Unexplained system crashes -- HELP

My main system is crashing with not alot in common. Each time, it's under a stress load of a full RAM check or in Audiosurf 2. I can't tell if the kernel get frozen, or the display crashes. I can't tell if it's from a weak CPU-RAM trace that gets damaged from heat or the GPU is doing something. The crashes are becoming for frequent.

In game, total system freeze. Not even holding the power button turns off the system. Usually.

In memcheck 5.04 it just freezes with all the CPUs appearing locked, but the heartbeat in the top left is still going.

My hardware is as follows:
Xeon E5 2620 v3
4x Kingston 2133 (clocked at 1866) DDR4 registered ECC memory
Samsung 840 Evo 250GB formatted at ext4
2x Western Digital 4TB drives formatted in BTRFS RAID 1
Asus R9 270x 4GB
Asrock extreme4/3.1 x99

This system is running Archlinux.

did you allready tried to re-seat the cpu, and check the socket for any bended pins?

Step 1 reinstall windo...

Step 1 reinstall OS

Step 2 takeout every component and check beep codes while adding them in one by one

Step 3 blame the power supply

@MisteryAngel I don't have the equipment (thermal paste) to mess with re-seating the CPU, but it's been in there rather stable for some time.

@Streetguru It's not software related. And the power supply should be about 400W over-spec'd.

I second this. Have you a spare PSU lying around that you can test the system with?

Have you tried testing the RAM individually?

Test the system out with 1 stick or ram using each stick 1 at a time if it crashes with each stick its not the ram.

I doubt its the ram but this is a hardware issue. Since holding down the power button is separate from the O.S. Your psu most likely is dieing which you should replace fast before it kills your whole system.

If the cpu is not fully seated correctly, (which happens very often with X99),
then weird system crashes can happen.

UUhhhhh......looks like LTT had their storage server die, and it looks like they have the exact same board.

What are the chances of a consistent manufacturing flaw causing these to die?