How to test CPU functionality in Linux?

I just upgraded my HP uSrv gen8 with a Xeon 1220lv2. I bought this CPU used of course. Arrived from the US of A in a rather crusty state - did a bottle of heat paste burst on it before they packed it? Anyway, after a 100 or so Q tips and a pint of IPA, I exchanged the Celeron for the Xeon and booted into Linux Mint.

All looks good sofar.

However, I’d like to verify that it’s in perfect working order. Running something like Sysbench does not do much, and doing prime number crunching for an hour also does not use all parts of the CPU.

Is there any test for Linux that really verifies the functionality of the CPU, not benchmarks it? Something like The Intel® Processor Diagnostic Tool (Windows only)?

I run Geekbench a couple dozen times for testing under linux personally. Also, does HP have a built in diagnostic tool? Dell has one under F12.

Hi geeklife. Yes, HP has an onboard diagnostic in the ILO, but it’s a general system test, and does not really verify the CPU in its entirety like the Intel tool does. Good for a quick check though.

I’m not proficient enough in Linux and Wine, and with the tool not being in the Wine database, I didn’t want to spend time trying that route. I’ll run geekbench for a bit like you suggest.

In the mean time, after a lot of tinkering I created a Win to Go USB stick and booted into Win10Pro on the server. Slow, very slow. But once you’re booted, installing and running the Intel tool went fine. The CPU looks good after all these tests.

What a pain that Intel does not support Linux properly with their tools. Now, on to my AMD machines… not even sure AMD has a tool like Intel’s :grinning: plenty of searching left to do.

Yeah, you download some projects and build them. Time how long one run takes, just using make’s dry run feature (I know it for bsd, but it might be a different for linux, I never looked myself). You can check

man make

To see info, or look up online.

You could uh, you could even build powerpc packages for me and @stenstorp xD

A processor is so delicate and complex that it either works or it doesen’t. It might overheat if the TIM between the heatspreader and the die is not good anymore (crusty old thermal paste or cracked solder due to heat cycles).
I wouldn’t worry that much about the fact that it was covered in thermal paste.

Phoronix test suite.

I use stress-ng for CPU, and boot to memtest86 for memory.


I do the same thing.

Thanks, I’ll give that suite a try as well!

Second on this. I used an automated routine of memtest86, stress-ng, and gpu_burn for my last hardware rollout.