In my office at work I have a server computer that has been crashing more and more lately. I very quickly ruled out the possibility of a software problem, if Windows were crashing I would get a blue screen of death (BSOD) or a mini-dump file. Instead the machine would just hard reboot at random intervals. The BIOS log contained nothing, so I was left to figure this out the hard way.
First I tried replacing the power supply, thinking maybe the output power would occasionally "dip" causing the crash. But it continued to crash. I noticed the crashes seemed to occur most often during heavy disk activity. So I suspected either the RAM or the disk controller. Since the disk controller was onboard the motherboard, I checked the RAM next. Reseating the RAM didn't help, so I ordered new replacement RAM (DDR3 is cheap right now). The new RAM appeared to fix the problem, it hasn't crashed since then.
But while testing the new RAM I noticed something else disturbing. Using HWMonitor I noticed the CPU (Core i5) was running hot! The idle temp was 50 degree Celsius and under load it got as hot as 104 degree Celsius. That's not good, you shouldn't be able to boil water on your CPU. When I built this computer I used the stock Intel cooler, but I've always hated the stock Intel cooler. Not because of performance or noise, but because they seem to be difficult to install. And the instructions don't help, they only contain pictures and no text.
I turned the machine off and removed the stock cooler. I discovered two faults when I installed the original cooler, so no wonder it was running hot. I cleaned off the CPU and cooler, applied some Arctic Silver 5 (consistently rated as the best thermal compound), and reinstalled the cooler. Now the CPU is idling at 30 degree Celsius and under load it goes no higher than 50 degree Celsius. Much better than before!
I'm sure between the bad RAM and the overheating CPU that I've solved this machines instability problem.
No comments:
Post a Comment