[SOLVED] Is my PSU causing CPU overheating?

[acerninetes]

My computer crashed while gaming today and wouldn't restart for around 20 minutes. After some diagnostics I concluded that the CPU overheated. Testing OCCD's CPU stress test simultaneously with Furmark I get CPUtemps around 80c, with both CPU and GPU around 100% usage simultaneously, however running OCCD's PSU test (which also puts the CPU and GPU under 100% strain) the CPU temp shoots straight up to 90c and keeps climbing (I stopped the test after it hit 95c). I repeated all this with my case's side panel off and got the same results. I'm confused as to why these two tests give such different outcomes - does this suggest my PSU is causing overheating in the CPU (and if so how is this possible)?

The relevant specs are:
PSU: XFX 550w Pro (I think this is a Bronze psu)
GPU: RTX 3070
CPU: Ryzen 5 3600

I know I'm pushing the PSU pretty hard with this set up, so it wouldn't surprise me if that is the problem. I would really like to understand what is going on here!

 

[DSzymborski]

This would be highly unusual. The far more likely culprit is something going on with your cooling solution. A PSU is only going to release so much heat before it shut down to OTP and the fan you see on a PSU is intake not exhaust.

 

[acerninetes]

This would be highly unusual. The far more likely culprit is something going on with your cooling solution. A PSU is only going to release so much heat before it shut down to OTP and the fan you see on a PSU is intake not exhaust.

That makes sense, I certainly don't think the PSU is physically putting a lot of heat out. I am assuming there is some nuance in CPU/GPU usage between the tests that means 100% load isn't equivalent for both. How might I try to isolate the problem here? I assumed my cooling was okay since under the two separate tests run simultaneously the temps were good, but clearly there is something wrong.

 

[Archaic59]

That makes sense, I certainly don't think the PSU is physically putting a lot of heat out. I am assuming there is some nuance in CPU/GPU usage between the tests that means 100% load isn't equivalent for both. How might I try to isolate the problem here? I assumed my cooling was okay since under the two separate tests run simultaneously the temps were good, but clearly there is something wrong.

• If you're using the stock Ryzen cooler you should consider replacing it with a good air cooler. Check your chassis (case) specs for cooler clearance before buying one. If you need help selecting one just ask but first update your specs with a complete list.
• Your PSU is very old, at around 10 years or close to it and while not much chance it's causing any problems, waiting for it to fail from age is not a great idea either. There are perishable components in the PSU and it can lose it's ability to provide clean stable power as it grows progressively older. If you're in the US you can expect to spend around $100-150 for a good brand/model.

 

[sizzling]

To me there are 2 strong indicators of a psu problem and together I’d say it’s very likely a psu problem.

First a crash while gaming. Overheating will normally present with lost performance first.

Second needing to leave it for 20 mins to be able to switch back on.

 

[kanewolf]

To me there are 2 strong indicators of a psu problem and together I’d say it’s very likely a psu problem.

First a crash while gaming. Overheating will normally present with lost performance first.

Second needing to leave it for 20 mins to be able to switch back on.

The motherboard wasn't listed. Nor was the type of CPU cooling. I might suggest a VRM cooling issue. Running with the sidepanel off will help identify insufficient inflow fans. BUT without a desk fan pointing at the motherboard to ensure the ENTIRE motherboard is not overheating, it is impossible to eliminate the VRMs.

 

[sizzling]

The motherboard wasn't listed. Nor was the type of CPU cooling. I might suggest a VRM cooling issue. Running with the sidepanel off will help identify insufficient inflow fans. BUT without a desk fan pointing at the motherboard to ensure the ENTIRE motherboard is not overheating, it is impossible to eliminate the VRMs.

It might be heat but it has to be a very fast increase to not notice any throttling first. Even VRM overheating will throttle the cpu. My vote is psu seems more likely.

 

[Juular]

I am assuming there is some nuance in CPU/GPU usage between the tests that means 100% load isn't equivalent for both.

OCCT has two types of CPU stress test, 'regular' one and Linkpack. 'Regular' test also has three settings for the data sample size, the smaller the size the higher the load and thermals. So called 'PSU' stress tests uses Linpack, which is according to my tests somewhat in between of large and medium data sample size of 'regular' CPU stress test, but smallest size still gives higher temperature. I don't think your problem is related to CPU overheating, unless you have an AIO with dead pump, 20 minutes is way too long for the cooler to cool down. Have you by any chance detached the power cord from the PSU or switched it off and on in that time ? That would explain the 'delay', if the PSU latched off because of tripped protection it wouldn't let you start the PC again unless you unpower it for a moment. Now, there's like 5 or 6 XFX PSUs with the 'Pro' in it's name, what's the exact model ?