PC Graphics

[enegra]

I have a reference 5700XT card, which has been giving me a rocky ride since August. Initially I was getting a lot of BSoDs through September, even with the "dangerous" settings off, like enhanced sync or hardware acceleration. Sometime in October the drivers seem to have been fixed and I haven't been getting that kind of crashes anymore.

When this card works, it works nicely. It seems to be mostly delivering performance and gets pretty average scores in benchmarks (something around 8900 in TimeSpy). It doesn't crash in benchmarks. But during longer sessions in more demanding games it does. Usually it starts with heavy stuttering, then there's a sudden black screen. Sometimes the image comes back after a dozen of seconds, but not for the active apps or desktop, so restart is needed anyway.

I tested it with my husband in another system with slightly higher end other components (newer) and the card has the same behaviour in there, in fact the crashes are more common in there. Since it is a reference design, initial thought was thermal throttling/crashes, but even with the fan curve cranked up all the way to 100% and/or undervolt it still does the same, despite the card not exceeding low 70s and low 80s on the hotspot and junction. It sure does sound like an angry hairdryer at 100% fan speed though!

What I noticed, is very high VRAM usage on the card, growing all the time while the game is running and the stuttering begins when the card is very close to exceeding its 8GB capacity and the crashes follow after that. But the games I've mainly played - heavily modded Skyrim and World of Warcraft have worked just fine with the same graphic settings on two different cards, of which neither had more memory - GTX 970 (3.5 GB) and GTX 1070 (8GB).

I have pretty much ruled out other issues with both setups and I haven't been able to find much about this particular issue - most of the people are just complaining about the heat, but if the card is allowed to get loud it doesn't actually get too hot. Is the card faulty or is it a driver issue?

The computers I tried the card with:

Setup 1: i7 7700K, 16GB 3000Mhz CL16 RAM, Z270 MSI Gaming M5, Seasonic Prime Ultra Gold 650W (Earlier was an old Corsair TX750 Bronze 750W, no change in behaviour), Evo 970 Plus 1TB + 2TB of spinning rust, 1080p 60Hz display.

Setup 2: Ryzen 7 3700X, 32GB 3600CL16 RAM, X470-i Asus ROG, Seasonic Prime Ultra Gold 650W, Evo 970 Plus 1TB + 1TB Evo 860, 1440p 144Hz display w/ Freesync.

What we tried:

• Fresh Windows install
• DDU in safe mode (multiple times)
• Different driver releases
• Agressive fan curve
• Undervolting
• Different settings - Enhanced Sync off, Freesync on/off, Hardware acceleration on/off, PCI-e 4.0 and 3.0.
• Different PSU (all the time the card was connected by two separate cables)

So far no solution found for the crashes and we are really out of ideas what else we could try. Appreciate any tips.

[averey]

What you maybe can try as well is making your virtual memory bigger, that is what I am currently testing. Increase the virtual memory size to something more than 800 mb in your advanced settings.

Also, what you can do and what I currently test as well so not to get crashes is to increase the TDR, as the crashes seem to occur because Windows attempts to restart the driver if it does not get a response in time. 

How to do it is described in this thread

Hope it helps, as I am currently testing it myself. Increasing the TDR with Enhanced Sync on seemed to help at first, and a stuttering loading screen which would have crashed my system did load for me. But then it crashed anyway, and now I turned Enhanced Sync off, so far no crashes at Battlefront 2.

Still, the driver issues with this card are rather unnerving. 

[enegra]

Thank you for the suggestion - will try it. Some of the crashes actually got entries in the Event Viewer that pointed to timeouts, so it's worth trying - I now vaguely remember that many years ago my 280X exhibited similar random display cuts and it helped then.

[hardcoregames_]

check your memory timings as that seems to the main reason for instability

this is due to a bug in the BIOS that sets timing as more aggressive than AMD processors can handle

[enegra]

This issue is happening with both Intel and AMD systems. Both of them are also stable when running with the 1070. We don't use the auto settings in BIOS, but manual memory timings, which are Memtest stable at least. I don't think that's the problem.

I was also able to confirm that it is VRAM crashing the card and found an easy way to reproduce it - the card in question doesn't free up resources fast enough, so just launch enough games until stutter begins and soon after that the video crashes - the system is still running and audio does work.

[hardcoregames_]

I  suggest slightly more relaxed timings to be safe. A BIOS may like it but Windows needs stable read and write.

[hardcoregames_]

driver 19.10.2 surfaced recently

[averey]

Driver 19.10.2 did crash on me as well, sadly, more than 19.7.5, while playing.

[enegra]

Already tried 19.10.2, unfortunately it brought no improvement in my case.

[Matt_AMD]

Based on what you have described, and the fact you have tried it in two different systems, it does sound like the graphics card could be faulty. 

[jsb]

I had exactly the same problem. Even just on desktop the 5700XT just always ran on 1750Mhz.

Tried everything, for me it had to do with the frequency of the screen. 

It is a 144hz panel and apparently, the card behaves strange when you select that frequency. 

You can try to select 120hz or 60hz on your monitor it should behave normal again.