Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

On a Ubuntu server that I am using for computations I have recently observed that some CPU extensive programs (GUROBI,CPLEX) often segfault.

Being in correspondence with tech support of the respective programs I was suggested that it may be a hardware issue.

The administrator of the server performed a detailed memtest and it turned out that the RAM modules appear to be fine.

Hence I've used the tool mprime to test the CPU and the following two lines appear multiple times durring the execution of the stress tests:

[Worker #4 Oct 18 18:47] FATAL ERROR: Rounding was 0.498046875, expected less than 0.4 [Worker #4 Oct 18 18:47] Hardware failure detected, consult stress.txt file.

The stress.txt file in itself is not very verbose about what could be the cause of this error so I would like to ask whether anyone here happens to know what could be the cause of this issue? Is there some other test I could perform to nail the problem even further?

The temperature of the system (and all cores) was fine during the entire stress test (+69.0°C (high = +80.0°C, crit = +98.0°C)) the CPU in question is a Intel Core i7-2600K CPU @ 3.40GHz and is not overclocked or modified in any way.

Also what is interesting that if I run mprime to only stress the CPU all tests pass fine. The error is only triggered when I let mprime stress the CPU+RAM.

share|improve this question
add comment

1 Answer 1

up vote 2 down vote accepted

mprime doesn't give false positives, so you can assume you have a hardware problem. You want to swap the components with known good parts one at a time, and run the same test. I would start with RAM (despite memtest passing), then power supply, then CPU, then motherboard.

If you don't have a spare system, or access to compatible parts, there are a few things you can try, but you may not find a definite cause. First, check that all the fans are spinning freely. Second, check the temperature of the power supply, and the VRMs and PCH on the motherboard. If your server doesn't have probes for these, ground yourself, and touch them with your finger. They might be hot, but not so hot that you can't leave your finger on them. Third, remove all the RAM modules except for the first slot (as labeled on the motherboard), and if the test still fails, swap that one out with one of the ones you removed. If two or more fail, you can safely assume it's not the RAM.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.