PPRuNe Forums - View Single Post - Fault tracing with a bootable USB stick
View Single Post
Old 20th Sep 2016, 21:29
  #13 (permalink)  
rans6andrew
 
Join Date: Sep 2004
Location: Berkshire, UK
Posts: 811
Likes: 0
Received 15 Likes on 6 Posts
I built a nice powerful Core i7 machine when the i7 processors were first about. It was brought alive with Ubuntu originally and every thing seemed OK. Then, one day, it just rebooted, just like that. Then it ran fine for a few more days and then, again, it just rebooted. This happened on and off. I checked the obvious stuff, the CPU temperatures never went anywhere the limits, the BIOS memory test ran time after time error free. It never lost any important data or corrupted the OS or BIOS settings. It nearly always crashed when running something audio/visual but that was not really surprising as I often stream music videos off the www when working. This observation suggested an issue with the graphics card so I bought another one and swapped them over. No difference. I then removed each of the three memory SIMs, one at a time and then I swapped out all of the memory with my partner's machine. Mine still crashed after a few days, hers was OK with my memory fitted. I swapped out the hard disk as I had several in swappable drive trays. One of the drives has WinXP, it crashed at about the same frequency as the Ubuntu ones. I bought a second (faster!) CPU which made no difference.

I had originally intended to have a play with overclocking. Whilst looking at the process for tweaking stuff I found that one of the things you do is to tweak settings and see if the machine crashes and if it does back off a bit. Then tweak in the other direction until it crashes again. Then set half way between the two positions and then move on to tweak something else. So I had a tweak of a few settings and found that by underclocking the machine it became stable. It still performed OK but it irked me that it didn't even run at normal speed. I carried on using the machine for over a year and then updated it to run Linux Mint. It ran OK until I tried setting it to run at normal speed and then the crashing started again. Eventually I saw a similar motherboard on Ebay. Got it for a song and swapped everything over onto it. As before, the machine ran OK for a day or so and then spontaneously rebooted. Aaarggghh!

By this point I had swapped out everything except the fans, the case and the PSU. I didn't even suspect the PSU. If you only draw 120 - 150 (measured mains power going into the box) Watts from a 550 Watt PSU you don't even consider that the PSU will cause problems. Anyway, I bought another PSU, swapped it in and the problem has gone away. Pity it took a shade over three years to find it as the warranty on the original PSU had just run out when I did.

Being of an electronics engineering bent I opened up the dodgy PSU for a look. The standard of some of the soldering wasn't brilliant but I couldn't find an actual faulty joint, even when looking with a decent microscope. After salvaging a few nice toroidal cores and a fan I binned the rest of the PSU.

I have now had a play with overclocking. With a bit of patience it is possible to make the machine go way over it's specification speeds. I eventually got the 3.2 GHz processor to run at 4.0 Ghz, apparently reliably but on days when my workshop is hot (it went up to 38 deg C on one day last week) the CPU fan goes up to a higher speed and becomes a bit noisy. I have backed off the settings to a very modest overclock.

With the exception of the PSU I now have a complete set of spare parts for my system. Perhaps I should buy another PSU and a case and build all of the spares into another system?
rans6andrew is offline