Hard Crash in Windows 10 when calculating eigenvalues of matrix

Put simply, I have reasonably large sparse matrix coming from a finite element simulation. When trying to call Eigensystem[] or Eigenvalues[] on this matrix, my Windows 10 machine consistently crashes in the next few seconds. Has anyone experienced this issue?

Attached is a simple notebook containing the code that blue-screens my machine.

Mathematica 12.0

Using Windows 10 Education 1803

Intel 7900X CPU with 32 GB of RAM

RunAtYourOwnRisk.nb (7.39 MB)

Hi Sam, it worked for me. When I added the Normal[] command worked without warnings, because the following message does not occur (without using Normal[]):

Using Normal[] worked without warnings:

Try this and see if it works for you.

Hi Claudio,

It still crashes with Normal@B on this machine too. To make matters more confusing: I have a laptop which is also running Windows 10 and the same version of Mathematica, and it does not crash when running Eigenvalues[B] directly. On the laptop, I get same warning that the matrix is being made dense, but it still completes.

Very strange … I could not reproduce this flaw here. You need to check if anyone else in the community knows what is going on. I don’t know if this is the correct procedure but you may need to ask this question at Wolfram Technical Service (stating the type of blue screen error - maybe taking a picture and attaching it to the message?!). Honestly, I don’t know why this is happening.

Works for me as well i7-9750H 2.6Ghz 16GB, home edition

The entire OS crashing may be a sign of hardware instability so the actual Eigenvalues computation might only matter as far as pushing the hardware by doing intensive computation on multiple threads.

My recommendation would be to run a hardware stress test on this particular machine (Prime95 has a built-in utility for that, or there are also some more specialized programs available).

Thanks for the suggestion, Ilian. Prime95 didn’t crash my machine, but resetting everything in the BIOS to defaults does seem to have resolved the issue!

Thanks everyone for helping me figure this out.