Question B450 motherboard and Palit 3060 Dual - - - GPU crashes frequently ?

[Nanocore]

Hi, I need some advice regarding my setup, the GPU crashes as soon as it gets to a intensive part eg 3d load, otherwise just running Windows, browsing etc is fine with no loss of power.

My equipment:
Gigabyte ITX B450 I Aorus Pro Wifi - on latest bios (F63c) - Link
Ryzen 7 3700X
32GB DDR4 (2x 16GB) PNY 3200MHz C16 RAM
1TB Intel 660P NVMe drive
Palit 3060 Dual 12GB
PSU >> Fractal Design SFX ION Gold 650w / 750W EVGA SuperNOVA 750 GT, 80 PLUS Gold (RMAed)
Windows 11 Home - fully updated, OS build 22621.963

So the GPU was bought 2nd hand on ebay - I initially thought this was the problem as it kept crashing on my Ryzen system, I have access to a micro Dell PC 7050 (Windows 10) so tried to test it using a PCIE to NVME cable, but using the same PSU to power the GPU. Same problem. So I thought GPU was knackered.
I took the GPU to my brother who runs a 3080 system and he plugged it in, and working fine! Ran multiple games on 4k high settings for at least two hours, not a single crash! I should have taken the PSU also to test but didn't think to do this at the time.

Anyway back home after few more tests, I concluded that the PSU was faulty - this was the EVGA from Scan, (many Google links stating incompatibility between EVGA and RTXs) so I RMAed it and ordered the Fractal Design unit (I needed a smaller PSU anyway to be honest as space was VERY tight with the EVGA and ITX case.

Fast forward few days (Xmas holidays etc) and the new PSU arrives, plugged in and SAME DAMN THING - as soon as a game or benchmark is loaded it crashes.

Here's what I've done so far (not in order) after reading other posts and youtube videos:

1. Reinstall Windows 11 (PC came with it installed) - this was after many attempts of other steps
2. Reset BIOS CMOS on mobo - loaded optimised defaults
3. BIOS update to F63c
4. Uninstalled/Reinstalled drivers many times - used DDU in safe mode
5. Changed PSUs
6. Re-seated RAM modules, isolated sticks so booted off 1 in Slot1 then switched and booted off 2 in Slot2
7. Unplugged keyboard from USB hub and plugged directly into PC / Disconnect wired mouse and used BT mouse
8. Set NVIDIA 3D settings to performance not quality
9. Set Windows performance settings to high
10. I updated the GPU BIOS from v94.06.14.00.A2 to 94.06.14.80.9B (there's a later version so can give this a try but I was reluctant to do anything to the GPU as it was working fine in my brother's PC) link

* However, after updating the BIOS the GPU was in the Dell at the time - it ran for several hours, I left it running Cyberpunk and went to sleep, when I woke it was still running so I thought yay, fixed! But after trying to move the character it crashed again!

11, Put the GPU back in the Ryzen PC as I thought the Dell config with NVME cable wasn't helping - still crashed so BIOS update didn't help
12, All mobo OC settings are off, not using XMP
13, Changed PSU cables, different wall sockets!
14, Used two different monitors, 1 is DP to DVI and other is HDMI to HDMI
I'm now trying to go back to older versions of mobo BIOS but it's not letting me downgrade.
I'm also running out of ideas, I'm planning to drop the whole system off to my brother's for a few days while he plays about and switches components between mine and his PC.

Is there ANYTHING else I can do, this is seriously frustrating!

Thanks

 

[Nanocore]

anyway after more reading, in the event viewer I see:
\Device\Video3
UCodeReset TDR occured on GPUI:a00
Reset TDR occurred on GPUID:a00

Running Rise of the Tomb Raiser benchmark option (TR), it crashed with this error:
DXGI_ERROR_DEVICE_RESET

15, I've added TdrLevel reg key, set to 0 and rebooted - crashed on TR but screen froze, had to hard reset. Set to 1, crash, set tp 3, immediaet crash! Removed the key as obv its made things worse!
https://www.thewindowsclub.com/dxgi_error_device_removed
16, Turned off all Ant-aliasing options in Nvidia control panel - after reboot and retest - got a bit further in TR benchmark but crashed gracefully eg not frozen screen! 2nd attempt, crashed earlier....gracefully!
17, Turned off Nvidia Shadow Play / Overlay Off - immediate crash on launch (launched from Nvidia Experience...results vary wildly step to step!)
--Rebooted and launched TR from Epic Games launcher - crashed with frozen screen - need hard reset!

 

[Nanocore]

hmmm think it's the GPU...
Ran OCCT on VRAM several times, even at 20% memory utilisation the damn thing crashes!
Taking it back to my brothers for a final re-test, just can't understand why it worked in his system with ZERO crash, running happily at 4k ultra high settings..... the thing must be possessed!

No more ideas!

 

[Nanocore]

18, Played about with PCIe Gen settings on x16 and system in BIOS. e.g. gen2/auto - auto/gen2 - gen3/auto - auto/gen3
I saw a post that someone fixed their GPU crashing by doing this, didn't work for me, but did notice a much slower response on Gen2 settings - to be expected.
19, Downgraded to v40 FW on mobo, this opened up a Gen4 setting(this was disabled by later BIOS upgrades), didn't work but also noticed fan speeds and noise were much rougher - so will go back to highest level.

taking it down to my brothers today, for a (hopefully) confirmation GPU is kaput!

 

[kerberos_20]

if GPU at your brother PC will be fine, then next suspect would be your power outlet wiring

 

[Nanocore]

ok back from my brothers, AND...
GPU works fine in his PC! Ran games, OCCT, all ok, not a single crash or stutter!
I took my PC down also, plugged his 3080 into my PC and all fine, no crashes at all.
Switched RAM sticks - I knew this wasn't going to be a problem given that the 3080 worked fine.

That also rules out power issues, my 650w Ion gold can easily run a 300w+ 3080.

https://nvidia.custhelp.com/app/ans...nvidia-gpu-firmware-update-tool-for-displayid
Didn't work for me!
More relevant to non-POST GPUs rather than crashing ones.

Come on tech gurus, what's going on here?

 

[kerberos_20]

ok back from my brothers, AND...
GPU works fine in his PC! Ran games, OCCT, all ok, not a single crash or stutter!
I took my PC down also, plugged his 3080 into my PC and all fine, no crashes at all.
Switched RAM sticks - I knew this wasn't going to be a problem given that the 3080 worked fine.

That also rules out power issues, my 650w Ion gold can easily run a 300w+ 3080.

https://nvidia.custhelp.com/app/ans...nvidia-gpu-firmware-update-tool-for-displayid
Didn't work for me!
More relevant to non-POST GPUs rather than crashing ones.

Come on tech gurus, what's going on here?

its your electricity wiring, old house maybe?

 

[Nanocore]

its your electricity wiring, old house maybe?

Yes old houses, but 2 of them, in 2 different cities.
My PC runs the 3080 perfectly without issues so rules out house power cabling.

 

[Multyspeed]

Try older drivers. Try a different power cable if you have one (from the wall to the PC).
Try to lower the frequency a bit with MSI afterburner or something similar.
I think you have tried everything else ...

 

[Phaaze88]

Different Nvidia Control Panel settings between the 2 PCs?
One is using Geforce Experience, the other isn't?

How does it crash? Crash to desktop? Shuts off? Black screen, but can still hear things?

 

[Nanocore]

@Multyspeed yes will try older chipset drivers and Nvidia drivers
@Phaaze88 The differences between the 2 PCs is moot, the 3080 works fine in my PC but the 3060 only works in my brothers PC. Not using Nvidia Experience on either PC.
Crash is usually software eg can be recovered, But a few of the things I've done above result in hard crash requiring manual shutdown or automatic shutdown and restart.

 

[Multyspeed]

@Nanocore Update everything to the latest version - Windows, BIOS, and so on.
Download older drivers from Nvidia. Remove the current driver using DDU in Safe Mode.
Install the older driver and don't update it. If It crashes again, try the same thing using an even older driver. Try this 2-3 times.
Do you have any overclockers tools like MSI Afterburner, Palit's ThunderMaster, or something else?
If you have any of them - uninstall them.

 

[Nanocore]

@Multyspeed yes I'm using base Windows install with all updates.
I just can't understand why the 3080 works with the same config e.g. the only change is GPU.
They both use the same drivers, even my brother is running the same, highest level of Nvidia drivers.
I haven't installed any of the 3rd party software or messed about with under/overclocking - I don't really want to go down this route

 

[Nanocore]

I've given up on this issue - mystery unsolved!
Spent too long on it.

Will report back when I get a replacement GPU that works!

 

[Nanocore]

Just to finish up this post, I wasn't successful in getting the RTX 3060 working on my B450 motherboard - YMMV!
My replacement RTX 2070 Super works just fine.
Sadly I can't perform any further tests as the 3060 have been packed and will be going to someone who has no issues running it on their system.