Possible Faulty GPU

[L33williams]

Hi everyone,

I've recently upgraded from a 5500xt to a 6900xt. Since then, whenever I'm gaming, after around 20-30 the system does a full reset, lets me jump straight back into a game and does the same.

I never had issues with the 5500xt (other than it being really weak) and swapping it back the system is stable again.

I'm hoping its the card but want to avoid wasting time and being charged to be told its not the card and wondering if the faster GPU is highlighting an issue that the 5500xt could present.

I haven't tested with many games (Forza 5 and Halo infinite) but to try to isolate the issue to the GPU I've ran OCCT on memory, CPU and then GPU and only on the GPU did it crash, again, it took 28 minutes for this to happen.

Also, if it is the GPU, why does it take so long of load for it to reset, running blender, unigine etc dont cause the fault so its has something to do with time

The system is not overheating and I've already replaced the PSU thinking it may have been caused by a dying PSU creating spikes.

System spec:

AMD Ryzen 5 5600x (max temp under 1hr 100% load 71C)
Gigabyte x570i
Corsair 16gb 3600mhz
Corsair mx500 2tb
Corsair rm850 PSU
Powercolor 5500xt 8gb/Gigabyte 6900xt OC (Max temp on 6900xt is 97 on the junction, 77 core)
Windows 10.

 

[varkanoid]

Monitor the temps on the GPU and report back. PSU should be fine I had similiar with my 6900.

 

[shankly1985]

On your power supply cables are you running a single cable with the extra connectors on the end?

If so try running two cables from the PSU each cable to the power on the GPU.

My experience power supply with these connectors on from a single cable are not very good for high powered GPUs.

 

[Buffalo2102]

I would suspect PSU too.

 

[Dervious]

This brings back (very bad) memories of a graphics card upgrade I did about 10 years ago. Absolute nightmare to diagnose. Turned out to be the GPU despite everything pointing to something else. GPU was eventually repaired by Gigabyte, but had trouble with it when I upgraded my motherboard and CPU. Makes pre-built seem tempting.

 

[Cadder]

On your power supply cables are you running a single cable with the extra connectors on the end?

If so try running two cables from the PSU each cable to the power on the GPU.

My experience power supply with these connectors on from a single cable are not very good for high powered GPUs.

Definitely check this is not the issue. It's far too easy to plug them in and think they're fine just because they fit...

 

[FredFlint]

I use 2 8pin cables from the PSU to power my (3 8pin) 6900XT and it works fine.

 

[uscool]

first point of call and easiest as suggested try 2 8pin single from PSU, next step if you can get hands on another PSU to test

 

[L33williams]

Cheers for the suggestions guys.

I've swapped back to my old GPU and now that's doing it too.

OCCT fails on GPU and RAM tests but passes the CPU no problems. blender does the same.

I'm running 2 cables to the 3 slots.

Something that is really annoying me is that it fails after 28 minutes every time.

Temps for the GPU never exceed 96 on the junction, 70ish on the core, the CPU never goes above 71 at full load and around 63 while gaming.

Hope everyone has had a nice Xmas by the way.

 

[uscool]

Try uninstalling the graphics drivers can use DDU to completely remove the drivers and reinstalling also could try setting the bios to default , leave ram at stock and test again

Also update the bios if available

 

[heatonpkmassive]

I'd be looking more at the board now, BIOS reset or update if not on the latest is where I'd be at.

 

[$anch3z]

I'd advise monitoring your voltage rails, sounds similar to a PSU issue I had

 

[Alan McFall]

Check the event log, it can give you a good clue as to where the issue is stemming.

My lads been using a 960T that crashes 5-10 times a day for the past 6+ months, never figured the issue regardless of troubleshooting because I never had spare parts.. new install of windows always helped in the short term (multiple HDD's slowly corrupted). It must be the Mobo, I've had to reset the BIOS every once in a while to get out of a boot loop, restarts are far too random, there's nothing specific that causes it to crash.

If you can't get parts to test with then stress the system as much as you can, prime95 and furmark are a good combo. Take intermittent screenshots of your Volts and Amps using HWMonitor and then Goooogle.

 

[Ico]

Check the event log, it can give you a good clue as to where the issue is stemming.

My lads been using a 960T that crashes 5-10 times a day for the past 6+ months, never figured the issue regardless of troubleshooting because I never had spare parts.. new install of windows always helped in the short term (multiple HDD's slowly corrupted). It must be the Mobo, I've had to reset the BIOS every once in a while to get out of a boot loop, restarts are far too random, there's nothing specific that causes it to crash.

If you can't get parts to test with then stress the system as much as you can, prime95 and furmark are a good combo. Take intermittent screenshots of your Volts and Amps using HWMonitor and then Goooogle.[/QUO






+1 check event viewer

 

[MrAshieB]

I know it's not much help to you, but my other half had this exact issue when he swapped out a GTX 970 for a 5700XT back in early 2020. We never did find out the issue, but a return was accepted and we opted to wait for the new cards in Sept 2020. Now running an RTX 3080 from launch, with no issues.

I know how frustrating it can be, with this issue, so fingers crossed you can find a solution where we failed. So far, my experience with AMD/ATI GPUs has been poor, and have always opted for Nvidia as a result. I must be the unluckiest AMD customer around.

 

[jayjayuk]

As above check event viewer -Windows Logs -> System, sort in date order and look for any Errors and it should give you some clues as to why its resetting.

Ie TDR Failure, Kernel unexpected power loss etc

 

[L33williams]

After being told off by the wife for spending too much time on sorting it out around xmas and then having time to test it I had to take a break.

Could see anything that was showing in the event viewer much to my annoyance.

I've re-installed windows and after the 5th install, it seems to be working just fine. No idea what was causing the issue but it seems to be working now. The only thing done different this time than any other was to install windows, wait a day then test. wait another day then install the GPU drivers. repeated this process for every driver/game/program. What a PITA haha.

generally speaking, I'm happy with the card. Id still love to have had a 3080Ti but really cant justify the price difference between the two even if the software is better. Who knows, I might get back to the green team with the RTX 40 series

 

[xPETEZx]

There should be no reason whatsoever why you need to wait any time between those steps.

Are you able to pass the OCCT tests you where failing on before?

The fact you got nothing in eventviewer makes it sound like a PSU issue. When the PSU is either pushed beyond its limit, or is failing all it can do is cut the power / reset. Thus nothing can be logged.
So while not the ONLY thing that can cause crashes and leave no traces in eventviwer, Its the first place Id look.

I had very strange crashing issues. Sometimes rebooting at desktop, other times mid game. I went through every other part thinking it was GPU, then RAM and even mobo. Swapped PSU and the issues went away.

If you have a spare PSU, Id recommend that as an easy test. I used a 12year old HX750 I had lying around, and it worked like a champion. (Albeit very loud... these things are pre-fan stop!)

Hope your clear of these issues. Nothing more frustrating then hunting down an inconsistent issue like this.