Drivers & Software

[AstroJam]

With chipset drivers 4.11.15.342 and an SKhynix P41 2tb nvme ssd, I'm getting 7200mbs read speed but only 2400mbs write speed. This is usually a problem with the chipset drivers.

[Earnhardt]

Info required when posting a question 

[AstroJam]

• AMD Graphics Card Make & Model
  • 7900 XTX Reference
• Desktop or Laptop System
  • Desktop
• Operating System
  • Windows 11 22H2 Build 22621.1265
• Driver version installed
  • Radeon Software Adrenaline 23.2.2
• Display Make and model and connection/adapter in use, resolution, and refresh rate.
  • LG C2 3840x2160 @120Hz HDMI 2.1 adapter
• CPU/APU Make and model number
  • Ryzen 9 5950X
• Motherboard Make & Model + Bios Revision
  • MSI B550I Gaming Edge Wifi BIOS 7C92v1B
• Power Supply Unit Make & Model + Wattage
  • Coolermaster 850W Gold
• System Memory Make & Model + Frequency
  • GSkill Ripjaws V 3866Mhz CL18 2x32GB

[goodplay]

Might also help, knowing what App. you're getting the readout from?

[AstroJam]

crystalDiskMark8_0_4c

Prior to recently testing with 22H2 and 7900XTX (from a RTX 3080). Both r/w 7000\6500. Now 7300\2400. Funny thing is, if I disable Windows write caching on the SSD it goes to 5800\3200. I don't know if it's the latest chipset drivers or Windows 22H2. Changing nvme drives yield the same results so it's not the Windows nvme driver.

I'll be building an AM5 system soon and imaging the same install, so I'll know if it is the AM4\5950X and chipset driver or Windows in a few weeks. I'd like to figure it out on this system though. All my reasearch points to AMD chipset drivers or Windows 11.

[AstroJam]

CrystalDiskMark8_0_4c

[goodplay]

Try the newer 5.02, see if you get the same low result.

I had no problem with either.

[AstroJam]

Didn't fix it, though overall benchmarks across system went up between 1-2%. I believe it's Windows. This happened in November of 2021 and MS released KB5007262 to address it. Looks like it's back. I'm going to submit feedback to them about it.

[Delver]

I am having the same issue across all my M.2 SSDs since March 18th. 

I applied both the newest windows update (the one related to the momentum 2 or something like that) and the recent AMD chipset driver for my x570. I have an x570s Aorus Master and only 3 M.2 slots in use (#1, #2 and #4). The first (on the CPU lane) was working fine until yesterday, the other two (on the chipset lane) have inconsistent low speeds since the updates (like 25% of the expected speeds, at best) and they also freeze the system a lot if I'm transferring stuff or during my nightly backup routines.

Sometimes the system just freezes for a few seconds and recovers itself without any BSoD, but some background software and drivers crash in the process. However I'm not in luck these past two weeks because it's freezing and crashing the system multiple times (black screen, then BSoD pointing this error message: "INACCESSIBLE BOOT DEVICE"). In games, they freeze sometimes but they don't crash.

I did a clean installation to the Win10 22H1 in a separate M.2 Gen3 yesterday thinking the issue could be related to Windows 11, and to my surprise, now all four SSDs are having the same symptoms. I have the latest AMD chipset driver on both Windows 11 (old installation) and Windows 10 (clean installation).

• NVidia Graphics Card Make & Model
  • MSI RTX 4090 Suprim Liquid X
• Desktop or Laptop System
  • Desktop
• Operating System
  • Windows 11 22H2 Build 22621.1413
• Driver version installed
  • NVidia GeForce Game Ready Driver version 531.41
• Display Make and model and connection/adapter in use, resolution, and refresh rate.
  • LG C9 3840x2160 @120Hz HDMI 2.1 adapter
• CPU/APU Make and model number
  • Ryzen 9 5950X
• Motherboard Make & Model + Bios Revision
  • Gigabyte X570S Aorus Master, AMI BIOS Ver.F3 (rev 1.0)
• Power Supply Unit Make & Model + Wattage
  • Corsair HX1200i 1200W 80 Plus Platinum
• System Memory Make & Model + Frequency
  • Corsair Vengeance RGB PRO SL 3600Mhz CL18 2x32GB

[AstroJam]

I built a new AM5 system and imaged over my old system drive. Same software but different mobo  chipset and different SSD (WD SN850X). No issues whatsoever so I believe it was tied to the hardware. Something going on with Windows 22H2 and Ryzen 5950X on AM4. Could be chipset drivers or a combination of Windows+Chipset+SSD.

[Delver]

Thanks for the reply.

I 'm confident it's related to Windows 11 itself because it started in March 18th, the moment I applied the security update from March 15th. I've awaited for the April 11 hotfixes but the problems persist. Since yesterday I am transferring everything to my old HDDs and I must format the two gen4 SSDs using the manufacturer software, as last week one of them started missing in Windows, but appears normally in the BIOS and I can browse its folders from a separate WindowsRE unit. Now I have to boot and reboot anytime it happens.

[petosiris]

Hi @AstroJam and @Delver ,

Have you measured the temperatures of your SSD?  If so, what is the temp during gaming? 
Low write speed and sudden halts could be thermal throttling due to overheating perhaps.

[AstroJam]

I did, temps were fine. I actually have active cooling on that board. It was fine until some updates, just not sure if it was a Windows 22H2 or chipset updates or a combination. It could also be the SSD itself. It's an SK Hynix P41 and they don't do firmware upgrades.  Before I pass that system down to a niece, I'll be changing the SSD out and I'll see if it's the SSD.

[Delver]

I've been checking temps and performance when I can and, in games, it's fine 95% of the times. There's the occasional stutters and eventually the long loading times, but other than that it's working OK in game. It oscillates in between 52 and 61 Celsius, at most, for the main drive and the other two (that most games are installed) are between 49 and 56 Celsius.

I updated the chipset driver in March 30 together with a new BIOS version (F5a) and nothing changed. The same day I swapped the drives around. Temps didn't change a bit, they remain within specs.

[t3chl0rd]

I was having freezing crashing, and had some reported slow speeds on my "HikVision E3000 NVMe SSD 512GB" which were a last minute budget purchase, this SSD was my boot drive in 1st CPU slot, I have an identical one in the next slot. This also started happening around march, I did a new install and upgraded from win10pro 22h2 which is when my SSD started giving problems, to win11pro 22h2 and when I first ran CrystalMark I noticed the boot SSD was slow, and the 2nd SSD was normal. I updated chipset drivers and it showed both running fast again.

However my freezes continued and I was troubleshooting  the Graphic Drivers because this was where I was getting my freezing, but before it was random freezing everywhere. Just yesterday I took out the boot SSD in slot1, put in an old Western Digital SATA m.2 and reinstalled windows with all my BIOs at stock settings except what I need to make it run properly, and am slowly testing things (started with 2 sticks of RAM no XMP, reseated all my hardware, checked all connectors, and so far no more crashing/freezing.

Note: My HikVision SSDs are still fairly new, SMART showed %98 on boot SSD and %99 on 2nd. All test I ran on them were passing, both had custom double sided heatsinks and never reached above 41c. Windows was misreading them and giving a disk identifier warning in event viewer and had same incorrect serial number in my registry, however disk programs saw the serial numbers normal and differently.

[t3chl0rd]

Freezing was my ULPS being auto-enabled, Chipset drivers did fix my slow speeds initially though.

[Delver]

Ummmm... You just gave a great insight to something I was neglecting all this time.

How did you find out your ULPS was the source of the issue? I thought the LPS was set in the active power plan and had mine configured to never allow the devices to go low, nor suspend anything.

And what do you mean by "initially though"? Are you still experiencing the issue after installing the chipset driver? Which version by the way?

#WARNING: incoming wall of text#
#Failure to ignore may cause damage to your eyes#
It's been a whole month of headaches for me, trying to solve this tantrum.

Back in March I suspected the issue could be related to power delivery after noticing my 4090 would have spasms and hard locks whenever I switched the power plan to Balanced or Eco. Every game or even a YouTube video in 4K was like a Powerpoint presentation, it would make the entire system have "virtual parkinson disease" in a click of a mouse. However, I checked the power plans at the time and each of them was configured how I wanted (never letting it suspend anything, always preferring maximum performance, except applying minimum CPU power where it's due, according to each plan of course), so I stopped looking into it because it seemed unrelated to the power plans and I end up suspecting it could be the NVidia Geforce driver itself, behaving after the updates (I often update the geforce driver and tinker with the power settings to play in VR) and switched back to the High Performance plan (I rarely switch to other power plans to avoid issue with my virtual machines and my VR games).

So I focused on the SSDs, the chipset and the CPU and I moved forward. Tried cleaning the CMOS in multiple occasions, removing the thunderbolt card, removing the Gen3 SSD (not the Gen4 SSDs because they had important stuff I use for work) and putting it back later on when I reseated all three SSDs in different slots. Nothing helped, just time wasted and excessive hair loss. I remember when the issue begin, only the SSDs under the chipset lane would have low write speed scores (single digit low scores). The SSD under the CPU lane was fine, until I messed that up updating and reinstalling drivers, and tried uninstalling the Windows Update from March 14 (which was released as a security update, not optional and, at least to my system, irreversible).

So I used a clean Windows 10 install in a separate 500GB Samsung 970 Pro but the issue persisted on the Gen4 and Gen3 SSDs (I even removed my thunderbolt add-in card, thinking that was the issue and used a fourth SSD for a week and all 4 drives had the slow write speeds issue, unbelievable), then I read some stuff about the QLC and TLC Nands on some forums (Tom's Hardware maybe) and someone pointed out that many layers of the Nand could be filled out with enough data there was not enough free space to perform the trim correctly/anymore. I had no idea if this makes sense but it sticked to my mind for 3 days, even though both Gen4 SSDs had plenty of free space available (15% and 23% if I recall correctly). Still, I decided to erase both Gen4 SSDs, filling them with zeros just to check if that would be the case.

So I prepared myself to backup 4TB of important stuff (milked every GB from my backup unit for that) and removed all three SSDs from the motherboard. I used my 970 Pro that I have as external drive for work, deleted everything and installed the latest Windows 11 22H2 from the official ISO. To my surprise, everything went fine and perfect; I carefully installed every single piece of software, I delayed the drivers and update as much as possible. Hours later I applied the remaining updates and installed the drivers, except the AMD Chipset and NVidia Geforce drivers (Windows auto-installed the "whatever version" it had in Windows Update).

I proceeded to erase the old system drive (the first Aorus Gen4 SSD) filling it with zeros. It took almost 55min and froze on 91% for almost 15 min (I thought the process had broken in a corrupt address or something, but then it unstuck itself and finished with no errors). Moments later I begin the benchmarks and the issue disappeared (or should I say "obviously" disappeared). No slow write speeds anymore, no read speeds lower than mediocre values, it is all fine - finally - although both Gen4 SSDs are installed on the chipset lane. Read and write speeds are not optimum due to that. I 'm not confident if formatting the SSD filling with zeros is a solution, and I felt really bad because I still believe the issue is something really stupid I 'm missing, even something I did back in March that could have caused this tantrum.

I went to the second Aorus Gen4 SSD (the one where I store my virtual machines, games and other work stuff). I checked if by any chance it was OK already, opened CrystalDiskMark to test it and saw the slow write speeds all over the scores again, restricted to this second drive. I tested the first Gen4 SSD (the one I wiped off) and it's fine, but if I switch to the second it immediately scores slow speeds lower than my IQ tests.

Then I read your post about the ULPS, started typing this wall of text and realized I still had to apply the AMD Chipset Driver on this clean Win11 installation to see if by some miracle it fixes the SSD like it did to yours (don't know which version you applied to your system). The motherboard app is listing the version 4.09.21.138 (not the latest chipset), but I have the most recent one downloaded and ready to overwrite if needed.

Holy cow! it fixed! The chipset driver (version 4.09.21.138) fixed the issue... I guess... even though it makes no sense in a clean install that was reproducing the same issue prior to the installation of this specific chipset driver. I spent 1 hour stress testing the second Gen4 SSD and it is no longer scoring low single digit write speeds, it returned to normal/acceptable scores. I had shutdown the PC for 5 min to confirm if a cold boot would resurface the problem, but it seems gone (hope it's gone forever!). Thank you for the insight. I even tried looking at the Registry a while back following some instructions on how to find the registry key that enables the ULPS, but my registry is almost clean from install and has nothing related to it or it may be specific to AMD graphics card.

I know I had plenty of resources to safely wipe everything and retry as many times as I wanted, but my laziness took hold of me after 3 weeks of anguish. My brain's processing speed has become single digits now.

Thanks to everyone that took the time to read my rumblings and tried to help.
Much appreciated