Ned help to understand whats wrong

[prl]

Hello!
I have an Intel NUC i7 gen 10. It has 16 GB RAM divided into two RAM modules.
A while ago I could not start Windows 11. All attempts to get Windows started failed. I subsequently tried reinstalling Windows which failed. I updated the UEFI but it did not help.
I now began to suspect a hardware failure. I tried installing on another hard drive with the same result.
I then ran memtest86, first testing one RAM module, then the other. So I tested the two 8 GB modules separately in the same memory slot in the computer. Both got a pass. Then I tested one of the memories in the other slot. Again a pass. Then I inserted both RAM modules again and ran another test. This time the test crashes and the computer shuts down. This happens during the first pass and at the end of test 7 (moving inversions, 32-bit pattern) The computer crashes at the exact moment test 8 starts. Test 8 is what I understand ‘Random number sequence’

What is the problem?

[Roxxors]

That could be a hardware issue, but it runs to test 7 then crashes so the hardware works. I would check the CPU cooler and make sure the fan is running or not clogged up. That smacks to me of a thermal shutdown because test 7 generates a lot of heat.

[David (PassMark)]

Testing two sticks at once changes the way the RAM works. It switches to dual channel mode (interleaving access). It also adds electrical load to the system & EMI. More current means potentially more voltage drop (if the power supply isn't sufficient). We've seen several cases now where RAM errors are only found in dual channel mode.

As it worked OK in the past, assume it is the RAM that has gone bad and replace it. Or borrow some known good RAM to test.

See also
https://www.memtest86.com/troubleshooting.htm

[prl]

I inserted two brand new RAM modules and tested.
But memtest crashes again in exactly the same place in the test.
RAM temperature according to memtest was 61 degrees (Celsius)

I then tried to limit the processor to four cores. (It has six)
Now the computer managed to get through tests seven and eight and completed the whole test without crashing.
At the end of test seven, the temperature was 53 degrees. That's eight degrees lower than when the system crashed with all six cores running.
Still, it has always managed to use all the six cores previously.

[David (PassMark)]

Temperatures of 50C - 70C are pretty common for RAM and CPUs. Some laptops run at over 100C for the CPU.

Doing an BIOS update (if you can) and a reset is probably also worth it, as it is free and quick.

As the fault didn't change with new RAM sticks, next most likely cause is CPU. With a small chance it might be the power supply or motherboard.
Was the CPU in a socket in this machine?

[prl]

I have already upgraded to the latest BIOS. It was among the first things I tried, but it did not help.

Yesterday I enabled all six cores but turned off in BIOS what is called “Intel Turbo Boost Technology”. This time the computer passed the whole memtest procedure, but it took a very long time to complete the test. Next, I went into UEFI again and selected “optimal defaults”, but it did not work. Then the computer crashes again after test #7.

The computer is an Intel NUC so the processor is not in a socket but soldered directly to the motherboard.

Does Intel NUC often break down? Does the NUC have known quality problems?

[prl]

It seems that the problem is solved.
It is the PSU that is failing.

I have an older 6th generation Intel NUC with an i5 processor. I tried using the PSU that the older NUC came with. The PSU for the older NUC has the same voltage of course, but it is capable of delivering almost twice the power. With this PSU, everything suddenly works as it should - at least what it seems so far.

Now that I think about it, I have for a long time had problems with Windoes suddenly shutting down, ie shutting down the computer. But it has always been possible to restart the computer again. Probably the PSU has had problems for a long time until now, when it suddenly became impossible to start the computer at all. Maybe it is capacitors that fail.

[David (PassMark)]

Posted before your reply:
There have been several case studies into CPU failures in data centers. Typical failure rates are around 0.5% per year, but they also noted that many of these failures were actually a result of cooling and power supply issues. And the failure rate is significantly higher in the first couple of weeks / months. After the initial period, failure rates were lower (more around 0.1%).

So maybe you got unlucky. Maybe it would be possible to under clock the CPU and get it working, or permanently disable some cores (if the BIOS allowed that).

Also as Roxxors said, it is worth cleaning out any dust in the system and check the fan(s).

After your reply:
Good news. Most PSU failures are a total failure. It is rarer to get have one partially working so that it always crashed at the same time, at test 8

[prl]

[B]

After your reply:
Good news. Most PSU failures are a total failure. It is rarer to get have one partially working so that it always crashed at the same time, at test 8

Yes, I know. That's why I didn't suspect the PSU. The suspicions did not go in that direction.
Thanks for the help anyway.