site banner

Small-Scale Question Sunday for August 27, 2023

Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?

This is your opportunity to ask questions. No question too simple or too silly.

Culture war topics are accepted, and proposals for a better intro post are appreciated.

3
Jump in the discussion.

No email address required.

Posted here for new eyeballs:

For the past few months, my pc has been consistently crashing under heavy load, in graphically demanding games like Escape from Tarkov, Warhammer 3 etc. In normal use and less intensive games like Rimworld, no issues.

After about 15-20 min of gameplay, I get a full crash to a black screen with the pc powered off, and it refuses to post for several minutes regardless of what I do, at which point it often restarts on its own. The American Megatrends screen doesn't usually show up unless I cycle power, at which point it doesn't tell me anything useful either. (I had a similar issue around 5 years ago, but that was almost certainly CPU thermals, since the American Megatrends screen called out CPU overheating, and I've changed CPUs since then and don't get the same error)

The crash seems to be so total and abrupt that I can't find any useful logs to figure out wtf is going on.

I've run CPU and GPU stress tests on OCCT and furmark, and they only seem to cause issues unreliably.

It seems thermally related, since the problem is less severe when the AC is running, but unfortunately the AC is currently on the fritz exacerbating the issue, but fixing the AC isnt really a definitive solution is it?

I noticed >85° C temps on my Ryzen 5600x, so I changed the thermal paste just a few days back, and while temps dropped by 5-10 degrees, the crashing hasn't abated.

GPU temps hover in the 50s-60s range in Tarkov, which seems quite reasonable. It's a 3070 for what that's worth.

The other potential culprit is my geriatric 600w power supply, over 10 years old at this point, but why would it be thermally related?

I'm not running any OCs, and I've maxxed out my fan curves to help, not that it's doing much. My case has two extra blowers, and I even took off the sides to help with airflow.

Anyone have any idea as to how I can figure out what exactly is wrong? I can't really afford to replace my GPU, but I could consider buying a new PSU if need be.

This issue didn't plague me when I first built this current setup with the same components, but it's been several months and I'm losing my mind :(

What I've tried:

  1. Switching GPUs with my brother's pc. Couldn't reproduce crashing.

  2. Dusting pc

  3. Repasting thermal paste

  4. Checking for any OC (none)

  5. Stress tests, which are unable to reliably cause crashes while games can.

It is worth getting a bigger power supply for sure before doing anything else; every weird computer problem I have ever failed to troubleshoot was resolved by getting a new PSU.

You can get a 750w 80+ gold pus on amazon with free returns; give it a shot and see if you can reproduce the crashes.

+1. You could get a power meter to see if you're near 600w, or use a calculator, but it seems likely. You've clearly put work into having good airflow/thermals, so this is the main thing that's left.

It might be expectedly failing before 600w, too. IIRC that rating is the sum of what can be put out at each voltage, and it's possible to exceed the maximum draw at one voltage without exceeding the total max, and combining newer components with older PSUs makes that imbalance easier to trigger.

it's possible to exceed the maximum draw at one voltage without exceeding the total max

This is also manufacturer-dependent; higher-quality power supplies tend to be able to supply the vast majority or all of its rating on 12V alone, while junk PSUs tend to list "500W" but neglect to mention that it can only output 300W as 12V (so from a modern PC standpoint it's only a 300W PSU- the better CPUs can pull 180W 24/7 and even ancient GPUs need about 200W- so if you try that you're already tripping overcurrent... if it even has one, that is).

Basically nothing uses 3.3 or 5V any more (aside from non-NVMe storage), to the point the new ATX standard removes them entirely (instead relying on the mainboard to downconvert 12V input, something it already does for the CPU to turn an input of 12 volts at 15 amps into 1 volt at 180 amps).