Hello folks,
I'm looking for some suggestions on where to look for troubleshooting.

Problem:
My desktop will power off (as if the power cable were pulled) when under high graphical load (gaming) with no warning or errors that I can find. The power-off is sudden, one moment I'm fine, then in a blink of an eye the screens are black and the PC is shut down. To turn the PC back on I need to toggle the switch on the PSU off, then on again. Then I can hit he power button.

No issues under normal usage, this only happens while gaming or stress-testing the GPU.

What have I tried to fix it?

• RMA'd the graphics card and they sent back a new replacement.
• RMA'd the Motherboard and they sent back a new replacement.
• Upgraded my PSU from a 500w to a new 600w PSU.
• Checked all the cables to ensure they're plugged in properly and snug.
• Compiled the Unreal Engine (which takes just under an hour on my machine) with no problems. I'm assuming this was a good CPU stress test.
• Run additional stress tests via Phoronix Test Suite. No problems with CPU tests but the problem persists with Graphics tests.
• I've checked
  Code:

  journalctl -b -1

  but I'm not seeing anything that would indicate a problem
• I've also run
  Code:

  journalctl -f > ~/journalctl.log

  while replicating the problem, but again no signs of a problem.
• Ran memtest86 to test the RAM. No problems.

My PC Specs:Any suggestions on where to start looking for clues to where the problem is would be very helpful. I don't think it's an overheat issue, I've watched the sensors output while testing and it never goes into a critical temp. I'm pretty much at a loss of where else to look for a problem.

Gaming is not only high GPU stress, it's high CPU stress. The APU hitting critical temperature is an automatic trigger for power off. Things usually shut down elegantly, but depending on the BIOS it could be kneejerk.

What video driver are you using, the default Radeon open source drier, or the propitiatory driver from AMD?
I assume that all the cooling fans are working and that the case has good air flow. Is the fancontrol program seeing the fans and temp settings?

Since it only happens when under high GPU load, it seems to be hardware or driver related.
You could try tailing /var/log/faillog, /var/log/kern.log, and /var/log/syslog, with the -f option to follow the log as it updates like so that way yo may be able to see the error if it does not get written to disk.

You could try the other AMD driver and see if that makes a difference in your stress test. Might be a bug in the open driver?

@scriptmonkey: Those temperatures you showed us were at tickover, not gaming.

Let's see some temperatures from high cpu/gpu loads - 5 minutes and 10 minutes into a game. You might seriously need to upgrade the fans and cpu heatsinks. If they're not black, painting them matt black improves conductivity. That's been laboratory tested. In fact there used to be a paint sold in motorbike shops called 'cylinder head black' for the 2 strokes which were always on the edge of heat problems, the way they were driven here anyway.

So I've adjusted the fans. Instead of just selecting "Performance" mode in the BIOS I went ahead and manually set the fans all to turbo. I haven't had a chance to really test if this change made a large difference or not. I played a game for about 20 minutes today mainly just to capture the sensors output after running the game a while.

I'll do more testing when I get a chance to sit down for a while.