• Notice regarding image hosts and attachments. See here.

PC restarting

Soldato
Joined
29 Aug 2010
Posts
8,901
Location
Cornwall
I posted this in general hardware but since my current suspicion is that it might be a PSU issue I thought I'd repost here:

So I had a Ryzen 5500 setup but every time I put my 7900xt in the system I would get usb disconnect issues.
Tried everything I could think of.

So upgraded to a 7500x3d and Asus TUF Gaming B850M-E WiFi.
Now I get occasional reboots. I've noticed it mostly after playing Diablo 2 Resurrected for a bit. When I save and go back to the character select screen it freezes for a second or two, then the screens go black, then the pc reboots.
Doesn't happen every time though.

Also got it while running OCCT (testing memory, 3D Adaptive (I think) and vram).

Event viewer shows an event ID 41 with a bugcheck code of 0.
AI tells me this means the power just died and suspects over current protection on the PSU. It suspect the GPU might be spiking a momentary high frame rate when it goes back to the menu screen.
The PSU is old, it's a 750W EVGA Supernova G3 from I'm not sure when but had to be 10+ years old.
I didn't notice this issue with my old setup though (just the usb disconnects).

I'm running memtest86+ and it's done 1 pass so far with no errors.

Anyone got any ideas?
I guess I'm hoping it is the PSU as it'd be cheaper to replace them the GPU, CPU or RAM.

I think they only recommend 750W for a 7900xt and the rest of the system is quite low power I believe with the CPU only being 65W.
 
Have you tried undervolting the GPU to see if it eleviates the issue.
I had a very similar issue which turned out to be a duff RAM stick. As you have now used two systems, I would aim down the PSU route first though.
 
I ran 3 passes of memtest86+ (v9) and no issues.
Then ran OCCT on the CPU and memory for an hour and it was fine.
Then I ran OCCT on the GPU and VRAM and it rebooted after 11 minutes.

I'm hoping it's just transient spikes or whatever from the PSU.
 
Thats why I was wondering if you could power limit the GPU and see how it goes from there.
Any indication as to what the GPU temps and hotspots are doing?
 
Thats why I was wondering if you could power limit the GPU and see how it goes from there.
Any indication as to what the GPU temps and hotspots are doing?
Tempts are all good (or where when running OCCT), mem junction was the highest I think at 78ºC.
I have set the power limit to -10 (which is the max I can do) and done a slight undervolt. I think it peaks at 288W TBP.
 
Small update...

I've ordered a new PSU, a Montech Century II 850W. I'm sorta hoping for miracles.

Another curve ball was thrown this evening when I was watching Amazon Prime Video on a 3rd monitor (well, TV. I've got 2 x 4K screens and a 1080p TV connected to it) and it froze and restarted on me then too.
Now I'm not sure if it is the PSU as that likely wouldn't have been stressing it too much.
 
If it ain't the PSU then it's probably the GPU. Any spare system to try out the GPU separately?
I'm hoping it's not.
I have tried a different GPU in the system and it happened with that GPU too.

EDIT: Today following the PSU swap it crashed while idling on the desktop with just a web browser open.
 
Last edited:
Just so I've not got the wrong end of the stick:

- The original 5500 system was working fine initially
- You added the 7900xt to the 5500 system and then the crashes started
- You then bought a new 7500x3d system and it crashes with the 7900xt
- You tried the 7500x3d system with a new PSU and it still crashes
- You tried the 7500x3d system with another graphics card and the new PSU and it still crashes.

Just my thoughts:
1) Is the bios reset and at stock? This would be my first suggestion if not already done so
2) If bios is reset, then add a +5 offset to curve optimiser and see if the crashes still happens.
 
Not quite.
5500 was fine with the 4060Ti. When I added the 7900XT I started getting USB dropouts where the keyboard and mouse would occasionally disconnect.
Bought the 7500X3D and motherboard and used the 7900XT. Crashes started.
Swapped in the 4060Ti and got another crash. At this point I disabled the PBO curve optimizer and set "Power Supply Idle Control" to "Typical Current Idle". Crashes seemed to stop.
Swapped the 7900XT back in and while less frequent I did get a crash. I disabled HAGS in Windows.
Then I swapped the PSU and set "Power Supply Idle Control" back to "Auto". It crashed again.
I set "Power Supply Idle Control" back to "Typical Current Idle" and am testing.

Unfortunately there's not an obvious way to test it. The last crash happened when I was AFK and it was just idling at desktop (screensaver had probably kicked in).

BIOS is now more or less defaults with the exception of fan curves, disabling onboard graphics and I do have EXPO enabled (but it's only 5600MT/s so not pushing anything silly).
RAM came from an old system so has been used previously (also ran memtest86+ 3 times with EXPO enabled). GPU also came from a previous PC.
 
I'd fully reset bios and run the ram at stock in the first instance.
If there's further crashes then put curve optimiser at +5 per core
 
Last edited:
Running completely at stock for a while seems like the cleanest way to separate a memory or curve-setting issue from the 7900 XT itself.
 
Back
Top Bottom