Weird Windows wide bug: parts of the screen sometimes not refreshing in several programs

I am having a weird issue with my computer and I am trying to narrow down the source. It seems like a graphic issue, but I doubt it is my video card.

In Chrome, Discord, and other non-game programs, it is like only parts of the interface will scroll. For example:

In youtube, if I use the scroll wheel the scroll, it will only scroll through the videos inside one video thumbnail (like it is an embeded website. It is like a individual video is a scroll page in its own. This also happens with Twitter. Sometimes a single twitter post will scroll, or only the center feed will scroll (if you have a graphical element spill over into the left or right margin, you will get a graphical artifact with it staying there until the page is refreshed). Now, this is definitely an issue with the rest of the screen not refreshing graphically, rather than the individual object scrolling. I can tell because alt-tabbing will refresh the page to what it should look like. Also I am sure this is a graphical issue because it isn’t always on an individual object and will happen with the screen blacking over in places just like a driver crash type of issue. This is similar in Discord too.

I also have weird things like a text cursor appearing in weird places.
Note: This seems to be worse when I have a game running (Throne and Liberty), which makes me wonder that it is CPU related.

The reason why I can’t narrow this down is that I just had Windows 11 upgrade to 24H2, and I also cleaned and reinstalled the AMD Adrenalin drivers to troubleshoot another problem. I have both my CPU and GPU undervolted, but reverting them both to stock did not fix the issue.

Old GPU? just because there’s official driver support…

New 7800xt. Also a 5800X3d for the CPU.

1 Like

Try turning off GPU acceleration in the browser and see if there’s any more artifacting. If so it’s due to the GPU. If not I’d investigate RAM. Maybe turn off DOCP, if you have it enabled, and see if the issue persist.

Have seen similar behavior recently with 24H2 (upgraded from 23H2) on a 5950X/X570 system, the cause here was a PCIe NVMe SSD, not the GPU (was a NVIDIA dGPU), as soon as the PCIe SSD (not the OS SSD, the OS was installed on a SATA SSD) was under a slightly more intense load (not even a full load like a synthetic benchmark), parts of the screen would stop refreshing, the mouse cursor would stop moving and after a short while it worked again for a few seconds until it froze again.

The system would not crash, after the load on the PCIe NVMe SSD was done the system would respond normally again. No entries in Windows Event Viewer, drive SMART values all good, also no PCIe Bus Errors (the SSD was PCIe Gen3 only).

A different SSD in the same slot, even Gen4 works absolutely fine.

Are you using a PCIe NVMe SSD?

That sounds like the PSU issue I had a few years ago.
It was pretty bad. Killed some mice.
EVGA pulled some bullshit on that.

Yeah I am using a Western Digital SN770

Did you have to unplug the drive completely, or just move the software you were loading off of it?

1 Like

Does it have the 24H2 patch?

1 Like

Something seems to be up with Western Digital M.2 SSDs and Windows 24H2, or maybe just Western Digital SSDs and changed behavior in Windows started to trigger the issues that were masked in the past.

The issues I had where with a WD SN750 2TB with the latest firmware to-date (102430WD). That model has DRAM cache and shouldn’t be affected by this.

I just swapped the drive for a Samsung 990 PRO 2 TB with its latest firmware and no issues since then.

Yeah and I still have the issue. I was thinking on getting a new PSU since my Supernova G2 is 11 years old, but I will try moving the Steam library to a different drive and disabling it in BIOS and see if that fixes it.

Hopefully it isn’t the PSU since it would be nice to hit up Black Friday.

EDIT

I disabled the drive and swapped the games onto a different disk. The issue persists.

Yeah, I was thinking it would. For this I’d tend to suspect display related software or maybe the LCD, display cable, or GPU hardware. Don’t think the odds are very good but if Adrenaline proper’s installed, along with the drivers distributed with it, I’d try uninstalling it to see if it makes any difference.

AMD Cleanup Utility + driver only Adrenaline reinstall might be another way to try to narrow the possible set of causes but it’s unclear to me AMD’s still maintaining the utility.

I’d try this as well. Resizable BAR’s also sometimes implicated in similar issues.

FWIW I’ve a 7700 XT in a build running 11 23H2 with current driver only Adrenaline and haven’t hit anything like this with Firefox.

Hmm, could be worth a try if other possibilities run out. Monitoring rails with HWiNFO64’s hit and miss with the sensors available but it might offer some insight into whether +12’s dropping too much or regulation on the minor rails has gotten too loose.

Another possibility’s the G2 isn’t keeping up with 7800 XT transients, though anecdotally it seems more common 7800s have larger problems with older 750s. I’d rate it unlikely given the use cases mentioned, but weird problems tend to have odd causes, and 7800s are capable of ~350 W 20 ms spikes.

Given Microsoft’s Windows client and WD’s flash group’s trends lately, my guess is WD’s patch is most likely a workaround for a 24H2 host memory buffer regression. No good external way to tell it isn’t more WD’s issue, though, but it is concerning either Microsoft’s testing’s so poor 24H2 released without the BSODs getting detected or that Microsoft knew about them and put out 24H2 anyways. With the lawsuits around SanDisk I’m also concerned about the reorg of WD’s flash group under SanDisk’s brand. Probably it’ll take a year or two to tell if that’s affect quality in the Black and Blue NVMe lines.

FWIW I’ve seen 10-40% downgrades on the DRAMed drives I’ve had a chance to compare across 10 22H2 and 11 23H2. 990 Pro had ~2.5x the degradation of SN850X on SEQ2M Q1T8 and generally seems fragile in the workloads I’ve been looking at, to the point of 990 activity sometimes freezing the desktop for ~3 s. Given how shit 24H2 is I’m not looking forward to updating. But can’t say I’m confident in Samsung’s firmware either.

It’d be interesting to compare with T500, KC3000, Fury Renegade, MP600, or Rocket 4 but I don’t have any of those drives in operation.