1

I have a build with a obvious hardware error, and I am stumped on how to debug it.

system:

  • i5-4690k
  • z97m-g43
  • 2+1GB avexir DDR3-1300
  • crucial m550 120GB SSD
  • corsair CX500M
  • kubuntu 14.04 (64bit)

symptoms:

  • ubiquity (the ubuntu installer) claims that the installation of GRUB failed, when it aparently did not.
  • after watching videos on youtube, the computer began completely freezing, but was able to recover. Following a couple of system updates, the system was rebooted, which led to the kubuntu consistently getting stuck on login
  • live CD works fine if you don't do intensive stuff, but locks up completely if you do.

things I did after randomly googling around:

  • badblocks -v /dev/sda reports no errors
  • memtest86+ refuses to boot

what are some general strategies which can be used to debug situations like these?

EDIT: derp, I was using 2GB + 1GB sticks. Now I have less symptoms and the same problems.

4
  • @Ramhound yes, boots fine. In the process I discovered I had been using 2GB+1GB RAM, but the crash under load problem persists.
    – Azsgy
    Jan 14, 2015 at 23:21
  • Is the system overclocked? What happens when you push it back to base settings?
    – Journeyman Geek
    Jan 14, 2015 at 23:33
  • @JourneymanGeek The first thing I did was reset my motherboard. It is at stock settings
    – Azsgy
    Jan 14, 2015 at 23:44
  • @Ramhound It appears then that my previous edit was not clear enough. I havr now edited the question to reflect the current state
    – Azsgy
    Jan 15, 2015 at 7:33

1 Answer 1

1

Test and Measure.

With no specific errors other than system instability, sometime testing and measuring is the most effective way to drive out a hardware issue.

First you need to be able to reproduce the problem on demand. If you cannot reproduce the problem, then you will have a difficult time with the measuring part. Systems that crash every 1 hour to 6 days are the worst!

Change the hardware in ONE way, then try to reproduce the problem.

First remove ALL unecessary peripherals. Start with Motherboard, CPU, Memory (One stick), Power Supply, Keyboard, Video, and Monitor. Test with each memory stick separately. Swap video cards. Add a drive and load with only one drive. Swap any parts possible. Try a different OS install like windows? Check for cooling problems. Make sure all drive connections are good. Swap cables.

The hardest are CPU and Motherboard since most people do not have these on hand and they can give extremely odd and elusive errors. I've had times when I could not even reproduce the errors on demand very well which made things near impossible to measure.

Point is a motherboard and CPU can be the problem and your symptoms are indicative of something to do with CPU/motherboard. If it is not hardware, maybe a BIOS setting? Definitely low level.

I've had write errors on hard drives that were power supply related; lock up errors that were CPU related; Memory errors and drive errors that were cooling (motherboard chipset) related. All of these solved through making a change and testing. Sometimes that required buying a part to test and measure; had to buy the CPU and Power Supply on faith.

Just a note, be careful with tests; like I said, I had a memtest throw errors on the memory. Then I had seatools throw errors on the drive. I assumed that not ALL of those components were failing at once and found that the heatsink on the chipset was sensitive to dust build up and air flow. So just because you have memory errors, it is not necessarily bad memory. The same goes for other tests.

In the end when all logs and errors run out of indication, test and measure. Good luck.

3
  • things I have eliminated: disk (badblock returns zero errors, motherboard was reset same errors occur with a Windows trial. That leaves CPU, PSU and MOBO, which I can't test and RAM, which will be tested this afternoon. In case the RAM does not fix it, how should I go about RMA? I don't see how I can differentiate e.g. between a error in the memory part of the MOBO vs. the memory part of the CPU vs. a PSU problem. (assuming the problem is something with the memory)
    – Azsgy
    Jan 15, 2015 at 14:01
  • Sometimes you cannot definitively pin point the error by pure inference without alot of intuition. I had weird problems and even asked for help superuser.com/questions/709058/ecc-errors-no-mem-test-errors and received none. In the end it was a combination of CPU and BIOS version. But it took me weeks to even have enough information to ask a question. But my intuition allowed me to push the computer in a way that produced an error I could work with where I had no tell tale errors up to the point. And that was with buying both a CPU and motherboard as part of the diagnosis process.
    – Damon
    Jan 16, 2015 at 2:28
  • If you can RMA, then at least try. It is highly unlikely you are having a software problem based on information in your question. And if it is hardware and you can RMA then let them deal with it. I never have hardware new enough to RMA and am left stuck diagnosing and repairing.
    – Damon
    Jan 16, 2015 at 2:30

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .