Wednesday, October 29, 2008

47 C

Computers get surprisingly hot.

For those of you that read my original post regarding Wagner, you'll notice that I said the CPU ran about 75 C under full load. Most reviews for my CPU say it should run at about 50 C under full load. So, yeah. It was running 50% hotter than it should, and in the CPU world, that's a huge difference.

So, yesterday I decided to figure out why it was running so hot. I changed the clock speed of my CPU (was at 3.6GHz, downclocked to stock at 3.0GHz), reset the CMOS to default, and checked the temps again: slightly cooler at 72C, but still nowhere near the 50C all reviews say it should run at (under load). I decided it must be a hardware issue. I shut everything down and take off the side panel; everything seems to be in working order. Still, maybe the thermal compound on the CPU didn't spread evenly or something, so I took off the heatsink to check.

Only about 80% of the heatsink thermal compound was touching the CPU, the rest wasn't. This could account for the temperature issues. I checked the heatsink and noticed one of the pegs was smashed against the motherboard

Thankfully, not my computer.

and not attached properly; a pair of pliers fixed that. After cleaning and then reapplying the thermal compound (probably more than I needed, but meh), I reattached the heatsink and booted up the PC. It was running at much better temps now, with the CMOS reporting about 40C. Mind you, that's prior to any real activity. So, I loaded up Windows and decided to see how it would perform under Super Pi. Within about a minute of monitoring the core speed I noticed something: it was climbing - and fast. I watched it go from 60 C at idle to 65, 70, 75, 80, 85 C - this freaked me out since that's redlining. It didn't show any sign of stopping so I immediately shut everything down as fast as I could. Looking inside the case, I found the problem...I forgot to attach the CPU fan. It was a rookie mistake and it almost cost me dearly. Still, I caught it in time.

Five seconds later it was hooked up properly and running again. this time reporting temps around 53 C idle. Better than the 60 C, but not great still. I decided that'd be enough tinkering for the night, talked to some friends for a bit, then went to bed.

This morning I powered on my PC and wanted to see if it would be stable. The past few days I've let it run while at work it's either locked up or crashed on me. I assumed this was a fluke of Vista and didn't think much about it. After talking to Will, he mentioned he's never had Vista crash on him. Maybe it was a hardware issue then? When I came back from work it was still on just as I left it: huzzah! I'm thinking the crashes were related to heat issues and perhaps an unstable overclock. I decided to check for sure and downloaded

I love prime numbers.

Prime 95, a great program for stress testing your CPU and RAM. I loaded it up and set it to go while my CPU was set to 3.6GHz; within a minute one of them had failed. I decided to try again; this time, the other processor crashed. Ok, that's not a good sign. I could up the CPU voltage to stabilize it, but upping the voltage raises the temp of the CPU significantly, and I don't want to run it any hotter than it is. I took it down to stock settings and tested it: it was stable for over 2 hours before I ended the program manually. Then I clocked it up to 3.3: stable for over 30 minutes before I ended it. Right now it's at 3.4 and hasn't had any issues. I might try going to 3.5 tomorrow, but I'm not sure it's worth it. 3.4GHz from a 3GHz CPU isn't bad, and as of this writing, it's idling at 47 C; as the thermal compound sets, it works better. Still not what it should be, but nothing to worry about, either.

Tomorrow I might try to go to 3.5GHz, but I'm not sure it's worth it. At 3.4 everything is working well so far; I'm gonna stress test it while I'm at work tomorrow to see if it holds up over a long period of time and make sure, but I'm confident. If it does, I'll try 3.5 the next day.

Best. Series. Ever.

[I'm also happy to note that this is probably why Fallout 3 kept crashing on me; when the CPU pukes, it takes whatever it's running with it. I didn't notice this because I'm not used to multiple core CPUs. I played it for about two hours today with absolutely no crashes or flaws at all.]


bigmark74 said...

I still think your computer was just suffering from some abandonment issues each time you went to work. But now that it's had some time to get used to being alone it's doing better.

--jeff * said...

i'm genuinely impressed by your computer knowledge, especially since the vast majority of it is self-taught.