Hello all,
I believe the processes are distributed and switched among the cpu's according to the work load. Are thermal readings affects these decisions, such as temperature or even temperature sensitivity (=the change in temperature per unit load) ?

In my computer temperature sensitivity of cores seems very different. I am looking for a way to use this info to make the machine quieter. See longer version below.

Thanks.
ilker.

Longer version : I observed something interesting in my new laptop. Annoyed by the fan noise on and off at night I started to watch the temps. Temperature of the one of the 4 cores is much more sensitive to the load. Same process changes its temperature much faster and higher in magnitude compared to other cores. For example, cpuburn with same# of threads can make a difference up to 16C in readings and that increases the fan noise. I have no idea if this is a sensor or cpu issue.
Actually I wouldn't mind whether fan is at level 2 or 3 when cpu is being abused. But when I do daily routine tasks, this core is the reason that fan kicks on/off constantly which is annoying (due to high sensitivity). As a workaround when I make this bad core offline I am left with 3 cores and quiter fan. Maybe there is a more clever way to do that ?? Any comments appreciated.

Also a side question,
Do you guys see significant variation in the temperature of the cores like me ?

Check below measurements with all cores online. 1st is taken while "cpuburn n=1". In this case 1-thread is dancing over all the cpu's randomly. But the temperature effect is much different.
2nd example is with "cpuburn n=8" so I assume it burns all cores equally.

("cpuburn n=1" is running)
$sensors
coretemp-isa-0000
Adapter: ISA adapter
Physical id 0: +52.0°C (high = +100.0°C, crit = +100.0°C)
Core 0: +39.0°C (high = +100.0°C, crit = +100.0°C)
Core 1: +52.0°C (high = +100.0°C, crit = +100.0°C)
Core 2: +34.0°C (high = +100.0°C, crit = +100.0°C)
Core 3: +39.0°C (high = +100.0°C, crit = +100.0°C)


("cpuburn n=8" is running)
$sensors
coretemp-isa-0000
Adapter: ISA adapter
Physical id 0: +71.0°C (high = +100.0°C, crit = +100.0°C)
Core 0: +56.0°C (high = +100.0°C, crit = +100.0°C)
Core 1: +71.0°C (high = +100.0°C, crit = +100.0°C)
Core 2: +51.0°C (high = +100.0°C, crit = +100.0°C)
Core 3: +61.0°C (high = +100.0°C, crit = +100.0°C)

Try to edit your posts instead of adding a second one. Makes it look like someone is helping you.

I never consider that cpu loads or temps must be equal. Too much going on. You could look at processes running maybe to decide. There might be a way to change how jobs get scheduled but that too may affect things worse. Some processes may still not be fully smp and the way the current scheduler works there may be reasons a cpu is running hotter.

There have been some issues with how temps are reported to on some systems.


I just set the bios and if temps get too high then let bios take action.

Might see if you could set cpu to a lower speed or other hacks to cpu like power. Some can be set in bios and I am pretty sure some linux ways to adjust. Usually laptops won't offer full control.

Could look at reasons for such load and try to offload if possible. Nic might could be set to offload checksum. Looking at how disk use like swap maybe would affect performance. Maybe some other ways.

Thanks for the reply. I don't have anything in bios also I already run in lower frequency settings. Even when I make the core offline in idle it still seems 7-8 degrees hotter. I don't know anything else to do other than turning off that core. (which helped a little)

However this issue made me think on how complicated it can get for an OS to select "the best" core/cpu for a given task if cores react differently under same load. An average laptop has 1 cpu with 4 cores, 1 gpu, 2 total fans one on each. Each fan is different size connects to both chips via heat sinks etc...