Persistent BSOD ntoskrnl.exe

I could easily see the pins of a Threadripper motherboard CPU socket getting physically damaged. I would expect the CPU to be considerably more durable but I don't have hands-on experience with a Threadripper CPU. Basically, I would expect a motherboard problem over a CPU problem unless the failed overclock included too much voltage.

I don't know if it's possible to damage a Threadripper CPU this way but did the sequence of events lead to the possibility of pins shorting while the CPU was getting power? Experience with other CPUs would make me expect the CPU to be far more unstable if such damage had occurred but perhaps it could be that subtle.

Got another BSOD as I was attempting to respond to this. Here's the minidump. It now definitely seems to be tied to those memory slots. But first I'd like to describe the sequence of events that led me to my current state:

1. I decided that I wanted to OC my 1950x and as a part of that wanted to re-apply arctic silver to get the best possible temps from Liqtech 360. I purchased some arctic silver two-step solution to remove the old arctic silver and (per my research) applied it in a way I thought was correct. I then reapplied a large blob (again per my research), reinstalled my AIO and attempted to OC.

2. Admittedly I wasn't very conservative in my attempts to OC the chip and did indeed add voltage without thoroughly testing it first (or in other words skipping voltage increments). I think the most I got up to was 1.35-1.3875. On my last Prime95 test the system seemed to be running good for about an hour and I decided to let it continue to run. Before I went to bed I came to check in on it and noticed the system was hung on a BSOD (Windows was stuck collecting info at around 30% and the percentage wasn't advancing). I don't know how long it had been stuck like that but I think it was for at least an hour or so. I turned the system off and went to bed.

3. I awoke the next morning to power on the system and it wouldn't boot. I first tested memory and could ultimately only boot with one DIMM installed. I ultimately took the CPU out to discover that a mix of two-step and arctic silver had gotten underneath the chip and into the CPU socket as well as on the gold pads on the bottom of the CPU. It was determined that this was from applying to much two-step remover with the CPU still installed in it's socket. It also looked like some arctic silver had mixed with the two-step and had gotten underneath the CPU as well. It should be noted that this was occurring while the CPU was on and attempting to be OCed.

4. I was not confident in my ability to clean the CPU socket so I purchased and installed a new, identical motherboard to replace the old motherboard. I then reinstalled all DIMMs, cleaned the underside of the CPU with arctic silver two-step and reinstalled the CPU. Shortly after is when the recurrent BSOD (typically tied to ntoskrnl.exe) started occurring.

5. Throughout the history of this thread I tested many things with the exception of a full Memtest86 on all memory (because it was time prohibitive) and was only able to gain long-term system stability with half my RAM installed, however (per the motherboard manual) the RAM was installed in slots D2, C2, B2 & A2. Now (per cwsink's suggestion) system instability was able to be reproduced when RAM was installed in the 1 slots (D1, etc.)

5. A side note is that the pump on my Liqtech 360 actually died right a few weeks after this occurred and was replaced with LiqtechII 280. I only note for posterity and don't get the sense that the current system instability is related to it.

Thoughts?

I think we can be confident the DIMMs are okay which suggests it's either a CPU or motherboard problem. It's not obvious to me which, unfortunately.

From a time/effort point of view I would want to try a different CPU since it will be easier to install compared to swapping out the motherboard and you could be back in business relatively quickly if it turned out not to be the problem. Do you have a good relationship with the company from which you'd get the CPU? To the point you could explain what's going on and might have to return the CPU if it turned out it wasn't the problem? I've never actually had to return a CPU so I'm not sure what the typical policy would be.

If the old CPU turns out to be okay you could try a different motherboard when it's convenient; or possibly not bother if 64GB quad channel is good enough for what you need to do.

It's certainly possible you got a faulty replacement motherboard but it seems like the CPU has been through quite a bit. I'm surprised at the subtlety of it but if a pin connection on the CPU is dedicated to that slot/channel configuration and can be damaged without having a noticeable effect on anything else it seems reasonable. I've not seen a pin chart for Threadripper but generic CPU charts suggest there are pins dedicated to communication with system memory.

Perhaps @axe0, @BSODHunter, @philc43, @zbook, @ztrucker might have more insight or better suggestions.

Ok understood and thanks cw. I'll wait a bit for some of your comrades to hopefully chime in and start looking at chips.

Did some research last night and apparently tr4 has issues with running 128gb of RAM. I'm going to attempt to manually set the RAM in BIOS and will let ya know.

Sounds good. If true, makes one wonder what the point of the QVL is when it states the memory you're using is good for a 2/4/8 DIMM configuration.