Like (I guess) most people, I end up doing tech support for a random set of friends and neighbours. One of them was complaining of problems with her computer locking up under load. After ruling out the usual suspects (an underpowered PSU being the main one), I realised the problem was the CPU overheating. She has a 2GHz P4 with a TR2TT M10 heatsink, which (when the machine was put together for her) was supposed to be silent (I have my own opinions on that, but that's beside the point).
Anyway, at 100% load the temperature shot up to > 80 degrees C within about 15 seconds, after which the machine locked up. Even at idle, the CPU was at > 40 degrees C. The best comparison I have is a 1.7GHz Celeron with the stock Intel heatsink, which idles at about 35 degrees C and never goes above about 50 degrees at 100% load (from Intel's TDP figures, the 1.5V 2GHz P4 and 1.75V 1.7GHz Celeron have about the same thermal characteristics).
The problem seems to be that the TR2 heatsink is totally incapable of removing any kind of heat from the CPU. Trying the stock Intel HSF in exactly the same setup produced the expected performance, using the TR2TT resulted in thermal runaway in the CPU. (Just for the record, I checked the usual suspects, the fan is working, the TIM is present, etc etc, so everything's working as it should).
I think I have a spare stock HSF somewhere, but the owner is very reluctant to use it because of the noise, and is also reluctant to buy a new HSF when the existing one should be doing the job. OTOH I've never encountered a situation like this before, I've found some HSF's that are a bit worse (or better) than others, but never one that barely seems to cool at all. Any ideas?
CPU overheating with TR2TT M10 heatsink
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- Posts: 333
- Joined: Wed Apr 05, 2006 9:54 pm
- Location: USA
-
- Posts: 2049
- Joined: Thu Dec 15, 2005 11:06 am
- Location: Klamath Falls, OR
-
- Posts: 39
- Joined: Tue Jan 20, 2004 2:48 am
Ahh, that was the magic clue, thanks! The giveaway wasn't so much the fact that heat wasn't being removed, it was the *rate* at which it wasn't being removed. The problem is that the motherboard (ECS, aka PCChips, ugh) has the CPU considerably offset to one side of the rectangle of the retention bracket. When I pulled out the TT2, I'd cleaned off the TIM to see whether there was a problem with the base of the heatsink, and to replace the heatsink grease in case that was the cause of the problem. Then when I dropped in the Celeron and stock HSF, I put grease on the CPU's heat spreader and dropped the heatsink on top of it.ultraboy wrote:Another possible area is HSF retention clip is losing its pressure. 40+ celcius different between idle and load in a very short time, in my opinion, indicates poor contact between HSF and heatspreader.
In both cases this hid the problem, which was that the CPU was sufficiently far offset to the side of the heatsink that the original TIM probably barely made any contact (the layer was also thick enough that in the area where there was some present it lifted the rest of the heatsink base away from the CPU). Replacing the original CPU and TT2 heatsink, this time with only a thin layer of grease applied to the CPU heat spreader, lead to a proper thermal contact (way off to one side of the heatsink, in fact right on the edge of the copper base plate). It's now idling at 30 degrees C and only goes to 38 degrees under load, which is a great relief.
(I thought the P4's clock-throttled when they got too hot? I have mainly AMD stuff here so I'm not too familiar with them, but I do remember the video Tom's Hardware had of running an Athlon and a P4 without a heatsink).
-
- Posts: 39
- Joined: Tue Jan 20, 2004 2:48 am
Yup, that was probably not too far from the truth before I inadvertently destroyed the evidence when I cleaned the TIM from the TR2 :-). In any case it's fixed now, I'm just going to give it a burn-in for awhile. Thanks for the help everyone!Felger Carbon wrote:In fact, considering the extraordinarily rapid temperature rise you're seeing, the TR2 must not be making contact with the CPU at all! :oops: