GPU Stressing Causes Power-Off

[ProteusCortex]

Hello,

 

I have quite a strange issue. When pushing my GPU to 100% my system just shuts down and then restarts. The power supply is more than enough (850W) and it used to power 2 GPUs at full load with no issues whatsoever. What I have tested so far is:

 

1. Switched to another power supply (just to be sure).
2. Switched to an older GPU (nvidia GeForce GTX 650Ti).
3. Switched PCIe slots (just in case there is an issue with the mobo's power supply to that slot...?)
4. Tried the latest "recommended" drivers from AMD.

Every time I'm running FurMark and it doesn't even get to show the furry donut. The system just shuts down.

 

Any ideas on how to narrow it down?

[Dissitesuxba11s]

Did you overclock it at all, because it might be an unstable OC?

[jj9987]

Very likely a hardware issue and my bet is on motherboard.

[ProteusCortex]

1 minute ago, Dissitesuxba11s said:

Did you overclock it at all, because it might be an unstable OC?

No. No overclocking for the GPU whatsoever. And for the CPU is just the default OC from the mobile (about 5%).

[Frankenburger]

The PSU might be going bad. How old is it? Do you have another PSU you can swap out to verify it's not an issue related to power delivery?

[ProteusCortex]

Just now, jj9987 said:

Very likely a hardware issue and my bet is on motherboard.

Any idea on how to test this?

[ProteusCortex]

Just now, Frankenburger said:

The PSU might be going bad. How old is it? Do you have another PSU you can swap out to verify it's not an issue related to power delivery?

The PSU is about a year old. I also tested my old one and I still gave issues.

[knightslugger]

did you ever increase the current capability of the VRMs?

Is anything overheating (CPU/GPU/VRM)?

[ProteusCortex]

3 minutes ago, knightslugger said:

did you ever increase the current capability of the VRMs?

Is anything overheating (CPU/GPU/VRM)?

Never touched the VRMs. I have also stress tested the CPU and it reaches a maximum of 70°C and the system seems stable. Also the last temperature I get to see for the GPU is 50°C before the system shuts down. I haven't checked the VRM temperature in any way.

[knightslugger]

3 minutes ago, ProteusCortex said:

Never touched the VRMs. I have also stress tested the CPU and it reaches a maximum of 70°C and the system seems stable. Also the last temperature I get to see for the GPU is 50°C before the system shuts down. I haven't checked the VRM temperature in any way.

It sounds to me that something is triggering a safety. Sudden shutdown are usually a ground fault or thermal protection from the MB. Especially with two DIFFERENT PSUs. Check your wiring loom.

[ProteusCortex]

17 hours ago, knightslugger said:

It sounds to me that something is triggering a safety. Sudden shutdown are usually a ground fault or thermal protection from the MB. Especially with two DIFFERENT PSUs. Check your wiring loom.

OK. It looks like the "minimum requirements" for the GPU's PSU are not that accurate or my older power supply cannot actually output 500W (which is also the minimum PSU requirement for the GPU). The fact that it couldn't handle the older GPU shows that it has its issues too...

[ProteusCortex]

17 hours ago, knightslugger said:

It sounds to me that something is triggering a safety. Sudden shutdown are usually a ground fault or thermal protection from the MB. Especially with two DIFFERENT PSUs. Check your wiring loom.

I measured the grounding from the outlet to the UPS. Also between the outlet and the case. Everything seems to be OK. Couldn't measure voltages on the motherboard, though. My hands are not that steady.

[Pi31415]

Hit or miss I guess they never miss huh? You got a power supply, I bet it can't handle your GPU ya! all jokes aside its your PSU even if it meets the requirements you can't have something off ebay (Generic and unbranded). Or it might be heat I didn't read all the replies sorry good luck!

[ProteusCortex]

5 minutes ago, Pi31415 said:

Hit or miss I guess they never miss huh? You got a power supply, I bet it can't handle your GPU ya! all jokes aside its your PSU even if it meets the requirements you can't have something off ebay (Generic and unbranded). Or it might be heat I didn't read all the replies sorry good luck!

Well, in short:

- My problematic PSU is an EVGA Supernova P2 @ 850W and has handled 2 GPUs in the past.

- The old PSU is a Corsair CX500 @ 500W. Nothing off ebay.

- Pretty much nothing gets to heat up because no benchmarks can actually run.

- Made a Frankenstein's Monster with a second PSU (that one is indeed no-name, but it works fine) to power the GPU only and everything seems to be OK.

- I'll just wait for the problematic PSU to be replaced.

[Pi31415]

1 hour ago, ProteusCortex said:

Well, in short:

- My problematic PSU is an EVGA Supernova P2 @ 850W and has handled 2 GPUs in the past.

- The old PSU is a Corsair CX500 @ 500W. Nothing off ebay.

- Pretty much nothing gets to heat up because no benchmarks can actually run.

- Made a Frankenstein's Monster with a second PSU (that one is indeed no-name, but it works fine) to power the GPU only and everything seems to be OK.

- I'll just wait for the problematic PSU to be replaced.

Can you do a CPU stress test then a GPU stress test and see which one powers the PC off when it is stressed? Might not be gpu...... lmk

[Domrockt]

Isnt that PSU released in 2014 ish? I bought myself a 2080ti and it keept crashing after short periods of Time. 

 

I owned a corsair HX850 from that time also 2014 ish.

Now i bought a Brand new 750 Watt Corsair HXi Series HX750i Modular 80+ Platinum

Its release was last Year. So babidibubidibob all problems solved. 

Newer Tech with less cushion for the pushin but does the Job.

[ProteusCortex]

15 hours ago, Domrockt said:

Isnt that PSU released in 2014 ish? I bought myself a 2080ti and it keept crashing after short periods of Time. 

 

I owned a corsair HX850 from that time also 2014 ish.

Now i bought a Brand new 750 Watt Corsair HXi Series HX750i Modular 80+ Platinum

Its release was last Year. So babidibubidibob all problems solved. 

Newer Tech with less cushion for the pushin but does the Job.

I bought it in 2017 brand new. So it doesn't really matter when it was first released (the P2 version was released in 2016, I think), because electronics have a huge shelf life. It's only when electrolytic capacitors are involved that you have to let them powered without load for a bit and that applies mostly to electric motor drives, that really push them to the limit.

 

So yes, it was as new as it gets. I was powering 2 GPUs with it when I bought it and for about 10 months (an RX480 and an RX580), so I don't think there was an issue until recently. I also had a UPS for the last 11 months, so there shouldn't be any issues with the grid power quality.

[ProteusCortex]

16 hours ago, Pi31415 said:

Can you do a CPU stress test then a GPU stress test and see which one powers the PC off when it is stressed? Might not be gpu...... lmk

I already did that. The CPU (overclocked @ 4.5GHz) would run OCCTPT's Linpack test with AVX, on all logic cores and using 90% of RAM for more than an hour, reaching a maximum of 70°C with no issues whatsoever. Only when pushing the GPU I encountered issues.

 

After adding a second PSU that powered only the GPU I could run all of OCCTPT's tests (even the "Power Supply" test) without the system shutting down.

[ProteusCortex]

The main issue was solved. I initially thought that it was the power supply and RMA'd it, but when using my old power supply I encountered the same issue and thought it must be something else.

 

My GPU (Sapphire Radeon RX580 Pulse Edition) requires a PSU with a minimum rating of 500W. My old PSU is a Corsair CX500 with a 500W output rating, so I thought it should be OK. Turns out my old PSU has some issues of its own and can no longer reach its rated output of 500W, which my system seems to pull at maximum load.

 

This is because I added another 300W PSU, just to power the GPU and everything could be stressed to 100% without any issues.

 

So I will just wait for the RMA of my main PSU to go through and everything should be fine.

 

Thank you all for your help. You really stirred me back to the correct solution.