Running Intel Tiger Lake On The Linux 5.10 Kernel
Linux 5.10 has many new features and improvements for this kernel that should debut as stable in December. When it comes to Tiger Lake specifically with Linux 5.10 there is the continued Gen12/Xe graphics work with the open-source Intel DRM driver as the main area seeing attention this cycle... In particular one item that came to mind and motivating this testing was the merging of Tiger Lake HOBL support. HOBL is short for "Hours of Battery Life" and thus curious what this means for the power consumption on Linux 5.10 relative to Linux 5.9.
The only change during testing on the Dell XPS with Core i7-1165G7 was switching out Linux 5.9.1 stable for a Linux 5.10 snapshot. 75 different benchmarks were run this round on the two kernels in looking for any changes in performance.
Of the 75 tests carried out, a majority of the time (57 tests) saw Linux 5.10 delivering better performance over Linux 5.9. But in most cases the difference on this Tiger Lake laptop was quite small.
If taking the geometric mean of all the individual benchmark results, Linux 5.10 with this Dell i7-1165G7 laptop was just about 1% faster overall than Linux 5.9...
There were only a few tests with measurable shifts in performance, for the most part leaning in favor of Linux 5.10. This includes Vulkan compute performance as one of the areas seeing better performance on Linux 5.10 with the updated DRM driver.
But what about the power consumption on Linux 5.10?
When monitoring via the Intel RAPL interfaces, the average power consumption on Linux 5.10 for the Tiger Lake SoC was 20.59 Watts compared to 20.89 Watts on Linux 5.9. The minimum power consumption recorded was also a half Watt lower. In my other tests in looking at the Linux 5.10 power consumption on this laptop, indeed when completely idle RAPL reports ~0.5 Watts lower which may be due to HOBL or other factors. But even under light load scenarios or where the CPU is busy but the GPU and screen idling there was no measurable difference in the power usage in going from Linux 5.9 to Linux 5.10.
The Tiger Lake CPU temperature also amounts to being the same throughout the entire span of testing given no major difference in the CPU power consumption.
So overall no major shifts in Intel Tiger Lake performance off Linux 5.10 but at least a few Vulkan compute tests did seem to benefit from the updated graphics module. Linux 5.10 tests on other hardware forthcoming.
2 Comments
