I felt like I had a few moments to spare this weekend, for the first time in quite a while, so I decided to address a long-standing and scary problem with my main desktop PC.
Understand that I'm still running the same basic build I outlined in my Damagebox 2011 write-up. I've added an SSD, more RAM, and a newer graphics card, but that's about it. So it's old-ish. I'm surrounded by amazing computers, but I don't like to tinker with my primary system. Downtime on it kills productivity. Plus, I usually game on one of my GPU test rigs.
Anyhow, the ol' Damagebox developed a problem at some point in the past few months. Actually, it's never been great about keeping itself cool. I talked about the problems with cooling the ICH chip in my initial build article. More recently, though, I'd noticed CPU temperatures above what one would like, even at idle. When I'd tried to do strenuous things, like transcoding video in Handbrake, the system had even rebooted a few times. A quick check several weeks ago confirmed the problem: the CPU reached temperatures approaching 90C in a load test using Prime95. So I knew this box needed some attention, but I just kept putting it off since I was busy.
This weekend, I finally pulled the system out from next to my desk and took a look. I was kind of hoping that reapplying thermal paste to the CPU cooler would solve the problem. When I pulled off the cooler and inspected the contact plate, though, the thermal paste situation was textbook. Nothing had dried up or flaked off. Hmm.
For science, I went ahead and reapplied paste. Then I booted up the system and ran another Prime95 test under close observation. Within seconds of starting it up, my CPU temperatures shot up to ~88C and kept climbing. Yikes. When I turned off Prime95, temperatures only dropped slowly. The cooler was totally ineffective, in a scary way.
My best guess is that this Corsair H60 either 1) lost all of its coolant over time or 2) quietly had a pump failure. I swear I could hear the pump going after booting up, though. I need to pull apart the plumbing on it and see if anything comes out. Betcha it's dry.
Realize this is the system that I've used to host four to five incoming video feeds, produce a composite image in XSplit, and them stream to Twitch for the TR Podcast's video version. All of that was happening on a box with pretty much no CPU cooling to speak of.
Anyhow, I ripped out the H60 and installed an old X58-compatible tower cooler from Thermalright with a 120-mm fan. That was shockingly more effective. CPU temps at idle were suddenly in the high 30s, rather than 50-60C, and under load, temps peaked at about 67C. Good grief.
After I'd reassembled everything and plugged it all back in next to my desk, though, I realized the repaired system was much too loud. The problem, turns out, was that the Thermalright cooler's fan was a three-pin DC type. My lousy motherboard wouldn't control its speed at all, so it was at a constant 1576 RPM. I pulled out the system again, swapped in a four-pin PWM fan, and soon achieved fine-grained linear fan control with similar temperatures and minimal noise.
So yeah, my really old computer now works perfectly again. I'm still shaking my head thinking about how broken it was and for how long. Thank goodness for Intel's CPU thermal throttling, I suppose.
I'm thinking I should upgrade, although honestly, I dunno. This thing has six Westmere cores and 12 threads at up to 3.6GHz with triple channels of DDR3, 12GB of RAM, a GTX 680, and a 256GB boot SSD. And I'm apparently not pushing it too hard, or it would have gone into nuclear meltdown by now. Maybe I'll just—finally—build those new GPU rigs instead.