[arch-general] Laptop overheating causes decreased CPU performance until reboot (even after cooling)
Hi, I have a dell 1564 laptop with an i5-M430 cpu, running arch linux. For a long while now I've had issues with very degraded CPU performance if my cpu temperature is very hot till around ~95 degree Celsius. The problem is that the degraded cpu performance stays inspite of cooling down back to 60 degrees, and even multiple suspend-resume cycles. The only way to reset the same is a reboot. I can't seem to track down the cause of this. The cpu frequency etc. look normal, all 4 vitual-cpu's are active from cpuinfo. top reports nearly 5x increased % cpu usage for process compared to what they normally use. Some simple CPU benchmarks (N-Queens) also take approximately 5 times longer. I'm using kernel: 3.14.6-1-ARCH 64bit Can you suggest some ways to track down this issue? or alternatively a way to reset this so that I don't need to reboot to restore performance. The issue is reproducable (I just need to use the CPU at max, for a while to overheat it), so I can provide any logs etc. if it will help to find the cause. I also previously posted with some more details on https://bbs.archlinux.org/viewtopic.php?id=183236 without any response.
On Thu, Sep 04, 2014 at 05:14:27PM +0530, Anish Shankar wrote:
Hi, I have a dell 1564 laptop with an i5-M430 cpu, running arch linux. For a long while now I've had issues with very degraded CPU performance if my cpu temperature is very hot till around ~95 degree Celsius. The problem is that the degraded cpu performance stays inspite of cooling down back to 60 degrees, and even multiple suspend-resume cycles. The only way to reset the same is a reboot. I can't seem to track down the cause of this. The cpu frequency etc. look normal, all 4 vitual-cpu's are active from cpuinfo.
top reports nearly 5x increased % cpu usage for process compared to what they normally use. Some simple CPU benchmarks (N-Queens) also take approximately 5 times longer.
Given that, I would suspect (WAG incoming) that the CPU is scaling back the clock as an emergency measure. 95 degC is WAAAAY to hot for a CPU to be. You need to fix that issue first. If the cooling system is not behaving sensibly, you may need to do some manual overrides: https://wiki.archlinux.org/index.php/fan_speed_control --Sean
On Fri, Sep 5, 2014 at 1:02 AM, Sean Greenslade <sean@seangreenslade.com> wrote:
Given that, I would suspect (WAG incoming) that the CPU is scaling back the clock as an emergency measure. 95 degC is WAAAAY to hot for a CPU to be. You need to fix that issue first. If the cooling system is not behaving sensibly, you may need to do some manual overrides:
https://wiki.archlinux.org/index.php/fan_speed_control
--Sean
The root cause of the heating, is bad physical heatflow design in general (My fans are already on max usually in these cases). I've taken some measures for cooling though occasionally temperature levels still peak. I can understand if the CPU scales down as an emergency measure (and is quite likely the case). My problem is that it doesn't scale back up when the temperature comes back to normal (~60 degC). Infact I've also suspended the laptop, left it for an hour to come down to room temperature and on resuming the scaled down levels persist. Although I'm not able to find a way to check that it is indeed scaled down, just simply the performance drops
The root cause of the heating, is bad physical heatflow design in general (My fans are already on max usually in these cases). I've taken some measures for cooling though occasionally temperature levels still peak.
You may want to look into a laptop cooler.
I can understand if the CPU scales down as an emergency measure (and is quite likely the case). My problem is that it doesn't scale back up when the temperature comes back to normal (~60 degC). Infact I've also suspended the laptop, left it for an hour to come down to room temperature and on resuming the scaled down levels persist. Although I'm not able to find a way to check that it is indeed scaled down, just simply the performance drops
Powertop may be of use, I'm not sure. Again, I suspect that it is a hardware failsafe mechanism, so there's probably nothing you can do about it. Your only option is to prevent the CPU from hitting 95 degC. --Sean
participants (2)
-
Anish Shankar
-
Sean Greenslade