[SOLVED] PC shuts down immediately when trying to stress GPU

PC shuts down under certain cases. Looks like it is on sleep/hybernation, since power light is ON, but does not wake. I can only fully shut down and boot again. This happens when stressing the GPU on the exact same point.

-Example 1: Running 3DMark, crashed when the animation of unfolding clock starts. Always on the same frame

-Example 2: Playing PUBG, at random points

-Example 3: Running Furmark in fullscreen on resolutions above 1920x1080. Runs with no problem fullscreen on lower resolutions

System Configuration

  • Motherboard: Gigabyte X570 Aorus Master (BIOS: 31J)
  • CPU: AMD Ryzen 9 5900X
  • PSU: Corsair CP-9020074-UK HX1000i 1000 W 80 Plus Platinum
  • Display(s) (Make, Model and Type): Eizo Foris FS2735 (1440p 144Hz FreeSync)
  • System Memory: Crucial Ballistix 2x32GB DDR4-3200 UDIMM 1.35V BL2K32G32C16U4R
  • GPU (VBIOS): AMD RX 6900XT 16GB (020.001.000.043)
  • OS (Version): Windows 10 x64
  • Driver: Radeon Software Adrenalin 2020 Edition 21.3.1
  • Background Apps: Steam (or none in case of benchmarks)
  • Radeon Software Settings: Default options for Gaming

Steps to Reproduce

  1. Start Furmark
  2. Set graphic settings to 1920x1080 Fullscreen
  3. Click “Stress test GPU”
  4. PC immediately shuts down but power led remains ON

It seems like power problem. Can you check the power kabels from motherboard (20pin/24pin and 8pin ore 6pin) and gpu are connected. do have 2 set power pins connected in both. Reconnect them in power supply asswell.

Thank you for your reply.
If it was a power issue, wouldn’t be happening all the time in all resolutions?

Yes and no. Stressful power hungry stress test will activate and ask for more power most of the time wil just use the basic power from one power line. Or just power from motherboard. It power up it wil open addre power lines when the GPU or CPU needs more can’t get the power second power line it wil kill and black screen ore frees the computer.

If it’s not power cable problems you can look at software based power problems than you can check if there are bios firmware and GPU firmware updates.

Could be memory errors. Run Memtest86+ from the BIOS level and stress-test the memory first. Make sure you’ve updated your BIOS to the latest version to support the CPU and/or GPU! (I got caught last week!)

Memory is also good one check out.

Removed all the cables and put them back in. None looked like it was loose or teared.
The only thing I noticed was that I have one cable for PCIe power with two 8pins at the end, instead of two cables with single pin each. But I remember when I got the PSU back in summer, that the manual said it doesn’t matter if I use one or two cables, as the power comes from the same lines.

I’ve run into this issue before and it turned out to be a power supply, when GPU power demand peaked the PSU couldn’t keep up and would shut down. Try swapping in a known good power supply.

Since it is not the easiest task, I have to be 100% sure that this is the problem.
What would you consider better than the “Corsair CP-9020074-UK HX1000i 1000 W 80 Plus Platinum” I already have?

I’d say that an easier troubleshooting step might be to disregard the manual and try 2 separate cables to the GPU. It can’t hurt to have more copper carrying your power, especially if it could be a steep increase in voltage requirements for the card that’s causing an issue. I say test that before anything else.

1 Like

That is what I just tried, and seems the problem solved.
I am using 2 different cables, instead of 1 with 2x8pins and I was able to run Furmark for 20 minutes in 2560x1440 fullscreen without an issue.

Thank you all for your input!

1 Like

It’s both amazing that “disregard the manual” was the answer and upsetting lol.

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.