RX 6700 crashes

[Danova Kontrola]

Hey guys, hopefully the right section.
I got new GPU - AMD RX 6700 from XFX and i have some crashes.

The crashes are kind of softs, the PC stops responding but first video pauses, few seconds later audio stops too, then after few seconds black screen.

The crashes only happens when playing Star Citizen, i dont play many demanding games (Arma Reforger works just fine).

I tried 3D Mark TimeSpy stress test, and passed just fine.

Win10 latest updates, i tried latest and one version older GPU drivers, no difference.

Google said i should try to disable FreeSync but didnt help.

Temperatures are more than better even under 100% load.
Issue never occured with my previous GPU ( RX570 ) on this system.
I asked in the game chat whats the experience with RX 6xxx cards, and i feel like the experience is mixed.

I have no idea what to do now.

 

My question is: Does this happens because

1.) the driver is not perfect because the GPU architecture is new and i should wait for more updates which will deliver fix
2.) or should i return it to the shop as a faulty? (  i bought it a week ago as a brand new, and got 5 years warranty )

When i check Event Viewer in Windows, i see this error logs, which i believe are related to the crash:
WudfUsbccidDrv:

An operation has failed (0xA, 0x0, 0x0, 0x0).
ScReadWrite: Failed to read reply.
HResult: {Device Timeout}
The specified I/O operation on %hs was not completed before the time-out period expired.

Service Control Manager:

The AMD External Events Utility service failed to start due to the following error: 
The service did not respond to the start or control request in a timely fashion.

 

A timeout was reached (45000 milliseconds) while waiting for the AMD External Events Utility service to connect.

 

[NF-A12x25]

This is definitely not normal behavior. Try using DDU to uninstall and reinstall your drivers. If that doesn’t fix the crashes, then return the card for a new one.

[Danova Kontrola]

So i decided to return the GPU to the shop and ordered same model again.
The bad part is the PC still crashes. ( still only in Star Citizen )
I tried undervoltage the GPU chip via the AMD tools, but it didint help. (i also noticed durring gameplay the voltage is always something around 0.9V, quite far from the maximum )
I tried to boost the cooling, but no difference. ( all tools were saying the GPU consumes 140W but temperateure was on 40C which is hard to believe to me, when i tried to touch the backplate after crash it was more then warm, but fans were still on something like 30%)
I tried memtest, because i got new pair of new 2x8GB (4x8GB now, all are the very same model), and after 6 hours it passed with green.

I ran Prime95 to stress test the CPU, and after 45minutes still fine. I also disabled the PBO feature few days ago.

I ran Prime95 to stress test the CPU while running OCCT stress test on GPU. So far running without issue.

I monitored the voltages of PSU (12V, 5V, 3,3V) via several tools and all of them looked fine. ( delta was less then 5% )

I tried drivers which are pushed by Windows Update, and same result.
I tried the absolute oldiest drivers for this GPU which i could find on AMD pages, same results. ( i always used DDU )

 

I noticed the OCCT doesnt use much memory of the GPU, it could be the memory chip on GPU is wrong. Not sure how to test it. (having 2 GPUs with wrong memory chip is unlikely but it could be caused by the buggy driver, not wrong memory chip itself)

 

Iam thinking of replacing my older PSU. I have it since 2013 ( and it wasnt new ). The PSU is Fortron Aurum Pro 1000. It handled i7-3930k 130W TDP OC@4.7GHz,1,22V + 2x AMD HD7870 175W TDP, with 11x 3.5" HDDs. I would say if it handled this then the new setup should be no problem. So far i am interested in "Fortron HYDRO G PRO 1000" but havent decided because i dont see the PSU as issue.

 

I also noticed the crash happens for the first time after 30minutes of gaming. But the second and third time are definetly less then 30 minutes. It could be something is overheating?

RAMs, CPU, MB in this setup were bought as a new this year.

After one crash/reboot AMD tool asked me to report the crashes which i had so i did my best in describing. 

 

I could be an asshole and order the new PSU (somethign cheaper but good enough), test it, and then return it back in 14 days&get my money back. Just to test with different PSU and not losing money.

 

I could return this GPU too and get one from NVidia this time.. (with my server lab and Windows Subsystem Linux i dont need Linux on my workstation anymore(nvidia didnt work on my linux well))

 

I feel lost..
but hey! i manage to make the GPU driver works on my Debian Linux!


//edit: i found the button in OCCT to test VRAM, i left it for 2GB and then to use 8GB of my 10GB VRAM. Both took 1 hour and were successful. 

//edit: once i lowered details in game to minimum and used the upscale resolution feature (which mean i render the game at 1080p i guess instead of 1440p) the crash didint happen in several hours. So i have powerful GPU which i cant use..

Edited November 11, 2022 by Danova Kontrola

[Danova Kontrola]

I returned it it too. And i bought RX 6750 XT from MSI.
The GPU power consumption shown in tools is almost doubled. Temperatures sky rocketed to over 70C and its also super loud in 100% usage. Which all seems right and makes sense.

The good thing is the PC didnt crash so far not even once! (Only game which is expected with Star Citizen.)

 

Problem solved - never ever going to buy from XFX.