"Thermal Pressure" Kernel Feature Would Help Linux Performance When Running Hot

[phoronix]

Phoronix: "Thermal Pressure" Kernel Feature Would Help Linux Performance When Running Hot

Linaro engineer Thara Gopinath sent out an experimental set of kernel patches today that introduces the concept of "thermal pressure" to the Linux kernel for helping assist Linux performance when the processor cores are running hot...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-Thermal-Pressure-V1

[schmidtbag]

Isn't that pretty much the reason why the Windows scheduler switches tasks between cores? That being said, isn't that why the Windows scheduler performs so much worse with Ryzen? Unless this basically forces a task to switch to another core after exceeding a high temperature (such as 80C), I question if this is actually a good idea.
That being said, I do think it's a great idea, if cores are switched based on a certain temperature.

[tildearrow]

That being said, I do think it's a great idea, if cores are switched based on a certain temperature.

Or if the feature can be disabled at will.

[boxie]

Isn't that pretty much the reason why the Windows scheduler switches tasks between cores? That being said, isn't that why the Windows scheduler performs so much worse with Ryzen?

One would hope that the Windows NT Kernel team would know about the Zen CCX layout and schedule accordingly.

[brrrrttttt]

Isn't that pretty much the reason why the Windows scheduler switches tasks between cores? That being said, isn't that why the Windows scheduler performs so much worse with Ryzen? Unless this basically forces a task to switch to another core after exceeding a high temperature (such as 80C), I question if this is actually a good idea.
That being said, I do think it's a great idea, if cores are switched based on a certain temperature.

Surely the implementation would need to be NUMA aware? That should help Ryzen.

[tuxd3v]

Isn't that pretty much the reason why the Windows scheduler switches tasks between cores? That being said, isn't that why the Windows scheduler performs so much worse with Ryzen? Unless this basically forces a task to switch to another core after exceeding a high temperature (such as 80C), I question if this is actually a good idea.
That being said, I do think it's a great idea, if cores are switched based on a certain temperature.

Maybe they will only switch between most closest cores..or cores of same CCX..

[starshipeleven]

Isn't that pretty much the reason why the Windows scheduler switches tasks between cores? That being said, isn't that why the Windows scheduler performs so much worse with Ryzen? Unless this basically forces a task to switch to another core after exceeding a high temperature (such as 80C), I question if this is actually a good idea.
That being said, I do think it's a great idea, if cores are switched based on a certain temperature.

Afaik the reason WIndows sucks on Ryzen is that it always was very meh on NUMA.