[SOLVED] ADATA XPG SX200 Pro 512GB running a half speed

[paddybliss]

I’ve just built my first PC and have been running some tests to check everything is working as it should be. All seems good apart from my my 512GB ADATA XPG SX200 Pro NVMe drive, which is running at nealy half the advertised speed (and perhps 60% compared to real life benchmarks):

CrystalDiskMark Default Profile (similar results with other profiles):
[Read]
Sequential 1MiB (Q= 8, T= 1): 1724.685 MB/s [ 1644.8 IOPS] < 4858.56 us>
Sequential 1MiB (Q= 1, T= 1): 1625.145 MB/s [ 1549.9 IOPS] < 644.87 us>
Random 4KiB (Q= 32, T=16): 772.768 MB/s [ 188664.1 IOPS] < 2710.97 us>
Random 4KiB (Q= 1, T= 1): 66.150 MB/s [ 16149.9 IOPS] < 61.80 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 1433.420 MB/s [ 1367.0 IOPS] < 5801.86 us>
Sequential 1MiB (Q= 1, T= 1): 1599.906 MB/s [ 1525.8 IOPS] < 654.98 us>
Random 4KiB (Q= 32, T=16): 898.161 MB/s [ 219277.6 IOPS] < 2332.21 us>
Random 4KiB (Q= 1, T= 1): 174.943 MB/s [ 42710.7 IOPS] < 23.27 us>

I've run this over and over agaIin with different profiles and swettings and the results are nevery very different. I have seen a few other threads on here regarding this device, but this is a slightly different set of results to others. The write speed isn’t a low as some have reported, but still slow, and the read speed is way slower than others.

My system is:
Asrock B450M Pro
Ryzen 5 3600
GEIL EVO II 16GB
HP NVDIA Geforce GTX 1060

The ADATA drive is my primary boot drive with a fresh installation of Win 10 Pro. It’s installed in slot M.2_1, which is PCI Gen3x4 and should give speeds up to 32Gb/s. Windows reports that the drivers are up to date.

I’ve also:
Ran System optimization (TRIM) in the ADATA SSD Toolbox
Checked that write caching of device was enabled

As this is my first PC build, so hopefully there's just something obvious that I've missed?

 

[EveryoneKnowsZee]

Im having the exact same problem with the ADATA XPG SX8200 Pro 1 TB right now.

Maybe we can compare what weve tried and solve the problem together.

My system is:
ASUS ROG STRIX Z370-E motherboard
Two ADATA XPG SX8200 Pro 1 TB (problem drive occupying M.2_2 and good drive occupying M.2_1)
Two seagate barracuda 2 TB (occupying sata ports SATA6G_1 - 2)
One 980 reference card


I just replaced all my old drives with the above listed drives. All of my drives speeds are good except for the SX8200 which i installed windows on. I did a fresh install of windows. I have basically nothing else installed right now except windows and CrystalDiskMark.

My mobo manual says the following:

And so i should be able to run both nvme drives using PCIe 3.0 x4. Is there some setting I need to set to make them both run using PCIe 3.0 x4? That shouldnt make a difference though because only the M.2_1 socket supports both SATA and PCIe and the drive in that slot isnt the problem so this shouldnt be a problem of using SATA instead of PCIe right?

I have tried numerous runs of the benchmark including right after my initial reinstallation and after all these other attempted fixes.

Here is the speed of the problem drive:

Here is the speed of the good drive:

I have:
Updated BIOS
Updated all other drivers from the motherboard website
From Device Manager I have the latest storage controller drivers (1x Microsoft Storage Spaces Controller and 2x Standard NVM Express Controller)
Ran System optimization (TRIM) in the ADATA SSD Toolbox
I have reseated the problem drive


What i havent tried:
Reinstalling windows on the non-problem drive and comparing the two drives speeds after that
Switching which drive is in which m.2 slot (dont want to do because the mobo heatshield is already on the one drive and i dont want to have to replace the thermal pad and the problem drive already has the packaged heatshield on)

Ive also noticed that my computer will now stutter/freeze for a few seconds infrequently (every 2 hoursish). When it comes back online my speakers arent recognized. Its very strange but it must be due to the new drives or some cables as its the only things to change.

The only thing i can think of at this point is either: 1) its a problem of having windows installed 2) a problem with the m.2 slot 3) a problem with the drive 4) something related to mobo whether its a setting or our misunderstanding

I plan on reinstalling windows on the non-problem drive and comparing speeds and ill update you with that maybe tomorrow.

 

[Maxxify]

Check lane usage with CrystalDiskInfo.

 

[EveryoneKnowsZee]

Check lane usage with CrystalDiskInfo.

Wow thanks for this suggestion i think you may have found my problem.

Bad drive

Good drive

So this seems to indicated that the bad drive is currently running in PCIe 3.0 x2 instead of PCIe 3.0 x4. Does anyone know why this is and how to resolve this?

It seems like this post (https://rog.asus.com/forum/showthread.php?109185-2-NVMe-M-2-PCIe-3-0-x4-ssd-on-Z370 ) covers what to do for my particular motherboard. Ill give it a shot and update soon.

@paddybliss
I hope this is useful to you too and i apologize if i just ended up stealing your thread.

 

[EveryoneKnowsZee]

Yeah that did it. I just needed to go into BIOS and set the M.2_2 PCIe bandwidth to x4 instead of the default x2. Amazing im so happy. Thanks @Maxxify!

 

[paddybliss]

Thank for your replies.

I had a look at BIOS settings and eventually found this:

PCIe x16 Speed
M.2 PCIe Speed
Promontory PCIEe Speed

All set to Auto

I changed PCIe and M.2 to Gen 3 and left the others as they were. I was sure this would solve the problem, but no, still the same. Tried again with PCIe set to Gen 2, but still the same.

The only other setting I could find which might have some bearing on this is:

PCIe x16/2x8, which has the following options (currently set at x16):
X16
X8x8
X8x4x4
X4x4x4x4

Should I try changing this? I didn’t want to fiddle with this too much, as my graphics card is on the primary PCIe slot, and If that doesn’t work then I won’t be able to see what I’m doing to change it!

Also saw this:
https://www.reddit.com/r/ASRock/comments/fddqga/nvme_drive_running_at_x2_speeds_instead_of_x4/fjgycpw/

Which recommended trying this:
https://www.windowscentral.com/how-convert-mbr-disk-gpt-move-bios-uefi-windows-10

But would like to explore the BIOS options first.

 

[EveryoneKnowsZee]

Did you download and try CrystalDiskInfo?

 

[paddybliss]

Did you download and try CrystalDiskInfo?

Yes, the figure in the first post were from CrystalDiskMark.

I can't find an easy way to show them as screen shots like the others posts in this thread?

 

[paddybliss]

I've emailed both ADATa and ASRock about this.

I had a reply from ADATA saying that it could be a faulty drive, and am waiting to hear back about what they propose doing about getting a replacement for me.

Then I heard fromASRock, who asked me to updaqte the BIOS (which was released on Monday, just after I first fired up the PC). This has helped, but the speeds are still low in comparison to what EveryoneKnowsZee has reported above and other benchmarks I have seen.

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): 2695.963 MB/s [ 2571.1 IOPS] < 3109.52 us>
Sequential 1MiB (Q= 1, T= 1): 2529.941 MB/s [ 2412.7 IOPS] < 414.00 us>
Random 4KiB (Q= 32, T=16): 744.066 MB/s [ 181656.7 IOPS] < 2815.53 us>
Random 4KiB (Q= 1, T= 1): 63.946 MB/s [ 15611.8 IOPS] < 63.92 us>

[Write]
Sequential 1MiB (Q= 8, T= 1): 1409.742 MB/s [ 1344.4 IOPS] < 5933.80 us>
Sequential 1MiB (Q= 1, T= 1): 1735.054 MB/s [ 1654.7 IOPS] < 603.80 us>
Random 4KiB (Q= 32, T=16): 1131.365 MB/s [ 276212.2 IOPS] < 1844.83 us>
Random 4KiB (Q= 1, T= 1): 176.567 MB/s [ 43107.2 IOPS] < 23.05 us>

Profile: Default
Test: 1 GiB (x5) [Interval: 5 sec] <DefaultAffinity=DISABLED>
Date: 2020/08/07 17:49:41
OS: Windows 10 Professional [10.0 Build 19041] (x64)

Any other ideas?

 

[paddybliss]

I was looking at the CystalDiskMark results above from EveryoneKnowsZee and noticed that the drives was only 4% full. Also, looking at the benchmarks online, noticed most of them were running the drive at 0% capacity (presumably running CystalDiskMark on another drive?) I think it’s true that drives run slower the more full they are so I deleted some stuff and got this result at 17% capacity (of the primary sector):

But subsequent runs were lower. This more typical:

Default profile:

But can be quite a lot lower, esp when running other apps:

Do these seem right for actual real world usage (as opposed to real world testing)?
@EveryoneKnowsZee I note that your drive was only 4% full when you tested. Did you have apps running when you did so? Have you tested subsequently?

 

[paddybliss]

I sent the above results to ADATA and had this responce:

"According your screenshot, the write speed is normal, The speed 512g in the state of empty disk is 3500MB/S read, 2300MB/S write. As the test you provide is for system disk, the system SSD will influence the speed for reading and writing."

I tested from my other installation of Windows on another drive. (so the disk wasn't running Windows. ASRock also suggested disabling internet and doing 5 runs:

So runs two and three were very good, but the others as before. Should it fluctuate so much?

 

[paddybliss]

Back on this again... and the plot thickens!

I got a replacement drive from Amazon and they will let me keep the old one until I have transferred windows over to it. This is what I’ve done:

• Put the existing card onto an NVMe PCIe Card and put it in the spare PCIe Gen 2 socket
• Put the new drive in the PCIe Gen 3 M.2 slot
• Formatted my drive to NFTS
• Tested it when formatted but empty, running windows on the old drive

I got this:

I did this test several times and got almost identical results. Much more like what I expected from this drive, through the write speeds are a little low.

I installed Windows and apps on the new drive from an image file created from the old drive, booted Windows from that drive (still in the PCIe Gen 3 M.2 slot), ran the tests again... and I got this:

I tested over and over and similar results every time. It appears to be performing about the same as the old drive. The write performance isn’t too bad, but the read performance is very slow.

Any ideas as to what’s going on here? Is it a software or hardware issue?

 

[alceryes]

First, try turning off 'Windows write-cache buffer flushing'. Other manufacturers recommend doing this when benching their drives.

Second, if it were me, I would never put this much strain (test, after test, after test, after test, etc.) on an NVMe SSD drive, chasing an artificial number that will have no real world consequence. You are literally hunting fractions of a second, and causing a lot of extra wear and tear on your drives in the process.

Third, because artificial benchmarks get down to the fractions of a second, they need to be run in EXTREMELY controlled environments for them to really be meaningful. Don't get me wrong, they are great for overall performance measurements or to find something obviously wrong (like the PCIe 3.0 x2 vs. x4 hashed out above), but many times the differences just come down to Windows doing something in the background during benchmarking.

 

[paddybliss]

Thanks for getting back to me. I’ve turned off the Windows write-cache buffer flushing as suggested. Maybe a slight improvement:

But I won’t re test, you make a good point about the wear & tear on the drives. I’ve done dozens of tests now, how much will this affect the lifespan / reliability of the drive?

I realise that real world usage is going to be lower than benchmarking in a controlled environment (I did turn internet off for some of them), but from what I can glean through user benchmarks and feedback, my speeds are still quite low for this drive in general usage. See EveryoneKnowsZee results (as an OS drive) above.

What sort of real world results are you getting on your WD Black? From reviews, the ADATA should perform at about the same level. If I end up replacing the ADATA, the WD would probably be my next choice.

Fractions of a second won’t be noticeable in general usage, but I do a lot of video work, including DCP conversation for cinemas, which involves generating 100,000 – 200,000 individual still images, this is the sort of work where these speeds can really make a difference.

 

[alceryes]

As long as you don't continually benchmark/test it you'll be fine. Don't forget to turn the write-cache buffer back on for normal usage.

My WD Black -
https://1drv.ms/u/s!AjLCgZ8P-HzihW5SUAK-OQ_srTSq?e=fOgpT3

Remember that your CrystalDiskMark speeds are a combined result of your motherboard, components, and OS working optimally, together. The theoretical speeds listed for drives are rarely (if ever) achieved.

 

[paddybliss]

Thanks for sharing your drive results. I'm not going to ask you to do any more testing, but if you've tested before, are these the same sort of results you've had before with the drive in use?

 

[alceryes]

I don't quite understand your question.
The test results I linked to are from my drive in December 2019. I haven't tested since then.

 

[paddybliss]

Sorry for not being clear. I was just wondering if you had run this test more than once and got similar results, but seems not.

 

[alceryes]

Ohhhh. Yes, I ran it several times and got similar results each time.

 

[paddybliss]

Werre they back to back runs, or on different days / times? I found that I often got similar results in the same seasion, but they would be more different after a reboot.

 

[alceryes]

Both. I probably ran it a total of 10-15 times over the course of a week.

 

[paddybliss]

Ok, thanks. I've ordered a WD SN750, is that the one you have?

 

[paddybliss]

I got a WD SN750 500GB to replace the ADATA drive (I don't need to return this yet). Installed it, did my first benchmark of it formated to NTFS, but empy, running windows on the other drive, and got this:


However, once I put windows on it, the results for some of the tests, particularly the writes, were very much lower than I’ve seen in benchmarks for any othe others NVMe drives I’ve looked at:

And far lower the one posted above for the same drive (I think?) from alceryes. I get very similar results to this every time. How important are these tests in real world usage?

More worrying, I though I’d try a different benchmark app, and gotn this us AS SSD:

The score is a lot lower than my original ADATA Drive and nearly half what I’ve seen published for other SSD’s in this price bracket.

But it’s the text in red in the top left hand box that’s the most troubling. It seems to indicate problems with the drive, though I can’t find any documentation that explains this figure. I’ve run chkdisk and S.M.A.R.T on the drive and they didn’t report any errors.

Should I be concerned about this?

 

[Deleted member 14196]

You are going to ruin your drives by Constantly benchmarking them

 

[paddybliss]

Thanks, you're the second person to mention this, and it is noted.

I've only done few tests on the new one and I won't be doing any more for the time being, it looks like it's not going to change.

What i'm trying to ascertain is whether there is a problem with the drive (I don't like those red figuresonn the AS benchmark) or something else in my system that is causing poor results in some benchmarks?