Tiger Lake + Renoir On Ubuntu Linux For Battery vs. AC Performance

Here is a look at the CPU peak frequency over the entire duration of the benchmarks. The Ryzen 5 4500U with the Lenovo laptop running off AC power had an average peak clock frequency of 2.7GHz but that fell to 2.4GHz on battery power. The Core i7 1165G7 in the premium Dell XPS 13 9310 laptop went from 2.6GHz to 2.4GHz on average.

For those curious about the CPU temperature difference for these two distinct laptops based on their power source, here is that data. Additionally, the battery power consumption data too over the entire span of benchmarks for those interested.

Here is the geometric mean of all the benchmarks ran in each configuration. The Core i7 1165G7 when running off battery power was at 88% that of the performance when on AC power for the Dell XPS laptop. The Ryzen 5 4500U meanwhile was at 86% the performance within that lower-end laptop. So overall the performance wasn't too relatively different at a high-level. Some tests did worse on battery with AMD while the Intel Tiger Lake laptop tended to do much worse in a smaller subset of tests.

But seeing the Intel Tiger Lake laptop tending to perform much worse in the latter tests got me wondering... At that point it was under 50% the battery capacity remaining. Perhaps as the battery was being closer to exhaustion, the Intel Tiger Lake thermal behavior was becoming more conservative? This would be through Intel's thermal daemon and their Dynamic Platform and Thermal Framework (DPTF) that is a murky black box to Linux. So I charged up the Dell XPS 13 laptop to around 80% and repeated those tests where at under 50% were showing a larger difference.

Yep, depending upon the battery charge level, the performance on the Dell XPS 13 9310 / Core i7 1165G7 was wildly different... Presumably for other EVO notebooks as well. It does make some sense to try to extend the battery life more if it's looking like the battery will be depleted soon while that logic appears to be all happening within the DPTF area that has already caused a mess on Linux as outlined yesterday. I'd assume too that for Intel's presentation on the matter they were showing all of their benchmarks for each application running on a freshly charged battery.

So, long story short, in some cases the AMD Ryzen 5 4500U being tested did see more conservative clocking behavior when running on battery compared to AC power. It depended upon the particular workload though as to the relative hit and in some areas particularly with multi-threaded workloads the performance could come out around the same. In some of those multi-threaded workloads, the much cheaper Ryzen 5 4500U laptop was already outperforming the much pricier Intel Core i7 1165G7. The Dell XPS 13 notebook with i7-1165G7 in some workloads was seeing a performance hit on battery, but especially more so as the battery was diminished. Intel DPTF appears to cause the performance behavior to shift depending upon the battery level at which point it seems to severely keep the CPU down-clocked in order to extend the battery life rather than being more conservative on a full charge.

Those wanting to dig through all of this data in closer detail can see this OpenBenchmarking.org result file for all the results plus a few more. That's all for now based on the available hardware. Those interested in seeing more laptops tested under Linux at Phoronix can always view this site without any ad-blocker or consider joining Phoronix Premium this holiday season.