Intel P-State Driver Might Still Be Causing Wonky Linux Performance
Last year on earlier versions of the Linux kernel we've uncovered performance issues around Intel's P-State / CPUfreq driver that was causing adverse performance in certain configurations. While it looked like most of the Intel CPU clocking issues were resolved in more recent versions of the Linux kernel, it looks like some Linux performance problems are present in Linux 3.14.
A new kernel mailing list thread is active about Performance regression in v3.14. Johan Hovold explained his performance issue that began with Linux 3.14, "After updating my main system from v3.13 to v3.14.2, I found that the git bash-completion was extremely sluggish. Completing a file name would take roughly six rather than one second on this Haswell machine (i7-4770). (Other things, such as git rebase, also felt slower, but the completion issue was much more obvious and easy to measure)."
Johan came up with an easy way to reproduce the problem to investigate further and compare against the Linux 3.13 kernel. "Taking all but one core offline seems to make the problem go away, and so does using the performance rather than powersave governor of the intel_pstate cpufreq driver (on at least one of two online cores). Moving the mouse cursor makes to loop finish faster, and so does switching to a another terminal to print cpufreq/cpuinfo_cur_freq which was around cpuinfo_min_freq several times (when tracing, see below)...Comparing the traces I see a lot of functions taking ten times longer to finish, but I guess that's expected if this is indeed a cpufreq issue."
A separate kernel developer, Romain Francoise, wrote, "I've also noticed some performance issues with intel_pstate in powersave mode, in my case playing fullscreen video was very choppy. Switching to the performance governor fixed things as well. Looking at turbostat, the cores remain close to their minimal frequency pretty much all the time in powersave."
The Intel P-State driver maintainer is currently looking into the issue and hopefully will see the issue rectified. Stay tuned for updates.
A new kernel mailing list thread is active about Performance regression in v3.14. Johan Hovold explained his performance issue that began with Linux 3.14, "After updating my main system from v3.13 to v3.14.2, I found that the git bash-completion was extremely sluggish. Completing a file name would take roughly six rather than one second on this Haswell machine (i7-4770). (Other things, such as git rebase, also felt slower, but the completion issue was much more obvious and easy to measure)."
Johan came up with an easy way to reproduce the problem to investigate further and compare against the Linux 3.13 kernel. "Taking all but one core offline seems to make the problem go away, and so does using the performance rather than powersave governor of the intel_pstate cpufreq driver (on at least one of two online cores). Moving the mouse cursor makes to loop finish faster, and so does switching to a another terminal to print cpufreq/cpuinfo_cur_freq which was around cpuinfo_min_freq several times (when tracing, see below)...Comparing the traces I see a lot of functions taking ten times longer to finish, but I guess that's expected if this is indeed a cpufreq issue."
A separate kernel developer, Romain Francoise, wrote, "I've also noticed some performance issues with intel_pstate in powersave mode, in my case playing fullscreen video was very choppy. Switching to the performance governor fixed things as well. Looking at turbostat, the cores remain close to their minimal frequency pretty much all the time in powersave."
The Intel P-State driver maintainer is currently looking into the issue and hopefully will see the issue rectified. Stay tuned for updates.
6 Comments