.24 [hinv]
Indigo R3000 33MHz Entry: 0.06 0.07 0.06 0.08 [hinv]
Test J: lothian400.iv (318402 polygons). This is a different
landscape model [screenshot],
400x400 grid, of the area of Lothian region, Scotland (East coast). I
live on this mesh somewhere! 8) NOTE: for O2, R4K/250 HighIMPACT
Indigo2, Onyx2 and other fast systems, this test is CPU-limited, ie.
the bottleneck is not in the graphics pipe (the extra three CPUs in
the Onyx2 are not used by ivview, so the gfx system is not going as
fast as it could if more than one main CPU could be utilised, or if a
better CPU were present). In fact, the R4K/250 HighIMPACT I2 becomes
CPU-bound at the 100x100 stage. Landscape data supplied by the Radar
Research Lab at Heriot Watt University, Edinburgh, Scotland.
MW MW/W SW SW/W
Fuel R16K/700 4MB L2 V10: 10.41 6.19 10.58 6.25 [hinv]
Fuel R14K/500 2MB L2 V10: 10.41 6.19 10.57 6.24 [hinv]
Octane R12000SC/300 2MB L2 V6: 10.39 5.05 10.47 5.23
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 9.48 2.40 9.48 2.40 [hinv]
Onyx2 IR R10K/195 1-CPU 1RM: 7.50 ? 7.50 ? [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 7.35 2.43 7.35 2.54 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 3.41 1.41 3.45 1.41 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 2.40 2.15 4.27 2.16 [hinv] [not convinced these are correct!]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 2.39 1.21 2.41 1.21 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 2.38 1.21 2.40 1.21 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 2.33 1.20 2.38 1.20 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 2.31 1.21 2.33 1.21
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 1.99 1.18 2.00 1.18 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 1.77 0.86 1.77 0.86 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 1.71 0.82 1.73 0.83
Octane R12000SC/400 2MB L2 SE+Texture: 1.67 0.97 1.69 0.97
Onyx R4400SC/250 4MB L2 RE2 2RM4: 1.67 0.86 1.68 0.86 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 1.66 0.86 1.66 0.86 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 1.64 0.84 1.67 0.87 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 1.64 0.85 1.64 0.85 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 1.63 0.85 1.68 0.86 [hinv]
Octane R10000SC/195 1MB L2 SE+Texture: 1.57 0.71 1.59 0.72
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 1.37 0.83 1.38 0.83
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 1.37 0.68 1.38 0.69 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 1.36 0.68 1.37 0.69 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 1.31 0.77 1.32 0.77
POWER Series R4400SC/150 RE 4RM: 1.13 0.83 1.13 0.84 [hinv]
POWER Series R4400SC/150 RE 2RM: 1.12 0.83 1.12 0.83 [hinv]
POWER Series R4000SC/100 RE 2RM: 1.11 0.83 1.12 0.84
POWER Series R4400SC/150 RE 1RM: 1.11 0.83 1.11 0.83 [hinv]
POWER Series R4000SC/100 RE 1RM: 1.10 0.83 1.13 0.83
O2 R12000SC 300MHz 1MB: 1.04 1.31 1.10 1.38 [hinv]
O2 R5200SC 300MHz 1MB: 0.89 1.35 0.89 1.36 [hinv]
POWER Series R3000SC/33 RE 1RM: 0.80 0.83 0.80 0.83
O2 R10000SC 195MHz 1MB: 0.73 1.14 0.73 1.15 [hinv]
O2 R5000SC 200MHz 1MB: 0.63 0.86 0.63 0.86
O2 R5000SC 180MHz 512K: 0.61 0.80 0.61 0.80 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 0.49 1.23 0.49 1.25 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 0.49 1.23 0.49 1.25 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 0.49 1.21 0.49 1.23 [hinv]
Indy R5000SC 180MHz 512K XL24: 0.28 0.24 0.33 0.29
Indigo2 R4400SC/250 2MB XL24: 0.21 0.20 0.23 0.24
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 0.21 0.54 0.21 0.55
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 0.21 0.54 0.22 0.54
Indy R5000SC 180MHz 512K XZ (4GE): 0.21 0.54 0.21 0.55
Indy R4400SC 200MHz 1MB XZ (4GE): 0.21 0.54 0.21 0.55
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 0.21 0.53 0.21 0.53 [hinv]
Indy R4600PC 133MHz XZ (4GE): 0.21 0.53 0.21 0.53 [hinv]
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indy R5000PC 150MHz XL24: 0.21 0.16 0.24 0.19
Indy R4400SC/200 1MB XL24: 0.17 0.16 0.19 0.19
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 0.13 0.33 0.13 0.33
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 0.12 0.32 0.12 0.32 [hinv]
Indy R4600PC/133 XL24: 0.09 0.11 0.10 0.13
Indy R4600PC/100 XL24: 0.08 0.09 0.08 0.11
Test K: arm2.iv (810 polygons). This is a simple but highly accurate
model of a robot arm [screenshot]; the model file actually
contains Inventor constructs to allow the joint angles to be
altered. The real arm [129K JPEG
Photo] resides in the Ocean Systems Research Laboratory at Heriot
Watt University, Edinburgh, where I conducted undersea modeling
research during early 1995.
MW MW/W MW/HL SW SW/W SW/HL
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 1365.38 1183.68 1794.30 1448.64 1252.46 2006.00 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 439.08 458.88 615.01 886.13 766.39 1190.88 [hinv]
Fuel R14K/500 2MB L2 V10: 384.79 386.50 609.96 1241.15 1060.49 1510.08 [hinv]
Fuel R16K/700 4MB L2 V10: 366.86 377.82 1244.79 1036.19 [hinv]
Octane R12000SC/300 2MB L2 V6: 314.25 286.27 406.04 960.41 742.32 968.74 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 252.87 207.09 324.62 268.88 297.01 385.20 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 175.67 172.54 271.90 415.10 358.60 565.72 [hinv]
Octane R12000SC/300 2MB L2 SSE: 167.31 171.97 253.23 412.58 358.46 553.97 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 191.61 167.50 255.01 276.21 285.21 370.38 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 189.89 166.22 252.47 278.46 284.01 369.82 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 187.82 165.99 257.43 276.62 284.12 367.37 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 188.54 165.83 257.36 272.98 274.21 363.53 [hinv]
POWER Series R4400SC/150 RE 4RM: 177.55 171.74 248.33 183.31 197.22 267.55 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 165.13 149.92 236.23 406.43 283.24 437.78 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 161.91 150.36 233.00 396.38 284.11 436.55 [hinv]
POWER Series R4400SC/150 RE 2RM: 176.34 163.51 222.31 183.58 195.96 265.95 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 155.23 143.16 221.63 345.55 254.91 385.09 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 141.93 140.51 205.53 322.52 251.17 375.40
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 148.24 136.57 207.47 313.40 235.16 343.01 [hinv]
POWER Series R4000SC/100 RE 2RM: 150.30 143.23 198.50 158.81 164.51 222.40
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 128.11 118.34 175.68 224.79 182.85 256.90 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 108.24 109.30 162.45 270.01 256.67 356.99 [hinv]
POWER Series R4400SC/150 RE 1RM: 109.78 107.29 157.00 176.27 178.82 249.18 [hinv]
POWER Series R4000SC/100 RE 1RM: 103.39 104.42 168.21 161.98 160.10 222.06
Octane R12000SC/400 2MB L2 SE+Texture: 87.47 88.77 138.46 257.65 217.23 329.30
Octane R12000SC/300 2MB L2 SE: 85.30 87.98 134.57 246.75 211.55 312.05 [hinv]
O2 R12000SC 300MHz 1MB: 86.20 89.47 133.88 230.46 204.09 293.84 [hinv]
Octane R10000SC/250 1MB L2 SE+Texture: 83.73 85.95 135.20 222.22 199.00 297.41
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 86.81 82.17 130.27 247.51 169.67 268.32 [hinv]
Indigo2 R4400SC/250 2MB HighIMPACT 1MB: 83.30 81.71 126.80 221.61 174.19 262.21
Indigo2 R4400SC/200 2MB HighIMPACT-AA 1MB: 80.00 83.08 125.37 214.00 186.31 274.18 [hinv]
Octane R10000/175 1MB L2 SI+Tex: 80.72 82.50 125.80 216.21 176.50 260.56
Octane R10000SC/195 1MB L2 SE+Texture: 79.73 81.64 129.06 183.62 167.18 249.80
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 84.03 78.71 123.90 212.94 151.67 237.22 [hinv]
O2 R12000SC 270MHz 1MB: 79.51 87.64 124.38 212.03 198.40 206.14 [hinv]
O2 R5200SC 300MHz 1MB: 78.00 86.53 123.16 187.75 197.00 268.08 [hinv]
O2 R10000SC 195MHz 1MB: 78.90 86.18 126.72 182.43 192.15 262.20 [hinv]
O2 R5000SC 200MHz 1MB: 72.45 81.25 118.97 153.45 165.13 222.73
POWER Series R3000SC/33 RE 1RM: 97.09 91.36 118.21 99.09 96.00 128.95
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 72.75 91.21 95.39 110.40 154.81 168.47 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 70.04 87.42 91.36 105.03 145.12 160.44 [hinv]
O2 R5000SC 180MHz 512K: 68.30 75.36 108.34 131.50 142.54 188.71
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 68.19 83.97 88.59 98.85 135.06 150.95 [hinv]
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 39.80 50.86 52.41 62.34 89.88 98.61
Indy R4400SC 200MHz 1MB XZ (4GE): 39.17 49.69 51.98 62.64 89.78 99.47
Indy R5000SC 180MHz 512K XZ (4GE): 38.90 49.40 50.84 62.63 89.85 99.43
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 38.90 48.20 48.80 60.93 88.55 97.37
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 33.72 46.65 48.16 50.58 77.49 86.15
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 36.54 46.33 48.01 55.67 76.65 84.70 [hinv]
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 33.29 44.63 45.83 48.58 73.39 80.24 [hinv]
Indy R4600PC 133MHz XZ (4GE): 34.13 42.78 44.02 49.51 67.72 75.27 [hinv]
Indigo2 R4400SC/250 2MB XL24: 29.72 37.59 37.10 52.32 55.57 60.30
Indigo R3000 33MHz GR2-Elan (4GE): 30.47 36.75 38.55 41.17 52.19 60.15 [hinv]
Indy R5000SC 180MHz 512K XL24: 29.46 36.38 37.86 56.06 53.53 61.22
Indy R4400SC/200 1MB XL24: 23.52 27.65 27.90 42.10 43.05 48.67
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 19.27 29.59 31.54 23.04 39.36 43.13 [hinv]
Indy R5000PC 150MHz XL24: 20.04 26.52 26.15 39.23 37.55 42.87
Indigo R3000 33MHz XS24Z (1GE): 17.49 25.74 27.67 20.24 32.23 36.40 [hinv]
Indy R4600PC/133 XL24: 13.10 19.00 17.72 20.33 26.28 27.93
Indy R4600PC/100 XL24: 12.01 15.21 15.20 19.12 21.08 23.78
Indigo R3000 33MHz Entry: 7.26 9.34 9.61 9.90 11.89 12.54 [hinv]
Test L: stars4.iv (0 polygons, 300000 points). This model is a
simple, very large point data set [screenshot], constructed by placing
random points in a spherical volume, which was then squashed
vertically. Unlike Test F, this test is a single defined Inventor
object; there is no screenshot available since the object just
resembles a white blob when shown in an Inventor window (one must
zoom-in to see the individual points). For the maximum size window
(MW), I zoomed the object so that it just filled the ivview window
area (this is only reliably possible to do because the object is
circular). Note that a system needs more than 64MB RAM to run this
test properly.
MW SW
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 37.40 38.00 [hinv]
Octane R12000SC/400 2MB L2 V12: 20.07 21.78 [hinv]
Fuel R16K/700 4MB L2 V10: 18.24 22.66 [hinv]
Fuel R14K/500 2MB L2 V10: 17.55 20.57 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 15.42 15.44 [hinv]
Octane2 R12000SC/400 2MB L2 MXE: 12.53 13.03 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 9.38 9.62 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 9.36 9.56 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 9.33 9.55 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 9.23 9.40 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 8.48 8.38 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 8.43 8.70 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 7.76 7.93 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 7.74 7.92 [hinv]
Indigo2 R10000 195MHz 1MB L2 SolidIMPACT: 7.64 8.20 [hinv]
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 7.64 7.83 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 7.36 7.53 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 7.16 7.51 [hinv]
POWER Series R4400SC/150 RE 4RM: 7.24 7.30 [hinv]
POWER Series R4400SC/150 RE 2RM: 7.23 7.30 [hinv]
POWER Series R4400SC/150 RE 1RM: 6.93 7.02 [hinv]
POWER Series R4000SC/100 RE 1RM: 6.71 6.75
POWER Series R4000SC/100 RE 1RM: 6.65 6.69
POWER Series R3000SC/33 RE 1RM: 6.14 6.23
O2 R12000SC 300MHz 1MB: 5.88 6.14 [hinv]
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 5.18 5.26 [hinv]
O2 R10000SC 195MHz 1MB: 5.03 5.22 [hinv]
O2 R5200SC 300MHz 1MB: 4.76 4.95 [hinv]
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 4.28 4.32 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 4.25 4.31 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 4.20 4.25 [hinv]
O2 R5000SC 200MHz 1MB: 3.63 3.76
O2 R5000SC 180MHz 512K: 3.25 3.32 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 2.74 2.80
Indy R5000SC 180MHz 512K XZ (4GE): 2.73 2.78 [hinv]
Indy R4400SC 200MHz 1MB XZ (4GE): 2.73 2.78
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 2.72 2.78
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 2.65 2.69 [hinv]
Indy R4600PC 133MHz XZ (4GE): 2.62 2.64 [hinv]
Indy R5000PC 150MHz XZ (4GE): [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 1.85 1.89 [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 1.66 1.69
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 1.64 1.65 [hinv]
Indigo2 R4400SC/250 2MB XL24: 1.10 1.20
Indy R5000SC 180MHz 512K XL24: 1.06 1.20
Indy R4400SC/200 1MB XL24: 0.81 1.02
Indy R4600SC 133MHz 512K L2 XZ (4GE): [hinv]
Indy R5000PC 150MHz XL24: 0.79 0.87
Indigo R3000 33MHz XS24Z (1GE): 0.69 0.69 [hinv]
Indy R4600PC/133 XL24: 0.43 0.68
Indy R4600PC/100 XL24: 0.40 0.44
Indigo R3000 33MHz Entry: 0.19 0.20 [hinv]
Test M: spacestation.iv (10237 polygons). This is the famous model
from the /usr/share/data/models/vehicles directory [screenshot].
MW MW/W MW/HL SW SW/W SW/HL
Fuel R16K/700 4MB L2 V10: 194.81 112.44 ? 400.30 162.59 ? [hinv]
Fuel R14K/500 2MB L2 V10: 190.76 111.50 144.72 388.26 165.17 197.00 [hinv]
Onyx2 IR2E R10K/250 4MB L2 4RM9/64MB: 223.79 102.42 159.78 223.93 103.79 150.37 [hinv]
Onyx R10000SC/195 1MB L2 IR 1RM6/64MB: 161.99 130.17 87.49 162.87 127.78 82.98 [hinv]
Octane R12000SC/300 2MB L2 V6: 123.04 62.95 84.39 255.32 107.10 129.38
Octane2 R12000SC/400 2MB L2 MXE: 75.47 41.12 59.85 118.38 47.88 72.50 [hinv]
Onyx R4400SC/250 4MB L2 RE2 2RM4: 68.69 31.13 46.93 87.06 35.66 54.27 [hinv]
POWER Onyx R8000SC/90 4MB L2 RE2 2RM5: 68.13 31.13 47.02 84.67 35.48 53.92 [hinv]
Onyx R4400SC/200 4MB L2 RE2 2RM5: 66.68 31.16 46.75 82.59 35.42 53.93 [hinv]
Onyx R4400SC/150 1MB L2 RE2 4RM4: 59.68 32.31 48.52 62.19 35.48 53.68 [hinv]
POWER Onyx R8000SC/75 4MB L2 RE2 2RM5: 62.94 30.67 45.93 68.74 34.94 53.35 [hinv]
Indigo2 R10K/195 1MB L2 MaxIMPACT 4MB: 54.20 28.99 42.93 76.19 32.78 51.76 [hinv]
Indigo2 R10K/175 1MB L2 MaxIMPACT 4MB: 48.46 28.88 40.38 77.88 32.84 50.29 [hinv]
Onyx R4400SC/150 1MB L2 RE2 1RM4: 48.70 26.01 38.35 62.35 33.53 50.48 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 4MB: 42.70 28.49 39.02 61.29 33.27 49.24 [hinv]
Indigo2 R4400SC/250 2MB MaxIMPACT 1MB: 41.01 28.01 38.57 62.66 32.50 49.57
Indigo2 R4400SC/200 1MB MaxIMPACT 4MB: 40.38 28.23 38.54 53.18 31.96 46.60 [hinv]
POWER Series R4400SC/150 RE 4RM: 57.90 25.72 38.36 54.88 26.96 40.49 [hinv]
POWER Series R4400SC/150 RE 2RM: 51.62 25.16 38.21 56.48 26.97 40.34 [hinv]
POWER Series R4400SC/150 RE 1RM: 42.48 23.13 33.89 52.99 26.44 39.29 [hinv]
POWER Series R4000SC/100 RE 2RM: 33.78 24.94 35.56 34.27 25.78 36.73
POWER Series R4000SC/100 RE 1RM: 33.09 22.51 32.90 33.88 25.92 36.58
Indigo2 R4000SC/100 1MB MaxIMPACT 4MB: 23.76 27.26 32.38 26.08 30.22 34.83 [hinv]
O2 R12000SC 300MHz 1MB: 21.11 17.31 23.26 35.07 25.94 36.62 [hinv]
Indigo2 R4400SC 250MHz 2MB L2 SolidIMPACT: 20.66 19.35 15.36 45.41 28.10 18.50 [hinv]
Octane R12000SC/400 2MB L2 SE+Texture: 19.88 17.72 20.63 49.87 23.82 34.23
Octane R10000SC/250 1MB L2 SE+Texture: 20.11 16.44 19.70 49.37 20.66 30.69
Octane R10000SC/195 1MB L2 SE+Texture: 19.06 14.48 17.80 42.67 17.38 25.95
POWER Series R3000SC/33 RE 1RM: 16.61 19.35 22.73 16.59 22.66 23.95
O2 R12000SC 270MHz 1MB: 25.37 32.60 28.89 36.04 ? 25.32 [hinv]
O2 R5200SC 300MHz 1MB: 15.15 15.98 18.83 27.43 31.60 40.56 [hinv]
O2 R10000SC 195MHz 1MB: 14.43 15.29 18.72 23.58 27.29 34.65 [hinv]
O2 R5000SC 200MHz 1MB: 12.85 14.60 17.35 20.03 24.26 30.73
Indigo2 R4400SC/250 2MB GU1-Extreme (8GE): 13.26 14.68 12.22 19.49 23.13 19.08 [hinv]
O2 R5000SC 180MHz 512K: 11.88 14.44 17.18 15.94 20.66 25.63 [hinv]
Indigo2 R4400SC/200 1MB GU1-Extreme (8GE): 12.84 11.79 14.13 15.97 16.67 18.32 [hinv]
Indigo2 R4400SC/150 1MB GU1-Extreme (8GE): 11.40 10.48 12.23 16.98 16.54 19.51 [hinv]
Indigo2 R4400SC/250 2MB GR3-Elan (4GE): 6.22 6.12 6.87 8.95 9.29 10.80
Indigo R4000SC 100MHz 1MB GR2-Elan (4GE): 6.16 6.53 5.72 7.70 8.61 7.63 [hinv]
Indigo2 R4400SC/200 1MB GR3-Elan (4GE): 6.13 6.12 6.90 7.85 7.88 9.20
Indy R4400SC 200MHz 1MB XZ (4GE): 6.05 6.09 6.70 7.97 7.90 9.13
Indy R5000SC 180MHz 512K XZ (4GE): 6.05 5.95 6.75 7.83 7.84 9.13
Indy R4600PC 133MHz XZ (4GE): 6.04 5.61 6.48 8.07 7.81 9.10 [hinv]
Indigo R3000 33MHz GR2-Elan (4GE): 5.64 5.38 6.19 7.35 7.47 8.53 [hinv]
Indigo2 R4400SC/200 1MB GR3-XZ (2GE): 5.33 5.75 6.51 6.50 7.17 8.05
Indigo2 R4400SC/150 1MB GR3-XZ (2GE): 5.22 5.57 6.33 6.33 6.89 7.91 [hinv]
Indigo R4000SC/100 1MB GR2-XS24Z (1GE): 2.82 3.24 3.65 3.04 3.43 4.09 [hinv]
Indigo R3000 33MHz XS24Z (1GE): 2.80 3.30 3.70 2.90 3.56 4.05 [hinv]
Indigo2 R4400SC/250 2MB XL24: 2.44 2.63 2.63 4.21 4.68 5.17
Indy R5000SC 180MHz 512K XL24: 2.14 2.34 2.36 4.72 4.20 5.10
Indy R4400SC/200 1MB XL24: 1.60 2.21 2.32 5.58 4.49 5.92
Indy R5000PC 150MHz XL24: 1.61 1.81 1.77 3.50 3.25 3.73
Indy R4600PC/133 XL24: 1.03 1.31 1.33 2.81 3.02 3.50
Indy R4600PC/100 XL24: 0.95 1.13 1.14 2.34 2.44 2.88
Indigo R3000 33MHz Entry: 0.67 0.84 0.82 0.84 1.09 1.00 [hinv]
I may add further tests of other models at a later date. If you have
any suggestions, feel free to email me.
Info on the R3000SC POWER Series RE system, from Simon Pigot:
Here are some tests for 8 x 33MHz R3000 256k/SC with 256Mb RAM and RealityEngine with 1 x RM4. The tests were all done under Irix 5.3xfs with the Open Inventor 2.1.1 eoe plus the 5.3xfs Recommended patches. Certainly lags behind the R4000 and R4400 RE but still isn't too bad.... BTW: Tried quake on 6.2 RE with R4400/150 the other day - seems to work just fine. In fact it looks to me to be somewhere between O2 and Octane in its performance - not bad.
Speaking of Quake, an Onyx2 R10K/250 (4MB L2) IR2E 4RM/64 system I tested gives 30fps at 3840x1024 (3-screen RealityCentre).
A note about the V12 results compared to V6: from what I can gather, only the power of the GE subsystem was increased from V6/8 to V10/12. Thus, please make careful note of the following table when looking at the V6 and V12 numbers:
CAD-oriented, fewer GIS, imaging, heavy texture
hw features, not so oriented, more VRAM/TRAM,
much VRAM or TRAM. more features, dual-head, etc.
Old Series VPro V6 V8
Original GE speed
New Series VPro V10 V12
2X Faster GE speed
Some observations which result from the above:
Curiosities:
GRAPHICS TEST 2: Buttonfly Powerflip Models: rotation speed
(double-buffered).
These tests give the frames-per-second rates whilst spinning various objects that form part of the Buttonfly/Powerflip SGI demonstration program.
NOTE: The frame rate figures given here are for double-buffered tests, ie. the drawing of a 3D image is always synchronised with the frame-by-frame display of pixel data on the screen - how quickly this can be done depends on the monitor refresh rate. Thus, a system's graphics hardware renders a model only as quickly as the restrictions of synchronisation allow, the precise nature of which may vary between different models and can easily be strangely obscure. Thus, frame rates for double-buffered tests can never be more than the monitor refresh rate, are usually an integer divisor of the monitor rate, and frequently do not represent the maximum capabilities of a graphics system - however, such tests often do represent the realities of someone's work, eg. real-time 3D animation, but they should not be used as any indicator of the maximum performance a graphics subsystem.
By contrast, single-buffered tests (see GRAPHICS TEST 1 above) make no attempt to synchronise the drawing of the 3D image with the display refresh of the monitor. Thus, a system's graphics hardware renders a model as quickly as possible, which means observed frame rates will usually be an obscure fraction of the monitor refresh rate, but can also be many times the refresh rate if a particular graphics system is sufficiently powerful or if the model is sufficiently simple (ag. the arm2.iv model data shown above).
I present results of both types of test here in order to offer as broad a range of information as possible, and also to show very clearly how different these two types of test can be. For full details on this issue, and I strongly recommend you read this information before looking at the results below, see my page on The Effects of Double-Buffering on Animation Frame Rates. In general though, one should always use single- buffered tests to compare systems.
For each object, a brief description is given and the number of polygons it contains. Monitor settings were always 1280x1024 at 72Hz with 32+32 or the nearest-best. Exceptions: some Onyx2 tests were run at 60Hz HDTV resolution, the Indigo2 GR3-XZ was using a 60Hz monitor, and the Octane R12K/400 V6 was running at 1280x720 60Hz. Remember: the fps score in these double-buffered tests cannot be higher than the monitor refresh rate; thus, they should NOT be used as a measure of best possible graphics performance, and especially not since these are IrisGL tests anyway (newer systems are better at OpenGL; IrisGL calls must be converted to OpenGL calls using the IGLOO system first, which slows down newer systems).
Key:
MWin (MW) = Maximum Window (click top-right button to maximise view)
SWin (SW) = Small Window (default size of ivview)
/xx = xx is the rendering mode; if none is given then the mode is
'default', ie. texturing is ON if the model contains texture.
Otherwise, see SH, W, T or HL below.
SH = shaded
W = wireframe
T = transparent
HL = hidden-line
AA = Anti-Aliasing ON
noAA = Anti-Aliasing OFF
Test A: F15 WireDefs (this is a collection of lines, not polygons).
MWin/AA MWin/noAA SWin/AA SWin/noAA
Fuel R16000SC 700MHz 4MB V10: 72 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6 (low res): 60 60 60 60 [hinv]
O2 R12000SC 270MHz 1MB: 21 24 32 34 [hinv]
O2 R5200SC 300MHz 1MB: 19 21 24 24 [hinv]
O2 R10000SC 195MHz 1MB: 18 18 24 24 [hinv]
O2 R5000SC 200MHz 1MB: 14 16 17 17
O2 R5000SC 180MHz 512K: 12 13 14 14
Indigo2 R4400SC 200MHz 1MB GR3-XZ (2GE): 6 11 12 13
Indigo2 R4400SC 150MHz 1MB GR3-XZ (2GE): 6 11 10 13 [hinv]
Test B: shaded Beathoven (4998 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 72 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6 (low res): 60 60 60 60 60 60 [hinv]
O2 R5200SC 300MHz 1MB: 14 14 18 18 18 24 [hinv]
O2 R12000SC 270MHz 1MB: 14 14 18 18 18 23 [hinv]
O2 R10000SC 195MHz 1MB: 12 12 14 14 14 17 [hinv]
O2 R5000SC 200MHz 1MB: 12 11 14 14 14 14
O2 R5000SC 180MHz 512K: 10 10 11 12 12 13
Indigo2 R4400SC 200MHz 1MB GR3-XZ (2GE): 6 5 4 7 7 4
Indigo2 R4400SC 150MHz 1MB GR3-XZ (2GE): 6 5 4 7 7 4 [hinv]
Test C: shaded Ranger (14174 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 36 36 36 36 36 36 [hinv]
Octane R12000SC 400MHz 2MB V6 (low res): 20 20 20 20 20 20 [hinv]
O2 R12000SC 270MHz 1MB: 7 7 7 8 7 8 [hinv]
O2 R5200SC 300MHz 1MB: 6 6 7 7 7 8 [hinv]
O2 R10000SC 195MHz 1MB: 5 5 5 6 6 6 [hinv]
O2 R5000SC 200MHz 1MB: 5 5 5 5 5 5
O2 R5000SC 180MHz 512K: 4 4 4 4 4 5
Indigo2 R4400SC 200MHz 1MB GR3-XZ (2GE): 2 2 1 2 2 1
Indigo2 R4400SC 150MHz 1MB GR3-XZ (2GE): 2 2 1 2 2 1 [hinv]
Test D: shaded SGI Logo (864 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 72 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6 (low res): 60 60 60 60 60 60 [hinv]
O2 R5200SC 300MHz 1MB: 24 24 36 72 68 72 [hinv]
O2 R12000SC 270MHz 1MB: 24 24 36 54 36 72 [hinv]
O2 R10000SC 195MHz 1MB: 24 24 36 36 36 36 [hinv]
O2 R5000SC 200MHz 1MB: 24 24 36 36 36 36
O2 R5000SC 180MHz 512K: 24 24 24 36 36 36
Indigo2 R4400SC 200MHz 1MB GR3-XZ (2GE): 15 11 17 25 20 20
Indigo2 R4400SC 150MHz 1MB GR3-XZ (2GE): 17 10 18 24 18 23 [hinv]
Test E: brick-textured Beathoven (4998 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 36 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6: 60 60 30 60 60 60 [hinv]
O2 R5200SC 300MHz 1MB: 9 8 10 17 17 12 [hinv]
O2 R12000SC 270MHz 1MB: 9 8 10 16 15 12 [hinv]
O2 R10000SC 195MHz 1MB: 7 7 8 12 12 9 [hinv]
O2 R5000SC 200MHz 1MB: 7 7 5 11 10 5
O2 R5000SC 180MHz 512K: 6 6 4 8 8 4
Test F: brick-textured Ranger (14174 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 36 36 24 36 36 36 [hinv]
Octane R12000SC 400MHz 2MB V6: 20 20 20 20 20 20 [hinv]
O2 R12000SC 270MHz 1MB: 6 6 4 6 6 4 [hinv]
O2 R5200SC 300MHz 1MB: 6 6 4 6 6 4 [hinv]
O2 R10000SC 195MHz 1MB: 4 4 3 5 5 3 [hinv]
O2 R5000SC 200MHz 1MB: 3 3 2 4 4 2
O2 R5000SC 180MHz 512K: 3 3 1 3 3 1
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 72 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6: 60 60 60 60 60 60 [hinv]
O2 R12000SC 270MHz 1MB: 15 14 24 36 36 36 [hinv]
O2 R5200SC 300MHz 1MB: 15 14 24 36 36 36 [hinv]
O2 R10000SC 195MHz 1MB: 15 14 23 36 36 36 [hinv]
O2 R5000SC 200MHz 1MB: 15 14 16 34 27 17
O2 R5000SC 180MHz 512K: 13 12 14 24 24 17
Test H: Cafe-Mapped Beathoven (4998 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 36 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6: 60 60 30 60 60 60 [hinv]
O2 R12000SC 270MHz 1MB: 10 10 10 14 14 11 [hinv]
O2 R5200SC 300MHz 1MB: 10 10 9 14 14 10 [hinv]
O2 R10000SC 195MHz 1MB: 9 9 7 11 11 8 [hinv]
O2 R5000SC 200MHz 1MB: 7 7 4 8 8 4
O2 R5000SC 180MHz 512K: 6 6 3 7 7 4
Test I: Cafe-Mapped Ranger (14174 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 36 36 24 36 36 36 [hinv]
Octane R12000SC 400MHz 2MB V6: 20 20 20 20 20 20 [hinv]
O2 R12000SC 270MHz 1MB: 5 5 3 6 6 4 [hinv]
O2 R5200SC 300MHz 1MB: 5 5 3 5 5 3 [hinv]
O2 R10000SC 195MHz 1MB: 4 4 2 4 4 3 [hinv]
O2 R5000SC 200MHz 1MB: 3 3 1 3 3 1
O2 R5000SC 180MHz 512K: 2 2 1 2 2 1
Test J: Cafe-Mapped SGI Logo (864 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 72 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6: 60 60 60 60 60 60 [hinv]
O2 R12000SC 270MHz 1MB: 15 15 24 36 36 36 [hinv]
O2 R5200SC 300MHz 1MB: 14 14 24 36 36 36 [hinv]
O2 R10000SC 195MHz 1MB: 14 14 22 35 34 24 [hinv]
O2 R5000SC 200MHz 1MB: 14 14 14 24 24 17
O2 R5000SC 180MHz 512K: 14 12 14 24 24 15
Test K: powerflip (fighter model, 572 polygons).
MWin MWin/T MWin/W SWin SWin/T SWin/W
Fuel R16000SC 700MHz 4MB V10: 72 72 72 72 72 72 [hinv]
Octane R12000SC 400MHz 2MB V6 (low res): 60 60 60 60 60 60 [hinv]
O2 R12000SC 270MHz 1MB: 36 36 36 72 72 72 [hinv]
O2 R5200SC 300MHz 1MB: 36 36 36 72 72 72 [hinv]
O2 R10000SC 195MHz 1MB: 36 36 36 72 72 72 [hinv]
O2 R5000SC 200MHz 1MB: 36 36 36 72 72 72
O2 R5000SC 180MHz 512K: 36 36 36 36 36 68
Indigo2 R4400SC 200MHz 1MB GR3-XZ (2GE): 30 20 20 30 30 30
Indigo2 R4400SC 150MHz 1MB GR3-XZ (2GE): 24 20 24 36 36 28 [hinv]
IMAGING TEST 1: GIMP V1.0.4: Filters->Distorts->Ripple
The tests are of two types:
Note that these tests will not be listed as specifically integer or floating-point tests; rather, they will be grouped is image processing tests. This is because I have no way of knowing exactly what each test is doing in terms of int vs. fp calculations - this is especially true of the combinational scripts.
The results tables will include the disk type for each system when a TYPE A test is being described. For multi-CPU systems, only one CPU is used for these tests.
Results notes: the advantages of having a more modern system are clear: an Origin200 with a 180MHz R10K often beats an Indigo2 with a 195MHz R10K, ie. the Origin design gets more out of the chip technology. The benefits for rendering (eg. BMRT, PovRay, etc.) should be even greater (I hope to do some rendering benchmarks later in the year). However, the differences become less if a task is mostly integer processing, eg. the Origin is half way between I2 R10K/175 and R10K/195 for the Circuit test below, probably because that test is mostly integer processing, while the Lava test is likely a mix of int and fp.
Imaging Test 1 is TYPE A, as explained above.
The data file to be processed is the 48MB image file that comes with the O2 'roam' demonstration program, called 'sanFran4k.tif' (a large aerial photograph of San Fransisco, USA), but for the purposes of the test I convert the image to RGB format using imgworks. The test file is thus a 4096 by 4096 RGB image of size 50,332,160 bytes (ignoring the image file header, exactly 48MB).
Since TYPE A tests always involve big image files, it's important that a test system has a large amount of RAM. I always use at least 256MB RAM (128MB is definitely cutting it too close when working with 50MB+ images). Note this means that even an old Indigo could run such tests since Indigo can have up to 384MB RAM. A lot of RAM is required because of GIMP's UNDO buffers.
How to run the test: first download and install GIMP V1.0.4 if you don't already have it (NB: this test is using V1.0.4; the latest version - which was V1.2.0 the last time I checked - can be downloaded from the main GIMP site). You will also need to install at least GTK V1.0.6 in order to install this version of GIMP. Next, convert the sanFran4k.tif image to RGB using imgworks (get the file from the O2 demo CD, or borrowed from someone who has the demo installed). Then, load the image into GIMP; use the right mouse button to bring up the main menu and select 'Filters', followed by 'Distorts', and finally 'Ripple'. Get ready with your stop watch: click on OK and start the timing. When the operation has completely finished (the new image is fully redrawn on the screen, ie. one is able to bring up the right-button menu again), stop the timing. Be warned: this operation could take a very long time on a system such as a low-end Indy (perhaps 15 to 20 minutes), so be patient!
Test Results
Clock Total Time Disk
System CPU Speed L2 (mm:ss) Ref
Fuel R16000SC 700 4MB 00:39 - [hinv]
Octane R12000SC 400 2MB 01:13 12 [hinv]
Onyx2 R10000SC 250 4MB 01:53 7 [hinv]
O2 R12000SC 270 1MB 02:17 11 [hinv]
Origin200 R10000SC 180 1MB 02:31 2 [hinv]
Indigo2 R10000SC 195 1MB 02:49 9 [hinv]
O2 R5200SC 300 1MB 03:05 10 [hinv]
Indigo2 R10000SC 175 1MB 03:17 9 [hinv]
O2 R5000SC 200 1MB 03:45 1 [hinv]
O2 R5000SC 180 512K 04:32 1 [hinv]
Indigo2 R4400SC 250 2MB 04:32 3 [hinv]
Indy R5000SC 180 512K 04:18 13 [hinv]
Indy R5000SC 150 512K 04:49 13 [hinv]
Indy R4400SC 200 1MB 05:34 4 [hinv]
Indigo2 R4400SC 200 1MB 05:36 3 [hinv]
Indy R5000PC 150 - 05:48 6 [hinv]
Indy R4600SC 133 512K 07:01 4 [hinv]
Indigo R4400SC 150 1MB 07:29 8 [hinv]
Indigo2 R4400SC 150 1MB 07:40 3 [hinv]
Indy R4600PC 133 - ? 5
Indy R4600PC 100 - ? 5
Indigo2 R4000SC 100 1MB ?
Indy R4000SC 100 1MB ?
Indigo R4000SC 100 1MB 11:12 8 [hinv]
Disk types: 1. IBM DCHS04Y 3030 (4.5GB XFS)
2. SEAGATE ST39173W SUN9.0G2815 (9GB XFS)
3. 2 x IBM DNES-309170 SA30 (9.1GB XFS) (XFS system/data)
4. IBM DNES-309170 SA30 (9.1GB XFS)
5. IBM 0664 M1H 6E60 (XFS system), SEAGATE ST52160N 0344 (XFS data)
6. SEAGATE ST51080N0950 (1.0GB XFS system), IBM DNES-309170 SA30 (9.1GB XFS data)
7. IBM DGHS18Y 0190 (18GB XFS)
8. QUANTUM XP34550SLXQ1 (4GB XFS)
9. QUANTUM XP34300 S89C (4.5GB XFS)
10. SEAGATE ST318275SLC 001 18GB XFS
11. IBM DDRS-39130W 595D 9GB XFS
12. SEAGATE ST39204 3234 (8GB XFS)
13. SEAGATE ST32155N (2GB XFS)
Table 37: GIMP V1.0.4 Image Processing Test No. 1
Filters/Distorts/Ripple, 48MB 4K x 4K RGB image
IMAGING TEST 2: GIMP V1.0.4: Script-Fu->Decor->Lava
See Imaging Test 1 for an introduction into the nature of these tests, their purpose, rationale, terms and definitions, etc.
This test is a TYPE B test.
How to run the test: first download and install GIMP V1.0.4 if you don't already have it (NB: this test uses V1.0.4; the latest version can be downloaded from the main GIMP site). Run GIMP, select 'New' from the File menu, change the image size to 1024 by 1024 and click on OK. An empty image window will appear. Using the mouse right button, bring up the main menu and select Script-Fu, then Decor, then Lava. A settings panel will appear. Do not change the default settings. With your stop watch ready, click on OK and start the timing.
This is a tougher task than Imaging Test 1. If you're using a low-end Indy or a really old system like R4000/50 Indigo, be prepared to wait a long time for this test to finish.
Test Results
Clock Total Time
System CPU Speed L2 (mm:ss)
Fuel R16000SC 700 4MB 00:51 [hinv]
Octane R12000SC 400 2MB 01:23 [hinv]
Onyx2 R10000SC 250 4MB 02:15 [hinv]
O2 R12000SC 300 1MB 02:26 [hinv]
O2 R12000SC 270 1MB 02:40 [hinv]
Origin200 R10000SC 180 1MB 03:04 [hinv]
Indigo2 R10000SC 195 1MB 03:18 [hinv]
O2 R5200SC 300 1MB 03:21 [hinv]
Indigo2 R10000SC 175 1MB 03:25 [hinv]
O2 R5000SC 200 1MB 04:26 [hinv]
Indigo2 R4400SC 250 2MB 05:27 [hinv]
O2 R5000SC 180 512K 05:39 [hinv]
Indy R5000SC 180 512K 05:40 [hinv]
Indy R5000SC 150 512K 06:12 [hinv]
Indy R4400SC 200 1MB 07:43 [hinv]
Indigo2 R4400SC 200 1MB 07:50 [hinv]
Indy R4600SC 133 512K 07:57 [hinv]
Indy R5000PC 150 - 08:34 [hinv]
Indigo R4400SC 150 1MB 08:56 [hinv]
Indigo2 R4400SC 150 1MB 09:23 [hinv]
Indy R4600PC 133 - ?
Indy R4600PC 100 - ?
Indigo2 R4000SC 100 1MB ?
Indy R4000SC 100 1MB ?
Indigo R4000SC 100 1MB 15:26 [hinv]
Table 38: GIMP V1.0.4 Image Processing Test No. 2
Script-Fu/Decor/Lava, 1024 x 1024 image
IMAGING TEST 3: GIMP V1.0.4: Script-Fu->Decor->Circuit
See Imaging Test 1 for an introduction into the nature of these tests, their purpose, rationale, terms and definitions, etc.
This test is a TYPE B test.
How to run the test: first download and install GIMP V1.0.4 if you don't already have it (NB: this test uses V1.0.4; the latest version can be downloaded from the main GIMP site). Run GIMP, select 'New' from the File menu, change the image size to 1024 by 1024 and click on OK. An empty image window will appear. Using the mouse right button, bring up the main menu and select Script-Fu, then Decor, then Circuit. A settings panel will appear. Do not change the default settings. With your stop watch ready, click on OK and start the timing.
If you're using an older system, be prepared to wait a long time for this test to finish.
Test Results
Clock Total Time
System CPU Speed L2 (mm:ss)
Fuel R16000SC 700 4MB 00:42 [hinv]
Octane R12000SC 400 2MB 01:11 [hinv]
O2 R12000SC 300 1MB 01:49 [hinv]
Onyx2 R10000SC 250 4MB 01:50 [hinv]
O2 R12000SC 270 1MB 01:59 [hinv]
Indigo2 R10000SC 195 1MB 02:24 [hinv]
Origin200 R10000SC 180 1MB 02:40 [hinv]
Indigo2 R10000SC 175 1MB 02:52 [hinv]
O2 R5200SC 300 1MB 03:21 [hinv]
O2 R5000SC 200 1MB 03:56 [hinv]
Indy R5000SC 180 512K 04:09 [hinv]
O2 R5000SC 180 512K 04:12 [hinv]
Indigo2 R4400SC 250 2MB 04:27 [hinv]
Indy R5000SC 150 512K 05:11 [hinv]
Indy R5000PC 150 - 05:13 [hinv]
Indy R4400SC 200 1MB 05:24 [hinv]
Indigo2 R4400SC 200 1MB 05:26 [hinv]
Indy R4600SC 133 512K 06:10 [hinv]
Indigo2 R4400SC 150 1MB 07:10 [hinv]
Indigo R4400SC 150 1MB 07:22 [hinv]
Indy R4600PC 133 - ?
Indy R4600PC 100 - ?
Indigo2 R4000SC 100 1MB ?
Indy R4000SC 100 1MB ?
Indigo R4000SC 100 1MB 11:32 [hinv]
Table 39: GIMP V1.0.4 Image Processing Test No. 3
Script-Fu/Decor/Circuit, 1024 x 1024 image
MEMORY USAGE: Default 6.2/6.3/6.5 memory usage given varying RAM
resources.
Table 10 shows the amount of memory used by the OS and base user environment, rounded to the nearest MByte. To do this test yourself, login as guest to a default installation and run gmemusage (the figures don't include the memory used by gmemusage, but do include the IconCatalog window that appears by default).
The results do not denote the memory used by IRIX on its own, but rather the amount one would typically see taken up by the system as a whole after one has logged in, ie. this table is supposed to be an indicator of available resources to an every-day user. Note that: O2 uses a large percentage of the available memory because of the UMA architecture (main RAM used for video, etc.), the 6.3 OS disk I used was not a default installation, and the video mode was always 1280x1024 72Hz on each system (32+32 visual on O2).
O2: RAM 6.3 6.5 256MB: 67 69 192MB: 62 59 128MB: 52 54 64MB: 41 42 Indigo2: RAM 6.2 6.5 384MB: 43 62 320MB: 42 60 256MB: 41 59 192MB: 39 52 128MB: 33 42 96MB: 29 35 64MB: 23 31 32MB: 18 26 Indy: RAM 6.2 6.5 256MB: 44 54 192MB: 35 46 160MB: 32 40 128MB: 25 36 96MB: 23 32 64MB: 20 29 32MB: 17 26 Table 10: Memory Usage
CDROM TEST 1: Analysing the 6.5 OS installation times to compare
CDROM performance.
This analysis compares the time taken to install the various default IRIX 6.5 CDs for different SGIs with two different CDROM types: 2X and 32X, though the O2 tests use an internal 12X to compare to the 32X.
The focus is on two areas: how much better a faster CDROM is compared to a slower CDROM for a particular system, and how much better a given CDROM can be utilised given the presence of a better CPU.
Refer again to tables 1 and 2:
******* Sub-Task Completion Times *******
(16%) (51%) (86%) (91%) (94%) (100%)
CD Read Pre- Inst End of End of
ROM Inst inst F 1 Apps F 2 Tools Exit rqsall
Tools Ends Ends Ends Ends Ends Coms ELF libs
O2 R5000SC/200: 32 0:42 0:54 06:31 15:07 16:46 17:45 22:52 0:34:06
O2 R5000SC/200: 12 0:46 0:54 06:58 16:04 18:05 19:20 24:08 0:35:28
I2 R4400SC/250: 32 0:47 1:11 06:45 15:17 16:57 17:55 23:12 0:36:05
I2 R4400SC/250: 2 1:16 1:11 09:18 21:06 23:28 24:38 29:57 0:42:31
Indy R4400SC/200: 32 0:42 1:26 08:32 19:42 21:40 22:50 29:41 0:45:24
Indy R4400SC/200: 2 1:16 1:26 11:11 25:49 28:29 29:58 36:53 0:52:35
Indy R4600PC/133: 32 0:44 1:45 11:48 28:26 31:28 33:07 46:11 1:08:45
Indy R4600PC/133: 2 1:16 1:45 14:10 33:37 37:12 39:00 52:05 1:14:39
Indy R4600PC/100: 32 0:35 2:04 12:36 30:18 33:36 35:20 48:28 1:12:11
Indy R4600PC/100: 2 1:16 2:04 15:01 35:33 39:32 41:28 54:54 1:18:36
Table 1: Detailed Timings for a Default IRIX 6.5 OS Installation
CD PRE- INST- EXIT RQSALL
ROM INST F1 APPS F2 TOOLS COMS ELF
O2 R5000SC/200: 32 0:54 05:37 08:36 01:39 00:59 05:07 11:14
O2 R5000SC/200: 12 0:54 06:04 09:06 02:01 01:15 04:48 11:20
I2 R4400SC/250: 32 1:11 05:34 08:32 01:40 00:58 05:17 12:53
I2 R4400SC/250: 2 1:11 08:07 11:48 02:22 01:10 05:19 12:34
Indy R4400SC/200: 32 1:26 07:06 11:10 01:58 01:10 06:51 15:43
Indy R4400SC/200: 2 1:26 09:45 14:38 02:40 01:29 06:55 15:42
Indy R4600PC/133: 32 1:45 10:03 16:38 03:02 01:39 13:04 22:34
Indy R4600PC/133: 2 1:45 12:25 19:27 03:35 01:48 13:05 22:34
Indy R4600PC/100: 32 2:04 10:32 17:42 03:18 01:44 13:08 23:43
Indy R4600PC/100: 2 2:04 12:57 20:32 03:59 01:56 13:26 23:42
Table 2: Individual times for each installation step (IRIX 6.5)
This time, we're interested in the stages which do involve the CDROM, namely: reading the installation tools, installing F1, APPS, F2, and INST TOOLS.
Table 11 shows how much faster a 32X CDROM is compared to a 2X CDROM (12X for O2) for each installation step that uses the CDROM. The figures are obtained by dividing the faster time by the slower time for each system (ie. 100*(1-x) is the percentage improvement, where x is a number from the table).
INST-
F1 APPS F2 TOOLS AVERAGE
O2 R5000SC/200: 0.93 0.96 0.82 0.79 | 0.91
I2 R4400SC/250: 0.69 0.72 0.70 0.83 | 0.71
Indy R4400SC/200: 0.73 0.76 0.74 0.77 | 0.75
Indy R4600PC/133: 0.81 0.86 0.85 0.92 | 0.84
Indy R4600PC/100: 0.81 0.86 0.83 0.90 | 0.84
Table 11: Speedup using 32X vs. 2X CDROM (vs. 12X for O2)
(remember the above table only includes the installation steps which use the CDROM)
Obviously, the better the CPU, the bigger the speedup when switching to a faster CDROM. For low-end Indys, it really isn't worth upgrading to a very fast CDROM because the main CPU can't exploit it (don't be fooled by the 0.84 figure; remember that half the installation time is taken up by tasks that do not use the CDROM - these tasks do not change when using a faster CDROM, so the benefit of having a better CDROM is averaged out).
The O2 results compare a 32X to a 12X; the main conclusion to be drawn is that using a faster CDROM than the O2's internal 12X isn't really worth it, but it would be worth it if one has an older O2 system with a 4X internal CDROM.
As further evidence, Table 12 shows how much faster each system is compared to the Indy R4600PC/100 when using a 32X CDROM.
INST-
F1 APPS F2 TOOLS AVERAGE
O2 R5000SC/200: 0.53 0.49 0.50 0.57 | 0.51
I2 R4400SC/250: 0.53 0.48 0.51 0.56 | 0.50
Indy R4400SC/200: 0.67 0.63 0.60 0.67 | 0.64
Indy R4600PC/133: 0.95 0.94 0.92 0.95 | 0.94
Indy R4600PC/100: 1.00 1.00 1.00 1.00 | 1.00
Table 12: Performance compared to Indy R4600PC/100
when using a 32X CDROM.
Quite clearly, better processors can exploit faster CDROMs to a greater degree. Note especially the almost insignificant change when comparing R4600PC/133 to R4600PC/100. Now compare Table 12 to Table 13 below which shows the same calculations when using a 2X CDROM (O2 results excluded).
INST-
F1 APPS F2 TOOLS AVERAGE
I2 R4400SC/250: 0.63 0.58 0.59 0.60 | 0.60
Indy R4400SC/200: 0.75 0.71 0.67 0.77 | 0.73
Indy R4600PC/133: 0.96 0.95 0.90 0.93 | 0.94
Indy R4600PC/100: 1.00 1.00 1.00 1.00 | 1.00
Table 13: Performance compared to Indy R4600PC/100
for each installation stage when using a 2X CDROM.
Notice that the figures are higher than those in Table 12 as one might expect (ie. slower), with the exception of Indy R4600PC/133; ie. low-end systems are bottlenecked by their limited CPU power, not the theoretical performance of the CDROM.
Table 14 shows the proportion of time spent using and not using the CDROM. Clearly, whatever the combination of CDROM and CPU, roughly half the overall installation time does not involve using the CDROM. The last column (FACTOR) shows how much faster each installatin was compared to the slowest installation (R4600PC/100 with 2X CDROM).
TASKS NOT
CD USING USING TOTAL CDROM
ROM CDROM CDROM TIME USAGE (FACTOR)
O2 R5000SC/200: 32 16:51 17:15 0:34:06 49% 0.43
O2 R5000SC/200: 12 18:26 17:02 0:35:28 52% 0.45
I2 R4400SC/250: 32 16:44 19:21 0:36:05 46% 0.46
I2 R4400SC/250: 2 23:27 19:04 0:42:31 55% 0.54
Indy R4400SC/200: 32 21:24 24:00 0:45:24 47% 0.58
Indy R4400SC/200: 2 28:32 24:03 0:52:35 54% 0.67
Indy R4600PC/133: 32 31:22 37:23 1:08:45 46% 0.88
Indy R4600PC/133: 2 37:15 37:24 1:14:39 50% 0.95
Indy R4600PC/100: 32 33:16 38:55 1:12:11 46% 0.92
Indy R4600PC/100: 2 39:24 39:12 1:18:36 50% 1.00
Table 14: Task Summary for CDROM usage.
In order to come to a conclusion as to whether a CPU upgrade or a CDROM upgrade is a better choice, one must compare the total time taken for the installation when using a slow vs. fast CDROM (this is not the same as Table 11 which only covered the stages which used the CDROM). The results for this are in table 15.
OVERALL
SPEEDUP
O2 R5000SC/200: 0.96
I2 R4400SC/250: 0.85
Indy R4400SC/200: 0.86
Indy R4600PC/133: 0.92
Indy R4600PC/100: 0.92
Table 15: Overall speedup using
32X vs. 2X CDROM (vs. 12X for O2)
The results are clear: if one has a low-end CPU, upgrading to a fast CDROM does not improve software installation speed because the main CPU isn't good enough to take advantage of it; plus, the portion of the installation which doesn't use the CDROM is going to be pretty slow anyway. Only when one moves to CPUs which have reasonable clock speeds and L2 cache does one see any real benefit from using a fast CDROM.
One can also glean from Table 15 that it isn't worth using a faster CDROM when installing software on an O2 which has a 12X internal CDROM, but it would be worth it if the O2 was an older model with a 4X CDROM. In case you've never tried, external CDROMs are handled just fine - one is given a list to choose from on the software installation menu, and IRIX shows numerically labelled icons on the GUI desktop (CDROM1, CDROM2, etc.)
This has been a detailed analysis, but it's not the whole story. What about other types of operation which use a CDROM? What about copying files straight to disk? See the next section for further CDROM tests.
CDROM TEST 2: Copying a large file straight to disk.
In this test, I timed how long the various systems took to copy the main 179685K (175.5MB) Quake2 pak0.pak data file from the Quake2 CDROM straight to disk. The file was always copied to /var/tmp, and all tests were performed using IRIX 6.5. Data rates are in K/sec. The command sequence used to run the test, which was entered into a raw xterm where the current working directory was /CDROM/install/data/baseq2, was:
date && cp pak0.pak /var/tmp && date
Results for Origin200 R10000SC 180MHz (1MB L2):
CDROM Time Data
Speed (m:ss) Rate
2X
32X 1:15 2396
Table 40
Results for O2 R5000SC 200MHz (the 12X is the O2's internal CDROM):
CDROM Time Data
Speed (m:ss) Rate
2X 5:01 597
12X 3:22 890
32X 2:30 1198
Table 16
Results for Indigo2 250MHz R4400SC:
CDROM Time Data
Speed (m:ss) Rate
2X 4:59 601
32X 1:34 1912
Table 17
Results for Indy 200MHz R4400SC:
CDROM Time Data
Speed (m:ss) Rate
2X 5:59 501
32X 2:30 1198
Table 18
Results for Indy 133MHz R4600PC:
CDROM Time Data
Speed (m:ss) Rate
2X 6:47 442
32X 4:35 653
Table 19
Results for Indy 100MHz R4600PC:
CDROM Time Data
Speed (m:ss) Rate
2X 6:58 430
32X 4:27 673
Table 20
As shown in CDROM TEST 1, Tables 19 and 20 prove that a faster CDROM
for a system with a slow CPU is barely worth the expense. But far
more interesting are the following two observations:
My theory is this: copying a file from a CDROM merely involves getting the data as fast as possible through the main CPU, ie. into the registers, out to the disk (loads and stores). These operations usually only take a single clock cycle. Thus, the design of the CPU won't affect how quickly the data is copied, ie. it's the raw clock speed that matters, assuming the CPU is of the type which does the generic 1 int operation per clock tick. This is why the Indy R4400SC/200 copied the file in exactly the same amount of time as the O2 R5000SC/200 (for the 32X CDROM) - the two CPUs run at the same clock speed. On the other hand, tasks like installing software require the CPU to process the incoming data in some way. How well the CPUs do this depends on their quality of design, so the R5000 performs better. Well, that's my theory anyway.
Tables 21 and 22 show the data from tables 16 to 20 reorganised to show the performance for each CDROM type separately (the 12X results for O2 have been excluded).
Time Data
(m:ss) Rate
O2 R5000SC 200MHz 1MB: 5:01 597
Indigo2 250MHz R4400SC: 4:59 601
Indy 200MHz R4400SC: 5:59 501
Indy 133MHz R4600PC: 6:47 442
Indy 100MHz R4600PC: 6:58 430
Table 21: 2X CDROM Performance
Copying a large file to disk.
Time Data
(m:ss) Rate
O2 R5000SC 200MHz 1MB: 2:30 1198
Indigo2 250MHz R4400SC: 1:34 1912
Indy 200MHz R4400SC: 2:30 1198
Indy 133MHz R4600PC: 4:35 653
Indy 100MHz R4600PC: 4:27 673
Table 22: 32X CDROM Performance
Copying a large file to disk.
CDROM TEST 3: Copying many smaller files straight to disk.
This test uses the IRIS Development Option 7.1 CD for IRIX 6.3, which has an 'insight' directory containing approximately 2200 files, representing some 76289K (74.5MB) of data, giving an average file size of 35K. I timed how long the various systems took to copy the insight directory straight to disk. The destination was always /var/tmp/t (an empty directory), and all tests were performed using IRIX 6.5. Data rates are in K/sec. The command sequence used to run the test, which was entered into a raw xterm where the current working directory was /CDROM, was:
date && tar cBpf - insight | ( cd /var/tmp/t ; tar xBpf - ) && date
Results for Origin200 R10000SC 180MHz (1MB L2):
CDROM Time Data
Speed (m:ss) Rate
2X
32X 1:18 978
Table 41
Results for O2 R5000SC 200MHz (the 12X is the O2's internal CDROM):
CDROM Time Data
Speed (m:ss) Rate
2X 4:46 267
12X 2:24 530
32X 1:48 706
Table 27
Results for Indigo2 250MHz R4400SC:
CDROM Time Data
Speed (m:ss) Rate
2X 4:50 263
32X 1:48 706
Table 28
Results for Indy 200MHz R4400SC:
CDROM Time Data
Speed (m:ss) Rate
2X 5:02 253
32X 2:03 620
Table 29
Results for Indy 133MHz R4600PC:
CDROM Time Data
Speed (m:ss) Rate
2X 5:22 237
32X 3:24 374
Table 30
Results for Indy 100MHz R4600PC:
CDROM Time Data
Speed (m:ss) Rate
2X 5:27 233
32X 3:26 370
Table 31
Compared to copying a single large file, the average transfer rate is
clearly much reduced. So much so in fact that, when transferring many
small files, a system can easily display an average transfer rate for
a 32X CDROM that is lower than a 2X CDROM would give when
transferring a single large file. Thus, don't assume a fast
CDROM is going to display a level of performance that's anywhere near
its theoretical maximum - it all depends on the task in question.
Also, faster systems do proportionally better than slower systems when transferring lots of small files, ie. the better main CPU helps alot in dealing with the file-creation overhead. Here are the summary tables:
Time Data
(m:ss) Rate
O2 R5000SC 200MHz 1MB: 4:46 267
Indigo2 250MHz R4400SC: 4:50 263
Indy 200MHz R4400SC: 5:02 253
Indy 133MHz R4600PC: 5:22 237
Indy 100MHz R4600PC: 5:27 233
Table 32: 2X CDROM Performance
Copying many small files to disk.
Time Data
(m:ss) Rate
O2 R5000SC 200MHz 1MB: 1:48 706
Indigo2 250MHz R4400SC: 1:48 706
Indy 200MHz R4400SC: 2:03 620
Indy 133MHz R4600PC: 3:24 374
Indy 100MHz R4600PC: 3:26 370
Table 33: 32X CDROM Performance
Copying many small files to disk.
Compare Table 33 to Table 22, and Table 32 to Table 21.
DAT TEST 1: Archiving a single large file to DAT (DDS1 vs. DDS3).
In this test, I timed how long the various systems took to archive the main 179685K (175.5MB) Quake2 pak0.pak data file to DAT. All tests were performed using IRIX 6.5. Data rates are in K/sec. The command sequence used to run the test, which was entered in a raw xterm, was:
timex tar cv pak0.pak
Table 23 shows the results for archiving the pak0.pak file. The DDS1
column is not yet complete, mainly due to lack of time.
DDS1 Python DDS1 Python DDS3 Sony
25601-XXX2.75 28388-XXX5.AC SDT9000
time / rate time / rate time / rate
O2 R5000SC/200: 8:26 / 355 2:00 / 1497 [hinv]
Indigo2 R4400SC/250: 16:48 / 178 2:07 / 1415
Indy R4400SC/200: 2:11 / 1372
Indy R4600PC/133: 3:23 / 885
Indy R4600PC/100: 3:14 / 926
Table 23: DDS1 vs. DDS3 DAT performance for
archiving a large file to DAT.
Just as with the CDROM tests, slower CPUs cannot fully exploit a fast DAT drive.
One might also expect the DDS3 results to show something similar to CDROM TEST 2, ie. transfer rate depends on raw clock speed. However, the figures are somewhat more even for archiving a large file to DAT. This could be because the DAT drive often takes some time to begin and end the archiving operation, but could also be because of architectural differences in I/O systems. Actually, the tests as a whole don't last very long - I may rerun the DDS3 tests with a much larger file, eg. the /CDROM/install/demos_O2.sw file O2 Demos 1.1.1 CD (the file is 498466056 bytes, ie. 475MB).
Note that I also have a DDS4 I could test, but I can't say when I'd be able to do this.
DAT TEST 2: Extract a single large file from DAT (DDS1 vs. DDS3).
In this test, I timed how long the various systems took to extract the main 179685K (175.5MB) Quake2 pak0.pak data file from DAT. All tests were performed using IRIX 6.5. Data rates are in K/sec. The command sequence used to run the test, which was entered in a raw xterm, was:
timex tar xv
Table 24 shows the results for extracting the pak0.pak file. The DDS1
column is not complete, mainly because I don't now have a DDS1 to run
tests with.
DDS1 Python DDS1 Python DDS3 Sony
25601-XXX2.75 28388-XXX5.AC SDT9000
time / rate time / rate time / rate
O2 R5200SC 300MHz 1MB: 7:59 / 375
O2 R5000SC/200: 1:54 / 1576
Indigo2 R4400SC/250: 17:19 / 173 1:53 / 1590
Indy R4400SC/200: 1:55 / 1563
Indy R4600PC/133: 3:56 / 761
Indy R4600PC/100: 3:47 / 792
Table 23: DDS1 vs. DDS3 DAT performance for
extracting a large file from DAT.
Again, the slower CPUs cannot exploit a fast DAT drive.
And as with DAT TEST 1, the top-three DDS3 test times are too similar to judge any performance differences which may exist between the systems, but there is another possibility: the data is being transferred at close to the maximum possible speed of the DDS3 drive. It's difficult to say. I will rerun the tests at a later date using a much larger file.
DAT TEST 3: Archiving a large number of small files to DAT (DDS1
vs. DDS3).
This test measures DAT performance when archiving a large number of small files. It's difficult to construct a standard test for this, so my solution was to use the data in the /usr/share directory which is created after a default IRIX 6.5 installation. Even so, different systems end up with slight differences in the size of /usr/share - this is because of differences in default software selections during the installation process. Thus, for this test, it is the data transfer rate which is more important.
The command used to time this test, which was executed in a raw xterm from within the /usr/share directory, was:
timex tar c .
In Table 24 below, the times are also included for reference, but do
not use the times for comparisons between systems. O2 and Indigo2
GR3-Elan end up with 90MB in a default /usr/share, while Indy has its
default /usr/share taking up 74MB. What matters is the aggregate data
transfer rate.
DDS1 Python DDS1 Python DDS3 Sony
25601-XXX2.75 28388-XXX5.AC SDT9000
rate (time) rate (time) rate (time)
O2 R5200SC 300MHz 1MB: 238 (6:27)
O2 R5000SC/200: 819 (1:53)
Indigo2 R4400SC/250: 194 (07:56) 819 (1:53)
Indy R4400SC/200: 455 (2:46)
Indy R4600PC/133: 326 (3:51)
Indy R4600PC/100: 356 (3:32)
Table 24: DDS1 vs. DDS3 performance for
archiving many small files to DAT.
The most interesting observations here are:
DAT TEST 4: Extracting a large number of small files from DAT
(DDS1 vs. DDS3).
This test is the opposite of DAT TEST 3 and is executed in much the same way. The command used to time this test, which was executed in a raw xterm from within an empty directory in /var/tmp, was:
timex tar x
Again, since /usr/share varies in size between systems, it's the transfer rates which are important here, ie. do not use the times shown in Table 25 to make system comparisons (they're included only for reference). I have not yet performed the DDS1 tests.
DDS1 Python DDS1 Python DDS3 Sony
25601-XXX2.75 28388-XXX5.AC SDT9000
rate (time) rate (time) rate (time)
O2 R5200SC 300MHz 1MB: 394 (6:05)
O2 R5000SC/200: 561 (2:45)
Indigo2 R4400SC/250: 638 (2:25)
Indy R4400SC/200: 379 (3:19)
Indy R4600PC/133: 258 (4:53)
Indy R4600PC/100: 282 (4:28)
Table 25: DDS1 vs. DDS3 performance for
extracting many small files from DAT.
Observations:
These results, given in Table 26 below, are really just for reference, but they are interesting in that the figures are higher than those in Table 24. This is probably because the filesystem as a whole contains many large files which causes much less overhead, averaging out those parts of the filesystem which contain lots of small files, most of which I suspect reside in /usr/share.
Again, it's the transfer rate that is important (default installation sizes vary between systems), but times are included for reference.
Rate Time Size
(K/sec) (mm:ss) (K)
O2 R5000SC/200: 1438 09:05 783820
Indigo2 R4400SC/250: 1369 09:21 768009
Indy R4400SC/200: 947 12:41 720853
Indy R4600PC/133: 682 17:39 721792
Indy R4600PC/100: 676 17:48 721792
Table 26: DDS1 vs. DDS3 performance for archiving
the entire default 6.5 filesystem to DAT
AUDIO TEST 1: Parametric Equalisation and Normalization of a Music
CD Track.
This test uses an audio track from a music CD. Two separate sub-tests were conducted using the SoundTrack program: a 3-notch parametric equalisation is performed on the track, and a Normalization.
The track used is the title track from the CD entitled, 'Force Majeure', by Tangerine Dream. The file from the CD is Track01.aiff, which is 18 minutes 24 seconds of 44.1KHz 16bit 2-channel data, totalling 190140K (185.7MB).
To perform the parametric test yourself, run SoundTrack and 'import' the file (copy the file from the CD onto disk first). Choose 'Select to End' from the Edit menu. Now choose 'Parametric Equalizer' from the 'Effects' menu and enter these settings:
Band 1: Frequency: 4000
Gain (dB): 10
Bandwidth: 2000
Band 2: Frequency: 10000
Gain (dB): -10
Bandwidth: 2000
Band 2: Frequency: 16000
Gain (dB): 10
Bandwidth: 2000
With your stopwatch ready, click on 'Ok' and time how long it takes to complete the task. The band-filter settings I've used are arbitrary, but they do ensure that the filters overlap on the graph.
To perform the normalization test, run SoundTrack and 'import' the file. Choose 'Select to End' from the Edit menu. With your stopwatch ready, choose 'Normalization' from the 'Effects' menu and time how long it takes to complete the task.
Of course, if you don't have the CD I've used, you can't run the specific test shown here, but any music track of similar length should give comparable results. Note that I originally used a shorter track for this test, but the old test was performed too quickly to give meaningful results. Tangerine Dream often wrote long tracks, so now the test lasts long enough to be useful as a benchmark (this new test takes about five times longer to run). A lengthy test makes it easier to be sure that the task is not being overly influenced by the speed of the system disk, although disk type can make a difference in this test.
Note that something odd happened when testing the Fuel/500; the Normalization test dind't work. I'll try again another day, so in the meantime I've included an estimate based on the difference betweem the Fuel/500 and the Fuel/700 for the P.E. test.
Here are the results, in minutes and seconds:
Parametric
Equalization Normalization
Fuel R16000SC 700MHz 4MB L2: 0:27 0:29
Fuel R14000SC 500MHz 2MB L2: 0:38 0:41 [est]
O2 R10000SC 195MHz 1MB L2: 2:06 2:14
Origin200 R10000SC 180MHz 1MB L2: 2:20 2:20
Indigo2 R4400SC 250MHz 2MB L2: 4:44 5:07
O2 R5000SC 200MHz 1MB L2: 4:50 5:21
Indy R4400SC 200MHz 1MB L2: 5:52 6:18
Indy R5000SC 180MHz 512K L2: 6:30 7:23
Indy R5000PC 150MHz: 8:20 8:40
Indy R4600PC 133MHz: 10:10 10:33
Indy R4600PC 100MHz: 11:39 11:44
I intend to conduct many more performance tests in the future, including tests which involve floating-point calculations:
My goal is to provide so much data that you can get a very broad picture of a system's capabilities from a wide variety of metrics, rather than just relying on one or two synthetic benchmarks such as SPEC95 or Viewperf.
I welcome your feedback! :)
Ian.