Need help troubleshooting a 6700 XT with very strange behavior

[jak_2456]

I should preface this thread by saying that this is my first post on here, but I have read numerous other troubleshooting threads (unrelated to this problem) on TPU.

A few weeks ago I noticed that my sim racing PC started crashing semi-randomly in the middle of when I was playing EA Sports WRC, BeamNG.drive, and occasionally DiRT Rally 2.0. The first two games will run for about 5-10 minutes normally before hanging and crashing. WRC doesn't give me an error code but BeamNG gives me either a generic error code or a memory access violation. The last game, however, will typically run much longer before it will crash. I was also playing Portal 2's co-op splitscreen with a friend and that crashed twice in the roughly two hours we were playing it. Sometimes the PC will crash so hard it will reboot, but once when I was playing WRC in VR it froze for a few seconds and recovered like nothing had happened. The weirdest thing about this problem is that some games will work completely fine; Assetto Corsa and Forza Horizon 5 run flawlessly. I was able to race for 20 minutes in AC and over 30 minutes in Forza Horizon 5. Not only that, I can run Unigine Valley and MSI Kombustor for ages; they work fine. All of OCCT's tests passed too, including the power test. I can also run Cinebench R23 for 30+ minutes without crashing (the CPU reaches 81 ℃ but the 5800x does that regardless). All of the system temps look OK (don't remember off the top of my head what they were exactly but they were well under 80 ℃). I should also note that I tried putting two different GPUs in this PC: a Dell OEM GTX 745 and a MSI ARMOR GTX 1070; both cards ran without issue.

Full system specs:

• CPU: AMD Ryzen 7 5800x @ stock
• Cooler: Arctic Freezer A35 CO
• MB: Asus TUF B450-PLUS II
• RAM: 2 * 8 GB G.SKILL Ripjaws V DDR4-3200 w/ DOCP enabled
• SSD: Crucial 1TB PCIe Gen 4 SSD @ Gen 3 (pulled from a brand-new OEM PC, don't know exact name)
• GPU: XFX SWFT309 AMD RX 6700 XT
• PSU: EVGA 600 BA 600W 80+ Bronze Non-Modular (note: can deliver full wattage on 12V rail)
• Case: Cougar MX-330G Air
• OS: Windows 11 Home 64-bit

I have tried:

• running DDU and reinstalling the newest drivers from AMD
• reverting back to older drivers
• disabling AMD Smart Access Memory
• reducing the GPU's power limit
• swapping the RAM sticks' sockets (moving one stick to the other's slot and vice versa)
• running MemTest86 (passed)
• switching the BIOS on the GPU

I have NOT tried:

• reinstalling Windows
• putting the GPU into another PC (don't have an adequate PC to put it into)
• changing RAM set
• reseating CPU
• changing PSU

Any suggestions on what this could be or how to fix it would be much appreciated! I will do my best to provide any additional info you need.

 

[regeesfu]

did you had any blue screen of death?
memory error code might have something in common with ram...when my 4gb failed mobo did not even noticed it and run system fine....apart from that i had 4 not 8. Try swapping different ram.

 

[jak_2456]

I did not have any blue screens when my PC hard crashed; it simply froze and rebooted. I also ran MemTest86 on the RAM and it passed. I have another kit of 16 GB DDR4 from another PC, so I'll try swapping that in. Changing the GPU also fixed the problem, so maybe the VRAM is wonky on my card?

 

[jak_2456]

did you had any blue screen of death?
memory error code might have something in common with ram...when my 4gb failed mobo did not even noticed it and run system fine....apart from that i had 4 not 8. Try swapping different ram.

Nope. Changing out the RAM kit didn't fix the problem. I think it's a problem with the GPU.

 

[P4-630]

Not sure if it would make a difference but according to AMD:

•Minimum 650W system power supply

https://www.amd.com/system/files/documents/quick-start-guide-6700xt.pdf

 

[eidairaman1]

Lower powered cards are not going to put enough of a load on the psu.

 

[regeesfu]

I did not have any blue screens when my PC hard crashed; it simply froze and rebooted. I also ran MemTest86 on the RAM and it passed. I have another kit of 16 GB DDR4 from another PC, so I'll try swapping that in. Changing the GPU also fixed the problem, so maybe the VRAM is wonky on my card?

if you changed gpu and that solved your problem we can close this thread but i would double check motherboard just in case of overheating vrm's.

 

[jak_2456]

Not sure if it would make a difference but according to AMD:

•Minimum 650W system power supply

https://www.amd.com/system/files/documents/quick-start-guide-6700xt.pdf

I've been running this PC for a long time with that PSU. I should also mention that setting my 5800x to "Eco Mode - 65 W" in the BIOS made the problem worse somehow. I also checked online with a PSU calculator and found that I should be fine.
https://outervision.com/b/Egq6Dn

Lower powered cards are not going to put enough of a load on the psu.

I have run into PSU problems in the past (trying to run a GTX 1080 Ti on a ""750 W"" TigerDirect PSU) and the PC would always crash instantly under a FurMark or MSI Kombustor load; however, my PC runs Kombustor flawlessly for 45+ minutes. It also passes OCCT's "Power" test which loads all components (including CPU) for one hour.

 

[jak_2456]

if you changed gpu and that solved your problem we can close this thread but i would double check motherboard just in case of overheating vrm's.

One of the GPUs I used was a complete potato (the GTX 745) and the other wasn't mine (the GTX 1070). I would like to fix the 6700 XT crashing.

 

[jak_2456]

Dude kombustor does not work on a gpu, go borrow a reputable 750 watt and get back to us or go to a shop.

I also tried Unigine Valley, is that enough of a load? Also, Kombustor's website says that it is a "graphics card stress test", so you are likely confused with another piece of software.

 

[eidairaman1]

I also tried Unigine Valley, is that enough of a load? Also, Kombustor's website says that it is a "graphics card stress test", so you are likely confused with another piece of software.

Um I know kombustor is originally furmark, Stop wasting time, and do the swap of psu, it is to determine if gpu or psu are bad, just swapping to a lower powered card is not going to help you.

Dude kombustor does not work on a gpu, go borrow a reputable 750 watt and get back to us or go to a shop.

 

[jak_2456]

Um I know kombustor is originally furmark, Stop wasting time

If Kombustor is not a piece of software you use to test GPUs and/or PSUs then what would you use instead?

 

[jak_2456]

Um I know kombustor is originally furmark, Stop wasting time, and do the swap of psu, it is to determine if gpu or psu are bad, just swapping to a lower powered card is not going to help you.

Dude kombustor does not work on a gpu, go borrow a reputable 750 watt and get back to us or go to a shop.

My 6700 XT has two 8-pin power connectors. I will connect a second 500 W power supply to one connector and keep the EVGA PSU on another since that doesn't require me to buy another PSU. I know that you have to jump a second PSU's 24-pin connector in order for it to power on, by the way. I should also note that Assetto Corsa made my card draw its rated TDP and that ran fine, so I doubt that it's the PSU.

 

[jak_2456]

I think we can cross power supply problems off the list here. I just tried connecting a second power supply to one of the 8-pins on the 6700 XT and it still crashed in BeamNG.drive (I've been using that game because I can test it unattended while EA Sports WRC has no such ability). This is more likely an issue with the GPU than the power supply. I don't think it's a problem with the VBIOS because I tried switching BIOSes and it crashes with either position. Really hoping that the GPU is still covered by warranty!

 

[eidairaman1]

I think we can cross power supply problems off the list here. I just tried connecting a second power supply to one of the 8-pins on the 6700 XT and it still crashed in BeamNG.drive (I've been using that game because I can test it unattended while EA Sports WRC has no such ability). This is more likely an issue with the GPU than the power supply. I don't think it's a problem with the VBIOS because I tried switching BIOSes and it crashes with either position. Really hoping that the GPU is still covered by warranty!

Use 1 psu not 2 at same time, borrow 1

 

[jak_2456]

Use 1 psu not 2 at same time, borrow 1

If the PSU was the problem then why were Assetto Corsa and OCCT able to run without issue? I don't see a reason to keep chasing the PSU.

 

[A Computer Guy]

I think we can cross power supply problems off the list here. I just tried connecting a second power supply to one of the 8-pins on the 6700 XT and it still crashed in BeamNG.drive (I've been using that game because I can test it unattended while EA Sports WRC has no such ability). This is more likely an issue with the GPU than the power supply. I don't think it's a problem with the VBIOS because I tried switching BIOSes and it crashes with either position. Really hoping that the GPU is still covered by warranty!

Why? By using the old PSU you invalidate your test. Just use one PSU at a time when testing as it may not be a total power issue but some other issue with power delivery with your EVGA unit.

 

[jak_2456]

Why? By using the old PSU you invalidate your test. Just use one PSU at a time when testing as it may not be a total power issue but some other issue with power delivery with your EVGA unit.

I will keep saying this until I am blue in the face: I was able to stress test my GPU just FINE and it worked; HOWEVER, a couple games would crash. If it were the power supply then the stress test would crash.

EDIT:
I would like to put this troubleshooting investigation to a halt right now as I have just fried the motherboard in my attempts to fix this issue. I thought it might be the CPU not making good contact in the AM4 socket that was causing the PCIe lanes to drop out. In a desperate attempt to fix this likely non-existent problem, I tried scrubbing the socket with IPA and a toothbrush to clean out the thermal paste that was gunked up in there; the motherboard now no longer turns on. I plan to buy a replacement motherboard and if the GPU doesn't work in that board, I will RMA the GPU as I believe it is the culprit. I would like to thank everyone here who tried to help me fix my issue; this will no longer be your problem unless something changes.

The moral of my story here is to not get desperate trying to fix a problem and do something you otherwise wouldn't do. What I did here was entirely my fault.

 

[A Computer Guy]

What percentage IPA did you use?

 

[jak_2456]

What percentage IPA did you use?

70%. I may have used too much of it too. Again, I was quite desperate and I didn't think rationally about it.

 

[A Computer Guy]

70%. I may have used too much of it too. Again, I was quite desperate and I didn't think rationally about it.

70% doesn't evaporate very well. Probably want to go 90% or higher next time and possibly finish with an electronics cleaner on a socket to ensure the liquid is fully displaced and evaporated.

 

[jak_2456]

70% doesn't evaporate very well. Probably want to go 90% or higher next time and possibly finish with an electronics cleaner on a socket to ensure the water is fully displaced and evaporated.

Yep, will do next time (or just be extra careful so it doesn't happen in the first place).

70% doesn't evaporate very well. Probably want to go 90% or higher next time and possibly finish with an electronics cleaner on a socket to ensure the liquid is fully displaced and evaporated.

I do not believe it. IT LIVES NOW! The motherboard came back miraculously.

 

[A Computer Guy]

Yep, will do next time (or just be extra careful so it doesn't happen in the first place).

Hit it with electronics cleaner and let it dry for a day and try again. You might get lucky unless you heard your PSU go click or witnessed a puff of smoke from somewhere.

I do not believe it. IT LIVES NOW! The motherboard came back miraculously.

Personally I would still be a worried with 70% IPA if you still got liquid trapped in small spaces and would allow for more drying time.

I will keep saying this until I am blue in the face: I was able to stress test my GPU just FINE and it worked; HOWEVER, a couple games would crash. If it were the power supply then the stress test would crash.

My general understanding is games can cause spikes in power usage where stress tests are more consistent in power draw. By using your old PSU you invalidate your ability to rule out the PSU as the culprit not to mention putting the other PSU at risk by powering the GPU that way. I guess you will find out eventually by either replacing the GPU or PSU. Good luck.

 

[lordmogul]

My general understanding is games can cause spikes in power usage where stress tests are more consistent in power draw. By using your old PSU you invalidate your ability to rule out the PSU as the culprit not to mention putting the other PSU at risk by powering the GPU that way. I guess you will find out eventually by either replacing the GPU or PSU. Good luck.

I would go even further. Sometimes games manage to stress a system in a way that stress tests and benchmarks won't.
I remember people having benchstable, prime95 stable, furmark stable systems, but some games still crash.
And who else remembers New World destroying cards.

If you still want to troubleshoot the card, and you already changed other parts and already tested another GPU, and it runs fine then, I would look at underclocking it.
Sometimes parts just aren't stable how they are shipped, that can happen, and we can RMA them for that. Especially factory-oc'ed cards can run into that.
If you lower the clocks (GPU and memory independently) and it is stable, at least you know there is something wrong with the card.

 

[SomeOne99h]

jak_2456
Welcome to the forum!
jak_2456, using logic isn't enough to understand things in any field. You actually have to be in the know to fully understand. Many techpowerup members are pretty good when it comes to tech and usually the ones who knows posts to help others. You posted the topic to get help but unfortunately you seem stubborn about what you think is right. If you keep this attitude, you will be - and the others with you- running in circles.