MATLAB crashes in one computer but not in the other running the same code

I am running the same code in MATLAB 2020 in two computers (what I do is basically matrix diagonalization with eigs). One of them is a desktop using a processor with 16 cores (24 threads) and 64 GB RAM, and the other is a laptop using a processor with 4 cores (8 threads) and 32 GB RAM. Interestingly MATLAB crashes in the desktop but runs flawlessly in the laptop (MATLAB is using around 1 GB RAM in both computers). Looking at the task manager, I realized that almost 100% of CPU is used when MATLAB is running in the desktop. So, I limited the maximum # of threads to 6 (which was by default equal to the # of physical cores, which in this case is 16); this helped reduce the CPU usage to ~%55 but did not solve the crashing issue. How can I fix this problem?

[Walter Roberson]

Would it just happen to be the case that the laptop is using an Intel processor, but the desktop is using an AMD processor?

[Onur]

Thanks for the answer, but both are using an Intel processor.

[Walter Roberson]

Is it possible that the desktop is using an Intel Gen 13 or Gen 14 processor, but the laptop is using a different generation?

Answer 1

[埃博拉酱]

Can you describe the details of your crash? Is it just your code crashing or is the entire MATLAB process crashing? Is the operating system Windows or Linux? Are you running pure MATLAB code or do you include C++ code that you wrote?

[Onur]

The entire process crashes, both systems are Windows. I am running a pure MATLAB code. Perhaps I should try reinstalling MATLAB.

[Torsten]

Yes, and if this doesn't help, you should contact Technical Support:

https://uk.mathworks.com/support/contact_us.html

[Thai Phuong]

I got the same problem, the codes worked perfectly before. And it doesnt crash on other computers, just mine. I've monitored Matlab memory use, there are some peak at the crash moment, but not systematicly. Matlab reinstallation does not help. I am struggling to debug as it is not code problem, but something else.