05-13-2012 04:18 AM - edited 05-13-2012 04:20 AM
I was running Prime95 to check my CPU (Core i5-2520M) temperature (using Real Temp) and I noticed it was a bit high, about 95 degrees celsius. I followed up by running FurMark to check how the NVS 4200 GPU was doing. It ended up sustaining at 87 degrees celsius. I did not run the two tests concurrently.
I decided to strip the heatsink off and apply new thermal compound. After the new compound was applied, the CPU temperature under Prime 95 peaked at 84 degrees celsius. Under Furmark, the GPU sustained at 75 degrees celsius. A very significant drop.
I decided to run the two tests concurrently (after new compound had been applied) and ended up with a CPU temperature of 100 degrees celsius and a GPU temperature of 97 degrees celsius. Pretty horrifying stuff. I can't imagine those values being within operating range.
These torture tests are rather extreme so I followed up with some gaming. I played DOTA 2 for about 3 hours and during that period my CPU peaked at 98 degrees celsius with an average temperature of 93 degrees celsius. I did not log the GPU temperature during this session.
I should mention that Turbo Boost is enabled and the fan was set at maximum speed.
Does anybody have any temperature figures for similair configurations? Specifically in situations where the GPU and CPU are both under high load. I'm curious to know if my temperatures are abnormal.
05-13-2012 08:57 AM
05-13-2012 12:33 PM
05-13-2012 09:30 PM
The system should have shut off at 100c, that is the temp thermal protection kicks in at.
http://ark.intel.com/products/52229/Intel-Core-i5-

05-14-2012 08:37 AM
I havn't seen a rash of heat issue with the T420 like I have with the T410
After applying a fresh coat of thermal grease our T410 untis will usually max around 70-75 just running the Lenovo stress test.
However on a T420 those figures do bump up to 80 - 85
And yes the unit should shut off at 100 degrees but it takes a few beats at that temperature before it trips. You may be hovering in and out of the 100s
05-16-2012 11:27 PM - edited 05-16-2012 11:28 PM
I started to think that it's not an issue with my sample. I went ahead and confirmed these temperatures on two other identically specced T420s (4178). Under extreme conditions the heatsink/fan unit cannot dissipate heat quickly enough causing both chips (CPU and GPU) to reach dangerous temperatures. If you haven't torture tested the GPU and CPU in your T420 concurrently, you might want to give it a shot. It gets alarmingly hot.
(Do this at your own risk, extreme heat can damage your hardware)
Prime 95: http://www.mersenne.org/freesoft/ (CPU Test)
Furmark: http://www.ozone3d.net/benchmarks/fur/ (GPU Test)
Real Temp: http://www.techpowerup.com/realtemp/ (CPU Temperature)
GPU-Z: http://www.techpowerup.com/downloads/SysInfo/GPU-Z
Tests were done while plugged in using the default Power Source Optimized profile.
Prime95 test was run using In-place FFTs. Furmark was run at 800x600 with anti-aliasing off. The tests were run for 20 minutes concurrently.
05-17-2012 03:49 PM - edited 05-17-2012 03:51 PM
People bad-mouth the factory thermal compound, but all the objective testing done by the people I refer to as the "overclocking crazies" indicate that the compound used on new heatsinks is at least as good as the arctic silver that most people (me included) use. Also, the little dots on new heatsinks provide better coverage than what is obtained when you reseat the heatsink yourself. When tests are run with a clear material pressed onto the CPU die, you can easily see the uncovered areas that are created with the common "blob" approaches (grain-of-rice, pea-sized-blob, thin-line), and a thin even layer, generates tiny air pockets that act like excellent insulation.