Question Strange SATA/AHCI perf numbers

Apr 5, 2020
22
0
10
Hello,

I own an oldish Crucial 64GB SSD and a Seagate firecuda 2TB SSHD. Running win10, Ryzen 5 2600, MSI B450 gaming + max MB, 16 GB DDR4-3000 RAM.
Both drives are SATA, AHCI is enabled in the BIOS, I use the cables that came with the MB (which is new - installed last week).

I find disk performance below par. Out of curiosity I compared perf, using CrystalDiskMark, with the recommended MS Std AHCI drivers and the older AMD AHCI drivers that AMD recommence not to use. Numbers below.

Not much of a difference for the SD, which performs OK. However the numbers don't make sense for the SSHD. 4 times higher random write speed with the AMD driver? But twice lower random read speed? And higher sequential read speed?
As a side note, I find that the SSHD underperforms according to reviews - CrystalDiskMark scores should be higher, more like 225 MBps seq read.

Thanks for your thoughts!

SSD - MS std AHCI drivers

[Read]
Sequential 1MiB (Q= 8, T= 1): 531.358 MB/s [ 506.7 IOPS] < 15758.86 us>
Sequential 1MiB (Q= 1, T= 1): 508.148 MB/s [ 484.6 IOPS] < 2062.57 us>
Random 4KiB (Q= 32, T=16): 107.352 MB/s [ 26209.0 IOPS] < 19488.90 us>
Random 4KiB (Q= 1, T= 1): 23.598 MB/s [ 5761.2 IOPS] < 173.33 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 111.146 MB/s [ 106.0 IOPS] < 74756.56 us>
Sequential 1MiB (Q= 1, T= 1): 100.242 MB/s [ 95.6 IOPS] < 10454.48 us>
Random 4KiB (Q= 32, T=16): 11.112 MB/s [ 2712.9 IOPS] <184313.30 us>
Random 4KiB (Q= 1, T= 1): 13.017 MB/s [ 3178.0 IOPS] < 314.16 us>

Profile: Default
Test: 1 GiB (x1) [Interval: 1 sec] <DefaultAffinity=DISABLED>
Date: 2020/04/16 20:36:02
OS: Windows 10 [10.0 Build 18363] (x64)

SSD - AMD drivers
------------------------------------------------------------------------------
CrystalDiskMark 7.0.0 x64 (C) 2007-2019 hiyohiyo
Crystal Dew World: https://crystalmark.info/
------------------------------------------------------------------------------
  • MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
  • KB = 1000 bytes, KiB = 1024 bytes

[Read]
Sequential 1MiB (Q= 8, T= 1): 554.269 MB/s [ 528.6 IOPS] < 15106.02 us>
Sequential 1MiB (Q= 1, T= 1): 516.541 MB/s [ 492.6 IOPS] < 2028.60 us>
Random 4KiB (Q= 32, T=16): 108.031 MB/s [ 26374.8 IOPS] < 19371.73 us>
Random 4KiB (Q= 1, T= 1): 25.241 MB/s [ 6162.4 IOPS] < 162.07 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 110.938 MB/s [ 105.8 IOPS] < 74881.09 us>
Sequential 1MiB (Q= 1, T= 1): 111.998 MB/s [ 106.8 IOPS] < 9345.52 us>
Random 4KiB (Q= 32, T=16): 11.179 MB/s [ 2729.2 IOPS] <183876.10 us>
Random 4KiB (Q= 1, T= 1): 12.166 MB/s [ 2970.2 IOPS] < 335.97 us>

Profile: Default
Test: 1 GiB (x1) [Interval: 1 sec] <DefaultAffinity=DISABLED>
Date: 2020/04/16 18:59:30
OS: Windows 10 [10.0 Build 18363] (x64)

SSHD - MS std driver
------------------------------------------------------------------------------
CrystalDiskMark 7.0.0 x64 (C) 2007-2019 hiyohiyo
Crystal Dew World: https://crystalmark.info/
------------------------------------------------------------------------------
  • MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
  • KB = 1000 bytes, KiB = 1024 bytes

[Read]
Sequential 1MiB (Q= 8, T= 1): 187.687 MB/s [ 179.0 IOPS] < 43931.04 us>
Sequential 1MiB (Q= 1, T= 1): 157.085 MB/s [ 149.8 IOPS] < 6639.18 us>
Random 4KiB (Q= 32, T=16): 5.356 MB/s [ 1307.6 IOPS] <271883.75 us>
Random 4KiB (Q= 1, T= 1): 1.907 MB/s [ 465.6 IOPS] < 2142.53 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 188.952 MB/s [ 180.2 IOPS] < 44136.77 us>
Sequential 1MiB (Q= 1, T= 1): 201.347 MB/s [ 192.0 IOPS] < 5203.25 us>
Random 4KiB (Q= 32, T=16): 1.310 MB/s [ 319.8 IOPS] <512353.87 us>
Random 4KiB (Q= 1, T= 1): 1.335 MB/s [ 325.9 IOPS] < 3064.63 us>

Profile: Default
Test: 1 GiB (x1) [Interval: 1 sec] <DefaultAffinity=DISABLED>
Date: 2020/04/16 20:41:19
OS: Windows 10 [10.0 Build 18363] (x64)


SSHD - AMD driver
------------------------------------------------------------------------------
CrystalDiskMark 7.0.0 x64 (C) 2007-2019 hiyohiyo
Crystal Dew World: https://crystalmark.info/
------------------------------------------------------------------------------
  • MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
  • KB = 1000 bytes, KiB = 1024 bytes

[Read]
Sequential 1MiB (Q= 8, T= 1): 204.017 MB/s [ 194.6 IOPS] < 40960.34 us>
Sequential 1MiB (Q= 1, T= 1): 204.488 MB/s [ 195.0 IOPS] < 5125.45 us>
Random 4KiB (Q= 32, T=16): 1.453 MB/s [ 354.7 IOPS] <488507.50 us>
Random 4KiB (Q= 1, T= 1): 0.612 MB/s [ 149.4 IOPS] < 6678.98 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 186.402 MB/s [ 177.8 IOPS] < 44699.13 us>
Sequential 1MiB (Q= 1, T= 1): 196.502 MB/s [ 187.4 IOPS] < 5329.94 us>
Random 4KiB (Q= 32, T=16): 4.882 MB/s [ 1191.9 IOPS] <276330.03 us>
Random 4KiB (Q= 1, T= 1): 4.858 MB/s [ 1186.0 IOPS] < 842.20 us>

Profile: Default
Test: 1 GiB (x1) [Interval: 1 sec] <DefaultAffinity=DISABLED>
Date: 2020/04/16 19:01:35
OS: Windows 10 [10.0 Build 18363] (x64)
 
SSHD's you have to understand use that SSD part (Usually 8GB SSD) as a Cashe an storage for most used files.

So it may write data to the SSD part giving you high numbers, but then then it may flush that to the platters and then read speed is from platters.

Drivers on the other hand can be all kinds of weird.

The only thing I don't understand is why would AMD say not to use their SATA drivers on a brand new board?? If it was something 10 years old i could understand if it is an unsupported board, but still thats weird.
 
Hi and thanks for your reply,

Agreed for SSHD operation - but crystaldiskmark should create a new file for each test, so one could expect that the measured speed is actually the platter speed. Anyway, the numbers are (a little bit) below par wrt this particular drive's reviews.
Re drivers, no idea why AMD recommends that. They say that they use the standard Windows 10 driver, and recommend to install their proprietary driver only for Windows 7 and below. What puzzles me if the clear difference in results...