SATA3/SATA2 Chipset Comparison Tests Ian Mapleson Last Change: 27/Apr/2013, 0130hrs P67 Board (Intel SATA2/3): ASUS Maximus IV Extreme, i7 2700K @ 5.0GHz, 16GB DDR3/2133. X58 Board (Marvell SATA3 and Intel SATA2): Asrock X58 Extreme6, XEON X5570 @ 3.5GHz, 24GB DDR3/1600. SSD Used: OCZ Vector 256GB, Firmware 2.0. AS-SSD/CDM/Atto/PCMark7 tests done with a 200GB partition. Both systems run Win7 Pro 64bit. ASUS ASUS Asrock Asrock Asrock Asrock Intel P67 Intel P67 Marvell 9120 Marvell 9120 Marvell 9128 Intel X58 SATA3 SATA2 1.2.0.1016 1.2.0.1032 1.2.0.1032 SATA2 HDTach Maximum Read: 375.00 213.15 302.00 293.00 302.00 222.00 HDtach Maximum Write: 330.00 206.80 200.00 200.00 172.00 232.00 HDTach Average Read: 342.40 204.20 288.10 279.90 294.20 221.30 HDTach Average Write: 273.20 189.90 178.50 178.80 161.90 210.30 AS-SSD Sequential Read: 516.72 266.53 367.33 322.72 365.41 266.67 AS-SSD Sequential Write: 503.12 255.92 233.27 237.42 185.78 257.68 AS-SSD 4K Read: 24.99 23.05 24.10 24.14 24.27 27.19 AS-SSD 4K Write: 68.01 54.28 60.70 61.88 59.98 98.21 AS-SSD 4K/64Thrd Read: 348.38 190.32 261.64 257.42 197.83 200.31 AS-SSD 4K/64Thrd Write: 280.04 164.37 167.09 169.88 95.38 182.34 AS-SSD Access Time Read: 0.053 0.063 0.056 0.055 0.055 0.038 AS-SSD Access Time Write: 0.052 0.060 0.051 0.051 0.050 0.030 AS-SSD Read Score: 425 240 322 314 259 254 AS-SSD Write Score: 398 244 251 255 174 306 AS-SSD Overall Score: 1048 615 740 734 565 701 CDM Sequential Read: 521.90 276.70 397.20 398.20 380.20 278.00 CDM Sequential Write: 519.60 267.10 251.50 252.90 201.50 269.30 CDM 512K Read: 417.80 249.90 343.00 342.50 337.50 252.30 CDM 512K Write: 508.40 264.10 248.10 249.80 200.60 268.20 CDM 4K Read: 26.65 25.01 25.95 26.04 26.15 29.12 CDM 4K Write: 77.68 64.60 72.15 73.16 69.93 120.10 CDM QD32 4K Read: 370.00 199.40 279.10 281.10 215.40 213.30 CDM QD32 4K Write: 341.70 186.40 193.20 194.90 106.50 206.60 Atto 4K Read: 245760 126030 198656 200724 127680 193536 Atto 4K Write: 311296 200704 178176 175104 103166 198656 Atto 64K Read: 534199 272399 378092 380838 347419 274411 Atto 64K Write: 486653 258269 248477 250256 193226 270415 Atto 512K Read: 555461 275457 405638 367491 387404 274976 Atto 512K Write: 528936 269326 250256 252061 197471 272121 Atto 8MB Read: 552762 281822 406720 379414 389742 282192 Atto 8MB Write: 533536 270804 250289 252052 197924 272177 PCMark7 Secondary Storage Score: 5420 5073 - 5195 5121 5104 Windows Defender: 5.61 5.54 - 5.54 5.54 5.58 Importing Pictures: 30.70 26.35 - 26.88 24.92 26.40 Video Editing: 23.01 21.98 - 22.42 22.40 22.09 Windows Media Center: 8.27 8.14 - 8.15 8.11 8.15 Adding Music: 1.41 1.41 - 1.40 1.40 1.41 Starting Applications: 57.87 47.57 - 53.22 52.21 48.33 Gaming: 16.91 16.31 - 16.53 16.54 16.42 PCMark7 Raw Secondary Storage: 6136 4326 - 4603 4329 4579 Windows Defender (raw): 60.34 53.19 - 53.21 53.40 57.52 Importing Pictures (raw): 152.80 83.84 - 89.47 70.95 84.42 Video Editing (raw): 211.97 148.24 - 170.83 169.66 153.09 Windows Media Center (raw): 385.14 220.07 - 223.63 197.63 229.20 Adding Music (raw): 117.05 91.40 - 86.78 83.18 108.57 Starting Applications (raw): 103.87 74.79 - 89.79 86.94 76.69 Gaming (raw): 117.82 93.80 - 101.67 102.07 97.87 Conclusions ^^^^^^^^^^^ The newer 1.2.0.32 drivers for Marvell don't give better results than the previous 1016 drivers. The Marvell 9128 chip has signifantly worse write performance than the 9120. Intel's P67/SATA3 is massively better than the Marvell SATA3 chips. Even Intel's X58/SATA2 is often faster than either Marvell SATA3 chip. The X58/SATA2 gives surprisingly strong 4K performance for AS-SSD/CDM/Atto, beating the Marvell SATA3 chips, though this isn't reflected so much in the PCMark7 results. Strangely, X58/SATA2 is also noticeably better than P67/SATA2 aswell, again especially for 4K performance (check the AS-SSD and CDM 4K results). Very odd. NOTE 1: I didn't start the PCMark7 testing until after I'd updated the Marvell drivers on the Extreme6, thus the lack of PCMark7 data for the Marvell 9120 with 1016 drivers. Downgrading Marvell drivers is a bad idea. Doesn't matter though, the data would basically have been the same as for the 9120 with the 1032 drivers. NOTE 2: although these boards have different levels of CPU power, this shouldn't affect the PCMark7 results much as I've only run the Secondary Storage tests, which don't depend on CPU power. I could tell this was true because the 2700K's PWM fans on the HS never sped up from idle during testing.