Major stability problems with Threadripper 3970x

Discussion created by jmcconn on Apr 28, 2020
Latest reply on Jul 26, 2020 by misterj

Several months ago I built a workstation for code development and running moderately-sized CFD simulations. This system runs Ubuntu 18.04 with kernel version 5.6. When I try to run a simulation or compile code with 64 threads, the system immediately dies. When compiling certain libraries (like boost 1.65.1) the system crashes no matter how many threads I use. Once the system crashes, it won't start back up for at least a day; attempting to start the system results in a failure to POST.

I've had nothing but trouble with this CPU. I've tried 3 different motherboards, 2 PSUs (both of which were in a different system, working flawlessly), and 2 sets of memory that are on the motherboard manufacturer's compatibility list. I also tried various BIOS versions with little success. I even RMA'd the processor, and subsequently waranteed the processor to AMD after the 2nd processor exhibited instabilities. I figured it's extremely unlikely to to get 2 bad CPUs in a row, but not impossible.

The 3rd CPU is also problematic. I was really excited to have such a powerful processor, but so far it's been a $2000 disappointment.