System shuts down during gaming semi-predictably but NOT due to overheating.

Rig:
R7 2700x (not overclocked)
ROG x470 ITX mobo
32gb 3200 GSkill Bdie RAM @3200
Vega 64 LC edition
970 EVO 500GB
Silverstone 800W SFX-L PSU (using PCIe cable for each GFX power connector, no daisy chains)
Custom LC Loop (420mm rad + 120mm rad)

Computer will shut down during gaming, but GPU temps <60C. CPU <60C. Replaced PSU twice, no effect.

Managed to watch GPU hotspot temps on GPU using Aida64 free edition and saw them over 100C. Thought that was the problem RMA'd the GPU. Got the new one in, hotspot is better but still does it.

Tried re-seating the CPU, no effect.

When it goes, it spontaneously turns off and requires the PSU to be cycled (by unplugging it or switching it off) before the power button will work (hence why I keep thinking its a power issue, but I replaced the PSU).

I've disconnected and reconnected every power cable in the system.

I can run torture tests (AIDA64, GPU-Z, Heaven Benchmark, Firestrike, etc) for hours. Temps stay under 60, and even stressing GPU and CPU at the same time to 100% and it does NOT crash.

Any thoughts? I tried WhoCrashed, but there are no logs.

Event Viewer just shows EVENT 41, Kernel-Power "The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly."

-----

Any apps out there that can log what's going on to USB or to the hard drive to catch what is going on right up to the instant of the shutdown? All I can think of is the motherboard or a driver issue - but like said it's NOT temperature or seeming power use. Maybe changing P-states? Ugh.

The five games I'm playing these days, it happens in two of them. Fallout 4 and Witcher 3 crash, Starcraft II, Heroes of the Storm, and Hearthstone are fine. I have found a part of the map in Witcher 3 there it happens consistently if the game in run in windowed mode 2560x1440 with everything on its max settings, so I've got a save file in that spot just for testing purposes.

Whats your thermal paste contact patch look like? thick on sides, thin in middle? or even across?

and have you updated all your drivers? could be a faulty GPU as well...

My only guesses are either power supply or motherboard related. Weird reboots and shutdowns can be caused by blown capacitors and such. Quite unlikely on a new system, but can maybe explain the irregularities.

Kernel power 41 shutdowns are only power or heat related.
Since you have tried the obvious PSU change out, the only things left are CPU and mobo.

CPU is a longshot as failure is rare at best.
Best advice: RMA the board.

I wonder if you are trying to run too much RAM at that frequency. Try giving the IMC a little more juice or lowering the frequency. The stress test you have run may not be adequately stressing the RAM whereas some of the games may be.

Try bumping up the amount of RAM being used during the stress test. You can customize this in P95, IBT and Realbench stress tests.

I got a Vega 64, R7 2700, Phanteks Evolv Shift (not the X) and a Vega 64 reference cooler to build my brother in law a christmas PC on black friday - I tried swapping his Vega 64 and PCIe riser cable for my Vega 64 and riser cable this afternoon and I haven't gotten it to crash yet, so either the liquid cooled card is pushing some part of the system past stability or the riser is the questionable item. I could swap his riser for my riser, but I'd have to dismantle the water loop (right now his Vega 64 is sitting on a box outside my case plugged in with the cable).

Anyone recommend a high-quality 220mm-250mm PCIe riser cable? Phanteks looks like they have a "normal" one which matches what I have and a "premium" one, but I'm not sure its any better.