Intel Linux Power Regression Still Being Worked On
A few days back I wrote about an Intel Linux developer finding a 50 Watt power regression within the kernel since the 3.10 release. At first it looked like it might be easily solved, but the issue is still being investigated and Intel hardware engineers are now getting involved in the matter.
It looks like the Intel power regression on the Linux 3.10+ (through at least Linux 3.13 Git right now) might be more widespread than originally thought and for large, multi-socket systems could come down to large power consumption differences in not being allowed to fully idle. Len Brown wrote on Wednesday in a new mailing list post that he's now checking with the Intel hardware engineers about the matter.
The kernel developers do appear to have an understanding for the issue so hopefully the power issues will be solved in time for the Linux 3.13 kernel final release in the weeks ahead. New Linux power benchmarks are still forthcoming on Phoronix; hit up the mailing list thread if you want all the latest on the technical discussion surrounding this Intel CPU power issue on Linux.
It looks like the Intel power regression on the Linux 3.10+ (through at least Linux 3.13 Git right now) might be more widespread than originally thought and for large, multi-socket systems could come down to large power consumption differences in not being allowed to fully idle. Len Brown wrote on Wednesday in a new mailing list post that he's now checking with the Intel hardware engineers about the matter.
The kernel developers do appear to have an understanding for the issue so hopefully the power issues will be solved in time for the Linux 3.13 kernel final release in the weeks ahead. New Linux power benchmarks are still forthcoming on Phoronix; hit up the mailing list thread if you want all the latest on the technical discussion surrounding this Intel CPU power issue on Linux.
12 Comments