I have an Ideapad S145-15IWL, it comes with an I5-8265U CPU plus MX110 discrete onboard GPU, and I'm curious if this is just normal for laptop CPUs or if it's wrong:
I was trying to run a heavy simulation on python3 (in Linux Mint Tricia 19.3, on all threads with maximum priority), and I noted that as soon as it started the CPU frequency fell from 3.9 GHz to 2.1 GHz, while on performance governor AND manually set to 3.9 GHz, so it's not ondemand. It kept a 65 ºC temperature, which seemed tolerable, because that's the frequency and temperature it keeps when running heavy applications and games.
The issue here is how fast it was happening, because other applications, like GTA V, take several seconds until it reaches this point, but this simulation was making this nearly instantaneously.
I then investigated closely and saw that, whenever I started the simulation, the CPU temperature would rise to 93 ºC in a matter 1 to 2 seconds only, then the frequency scaling governor would immediately throttle not to burn the CPU. It went from 45 ºC to 93ºC and then to 65 º C so quick I wasn't seeing it happen.
I thought this was wrong, so I changed the thermal paste. In the first try I ended up putting a way too thin and smooth layer in both sides (the components and the cooler), and only worsened everything, making the frequency jump like crazy between 400 MHz and 2.4 GHz, and the temperature would also oscillate so fast the sensor latency wasn't enough to follow it.
I disassembled and made a similarly thin layer on each side again, but this time now using my fingers to create a rougher surface (with my fingerprint) and made sure to wobble the cooler laterally a bit after putting it on place to even it out before tightening the screws.
It basically came back to how it was before, just a bit better, now it spends more time at 90 ºC (some 4 seconds instead of just 1) before the frequency scaling throttles and it drops to 2.2 GHz, not 2.1 GHz anymore, and the temperature is now sitting at 61 ºC, not 65 ºC.
The thermal paste I'm using is Arctic MX-4.
I'm starting to feel this is just how it is, even why running at 2.2 GHz is not so low, I think that a really faulty thermal contact wouldn't manage to dissipate enough heat.
Maybe it just has a negligible thermal inertia because it has no integrated heat spreader? And perhaps the thermal design is a bit limited for this kind of activity?
Thanks!
I was trying to run a heavy simulation on python3 (in Linux Mint Tricia 19.3, on all threads with maximum priority), and I noted that as soon as it started the CPU frequency fell from 3.9 GHz to 2.1 GHz, while on performance governor AND manually set to 3.9 GHz, so it's not ondemand. It kept a 65 ºC temperature, which seemed tolerable, because that's the frequency and temperature it keeps when running heavy applications and games.
The issue here is how fast it was happening, because other applications, like GTA V, take several seconds until it reaches this point, but this simulation was making this nearly instantaneously.
I then investigated closely and saw that, whenever I started the simulation, the CPU temperature would rise to 93 ºC in a matter 1 to 2 seconds only, then the frequency scaling governor would immediately throttle not to burn the CPU. It went from 45 ºC to 93ºC and then to 65 º C so quick I wasn't seeing it happen.
I thought this was wrong, so I changed the thermal paste. In the first try I ended up putting a way too thin and smooth layer in both sides (the components and the cooler), and only worsened everything, making the frequency jump like crazy between 400 MHz and 2.4 GHz, and the temperature would also oscillate so fast the sensor latency wasn't enough to follow it.
I disassembled and made a similarly thin layer on each side again, but this time now using my fingers to create a rougher surface (with my fingerprint) and made sure to wobble the cooler laterally a bit after putting it on place to even it out before tightening the screws.
It basically came back to how it was before, just a bit better, now it spends more time at 90 ºC (some 4 seconds instead of just 1) before the frequency scaling throttles and it drops to 2.2 GHz, not 2.1 GHz anymore, and the temperature is now sitting at 61 ºC, not 65 ºC.
The thermal paste I'm using is Arctic MX-4.
I'm starting to feel this is just how it is, even why running at 2.2 GHz is not so low, I think that a really faulty thermal contact wouldn't manage to dissipate enough heat.
Maybe it just has a negligible thermal inertia because it has no integrated heat spreader? And perhaps the thermal design is a bit limited for this kind of activity?
Thanks!
Facebook
Twitter
Instagram