Whilst temporarily having access to two different O2 systems, one using a 512K L2 R5000SC/180MHz CPU and the other using a 1MB L2 R5000SC/200MHz CPU, I took the opportunity to compare the two by running a number of simple tasks on both and timing the results.
The hinv outputs for the two systems are given at the end of this page.
When comparing the performance measurements, it's worth noting that if a performance improvement was due to clock speed alone, then one would expect an 11% increase on average. Thus, if some tests show better results than this, then it's a clear indication that the larger L2 cache of the 200MHz R5000 is helping as well.
Obviously, if I ever get an opportunity to use an R10000 O2, I will run the same tests.
Some of the tests used a collection of RGB frames. I have them available if you should want to run any of these tests on your own system, but you'll need to find a way of downloading 563MB of data (ouch!).
Movie Conversion.
A sequence of 659 frames was captured from a PAL video tape at full-size (768x576) full-frame-rate (25fps) and stored as RGB images (each image is 1327616 bytes). These frames were converted into half-size and quarter-size, the new frames stored in two separate directories.
Note that the processing times listed do not include the 'Scanning Input' stage. All times were obtained by using dmconvert's -v option, except where otherwise stated.
Test 1: Converting 659 uncompressed RGB frames into an MJPEG QuickTime movie file. Each frame is an individual image file. Command used in each directory (split over two lines for readability):
dmconvert -v -f qt -n ###.rgb,start=000,end=658,step=1 -p video,rate=25,
il=none,size=orig,comp=jpeg,engine=sw,squal=1.0,tqual=1.0 ###.rgb out.mov
Results:
FULL-SIZE (768x576)
Elapsed Time (min:sec)
R5000/180: 8:51
R5000/200: 7:48
Compression ratio achieved: 4.7:1
Percentage improvement: 12%
HALF-SIZE (384x288)
Elapsed Time (min:sec)
R5000/180: 2:52
R5000/200: 2:44
Compression ratio achieved: 4.1:1
Percentage improvement: 5%
QUARTER-SIZE (192x144)
Elapsed Time (min:sec)
R5000/180: 0:56
R5000/200: 0:52
Compression ratio achieved: 3.1:1
Percentage improvement: 7%
Test 2: Converting 659 uncompressed RGB frames into a QuickTime Video
movie file. Command used in each directory:
dmconvert -v -f qt -n ###.rgb,start=000,end=658,step=1 -p video,rate=25, il=none,size=orig,comp=qt_video,engine=sw,squal=1.0,tqual=1.0 ###.rgb out.mov
Results (gr_osview showed full-size conversion to be significantly I/O-bound):
FULL-SIZE (768x576)
Elapsed Time (min:sec)
R5000/180: 13:44
R5000/200: 13:32
Compression ratio achieved: 7.3:1
Percentage improvement: 1.5%
HALF-SIZE (384x288)
Elapsed Time (min:sec)
R5000/180: 3:57
R5000/200: 3:36
Compression ratio achieved: 7.3:1
Percentage improvement: 9%
QUARTER-SIZE (192x144)
Elapsed Time (min:sec)
R5000/180: 1:08
R5000/200: 0:53
Compression ratio achieved: 7.2:1
Percentage improvement: 22%
Test 3: Converting 659 uncompressed RGB frames into a Cinepak
QuickTime movie file. Command used in each directory:
dmconvert -v -f qt -n ###.rgb,start=000,end=658,step=1 -p video,rate=25, il=none,size=orig,comp=qt_cvid,engine=sw,squal=1.0,tqual=1.0 ###.rgb out.mov
Results:
FULL-SIZE (768x576)
Elapsed Time (hours:min:sec)
R5000/180: 1:28:27 (*)
R5000/200: 1:18:40 (#)
Percentage improvement: 11%
QUARTER-SIZE (192x144)
Elapsed Time (min:sec)
R5000/180: 7:06
R5000/200: 6:06
Compression ratio achieved: 13.9:1
Percentage improvement: 14%
(*) ESTIMATED. It was going to take too long. Cut short and elapsed time (0:4:58) added to estimated remaining time (1:23:29).
(#) ESTIMATED. Cut short and elapsed time (0:5:08) added to estimated remaining time (1:13:32).
Test 4: Converting 100 uncompressed RGB frames into an MPEG movie
file. MediaConvert used with default settings but with 25fps. Times
obtained by stopwatch. Unfortunately, I only had time to run the
quarter-size test.
Results:
QUARTER-SIZE (192x144)
Elapsed Time (min:sec)
R5000/180: 1:14
R5000/200: 1:04
Compression ratio achieved: 14.2:1
Percentage improvement: 14%
Test 5: UltraSound demo (Colour Doppler, Animate, M-Mode, time trace
in middle of doppler).
Performance
R5000/180: 2.23 / 2.40
R5000/200: 2.58 / 2.68
Video: MVP unit 0 version 1.4 AV: AV1 Card version 1, O2Cam type 1 version 0 connected. FLASH PROM version 4.7 On-board serial ports: 2 On-board EPP/ECP parallel port 1 180 MHZ IP32 Processor FPU: MIPS R5000 Floating Point Coprocessor Revision: 1.0 CPU: MIPS R5000 Processor Chip Revision: 2.1 Data cache size: 32 Kbytes Instruction cache size: 32 Kbytes Secondary unified instruction/data cache size: 512 Kbytes on Processor 0 Main memory size: 256 Mbytes Iris Audio Processor: version A3 revision 0 Integral Ethernet: ec0, version 1 CDROM: unit 4 on SCSI controller 0 Disk drive: unit 1 on SCSI controller 0 CRM graphics installed Integral SCSI controller 1: Version ADAPTEC 7880 Integral SCSI controller 0: Version ADAPTEC 7880
R5000/200 system used:
Video: MVP unit 0 version 1.4 AV: AV1 Card version 1, O2Cam type 1 version 0 connected. FLASH PROM version 4.4 On-board serial ports: 2 On-board EPP/ECP parallel port 1 200 MHZ IP32 Processor FPU: MIPS R5000 Floating Point Coprocessor Revision: 1.0 CPU: MIPS R5000 Processor Chip Revision: 2.1 Data cache size: 32 Kbytes Instruction cache size: 32 Kbytes Secondary unified instruction/data cache size: 1 Mbyte on Processor 0 Main memory size: 256 Mbytes Iris Audio Processor: version A3 revision 0 Integral Ethernet: ec0, version 1 Tape drive: unit 6 on SCSI controller 1: DAT CDROM: unit 4 on SCSI controller 0 Disk drive: unit 2 on SCSI controller 0 Disk drive: unit 1 on SCSI controller 0 CRM graphics installed Integral SCSI controller 1: Version ADAPTEC 7880 Integral SCSI controller 0: Version ADAPTEC 7880 Vice: TRE