Thinkpads Forum

theterminator93 wrote:A bit off-topic but when I upgraded from the i5-2540m (2.6 GHz/3.3 GHz turbo) to the i7-2640m (2.8 GHz/3.5 GHz turbo) I couldn't find my tube of Arctic Silver - so I used whatever I had that I could find - the "cheap white stuff" you're never supposed to use on anything you care about. I finally ordered new Arctic Silver and decided to do a temperature comparison before/after... with 4 Prime95 FFT/maximum heat threads running before AS5, TPFanControl instantly reported the highest figure it could on this system, which means temps exceeded this value (96°C (!!!)). With AS5 applied, temps are now 88°-89°C. Still a good 10° hotter than I'd like, but with at least a 7-8°C drop (probably closer to 10°-12° in reality) at least it's not as absurd as before. This is with the nVidia GPU shut down. Just goes to prove what we all already know - don't cheap out on thermal paste!

On the other hand... If I load up the NVS 4200 in Furmark with Prime95 still running, CPU temps again go above the 96°C max shown by TPFC and the GPU tops out at 82°. After terminating Furmark, it took about 60 seconds for the temps to start showing less than 96, eventually (after a several minutes) dropping back down under 90. My guess is CPU core temps during this got close to the maximum allowed temperature of 100. I can't imagine what temps would do with an extra 10W TDP thrown at the heatsink, even with good thermal paste. Thermal shutdowns at full load, my guess.

No Prime95 running, Furmark running, GPU climbs to 81° and a single full load thread on the CPU brings it to 88°.

This is all in a room with ambient temperatures of about 28°C (84°F).

theterminator93 wrote:I decided to replace the thermal paste with Arctic Silver 5 this evening, and while I was at it I decided to do some temperature measurements before/after the reapplication. The results speak for themselves. The heat sink and fan was already cleaned before either of these trials, so the only variable was the thermal paste.

Conditions:
CPU: Core i7-2860QM @ 2.5 GHz (3.6 GHz Turbo)
GPU: nVidia Quadro 2000M
Ambient temperature: 19-20C

Multi-threaded Prime95 CPU and FurMark GPU torture

Before AS5 application, I loaded 8 threads of Prime95 for small FFT testing (maximum heat). After a short time the CPU temperatures stabilized at 80C, with CPU-Z reporting the clock at 2.5 GHz, occasionally jumping to 2.6 GHz. TPFC fan speed was 64 (maximum) at roughly 4700 RPM. To load the GPU after this, I loaded FurMark. After five minutes the temperatures stabilized with the GPU at 89C (or 75C as reported by FurMark) and the CPU up at 96C, although the CPU had throttled back to 2.4 GHz, occasionally dropping to 2.3 GHz.

After the AS5 application, I repeated the same series of tests. 8 Prime95 threads for small FFT tests had the CPU stabilize at 72C - with the fan running at level 7 (3800 RPM) AND the clock rate was 2.6 GHz, occasionally dropping to 2.5 GHz. Five minutes after loading FurMark, the GPU ramped up to and stabilized at 91C (or 77C as reported by FurMark), although the CPU only reached a peak of 87 at a steady 2.5 GHz. Quite an improvement (at least in CPU temps)!!!

Single-threaded FurMark GPU torture

Five minutes after unloading the CPU of the Prime95 torture (only FurMark running), the CPU was mostly 3.3 GHz, occasionally jumping to 3.5 or peaking at 3.6, regardless of pre or post application. Temps before application were CPU - 81C, GPU - 71, after application CPU - 73C and GPU - 72.

Conclusions

Overall, the GPU temperatures didn't change enough to warrant any claim that the AS5 paste reapplication improved or worsened its temperatures. The reapplication of AS5 to the CPU, however, showed impressive results in both temperature and throttling. The multi-threaded tests ran 8C cooler while at a lower fan speed setting AND a higher clock frequency. Multi-threaded tests with a loaded GPU allowed the CPU to run 9C cooler without throttling back to less than full clock speed.