[SOLVED] GPU crashes in-game due to low power, strange issue

Jan 18, 2022
1
0
10
Okay so a few weeks ago I ran unto an issue where in some games the PC crashes with a BSOD (Video TDR failure). Using MSI afterburner's overlay I noticed that the GPU power levels fell too low for the load (it was running at 40ish% power with a 90%+ load at max frequency) which I acertained to be the cause of the crash.
I used DDU to wipe the old driver and installed a new one (the december studio driver), and the issue seemed to go away...

...Until today when it reared its ugly head again while playin god of war, the power levels drop too low and the PC crashes, it happened twice.
The weird thing is the previous day I played god of war for a solid 6+ hours with no issues.
It's worth noting that my GPU is undervolted, but even removing the undervolt doesn't fix the issue.
It's also worth noting that not every game triggers this behavior.
I suspect my PSU may be faulty (it is 3 years old) or my GPU is diying (which is even worse considering the crazy high GPU prices).
Are there any fixes I should attempt before replacing any hardware and is someone familiar with this issue?

Specs:
GTX 1080
R7 1700
32GB RAM
Corsair CX550 550W
 
Solution
Not really, not for that pc. The Ryzen will pull @ 90w when topped out, the rest of the pc might at most pull 75w±. Add in a 180w 1080 and max power load is close to 350w±. That's fine for a 550w psu.

If it's a CX, that's not so bad, it's on the lowest tier of decent(ish) Corsair, better by a good margin than the CV or CXM, and just behind the RM.

It did 6hrs of GoW without a hiccup.

So what's left is a gpu trying to use 40% power with a 90% load. Which becomes a question of chicken or egg. Did the psu only supply 40% power or is there an issue with the power delivery on the gpu and it could only use 40% power in partial delivery, eventually overheating the VRM's actually powered.

The only way to test is swap out one of the...

Karadjgne

Titan
Ambassador
Not really, not for that pc. The Ryzen will pull @ 90w when topped out, the rest of the pc might at most pull 75w±. Add in a 180w 1080 and max power load is close to 350w±. That's fine for a 550w psu.

If it's a CX, that's not so bad, it's on the lowest tier of decent(ish) Corsair, better by a good margin than the CV or CXM, and just behind the RM.

It did 6hrs of GoW without a hiccup.

So what's left is a gpu trying to use 40% power with a 90% load. Which becomes a question of chicken or egg. Did the psu only supply 40% power or is there an issue with the power delivery on the gpu and it could only use 40% power in partial delivery, eventually overheating the VRM's actually powered.

The only way to test is swap out one of the components and see if the issue continues. Easiest route is swap gpu. Cheapest route is swap psu.
 
  • Like
Reactions: jaybarillo
Solution