Intel: AMD Weak On Battery-Powered Laptop Performance - But DPTF On Linux Still Sucks
Intel held a virtual event last week to basically plead their case that AMD Ryzen laptops are coming up short on battery-powered performance compared to their own offerings. It was all Windows focused, but at least given their emphasis now on battery performance gave me another opportunity to prod over the lackluster state of Intel DPTF support on Linux with it not being pleasant out-of-the-box and one of the few areas encumbered by blobs or lack of public documentation.
Intel's own testing feels that "[AMD] competition sacrifices significant performance to be on-par with Intel on battery life" and provided their own analysis of five AMD laptops and five Intel laptops.
Under Windows they find that the Intel laptops largely delivered similar performance between AC and battery states during their benchmarking. But on the AMD Ryzen laptops they found as much as a 38% drop in performance when running off battery.
Or in some cases their own tests argue that AMD laptops can lose ~48% performance when unplugged.
They basically promote that Intel Tiger Lake notebooks deliver better plugged/unplugged performance than AMD Ryzen systems... And they also were suggesting to reviewers that they do their notebook benchmarking in an unplugged state or otherwise test both modes.
Again, all Windows data so Linux laptop users... But with my love for benchmarking, I do plan to devote some time over Thanksgiving with my limited selection of notebooks to carry out some similar tests on Ubuntu and see how the performance differs from AC to battery between AMD and Intel. I'll be highly surprised if there is any big AMD vs. Intel battery performance difference on Linux.
I didn't expect them to have any Linux numbers at the presentation, but I dialed in so I could ask about DPTF being one of the few sore spots in their Linux support. While for nearly all areas the Intel Linux laptop support is in good standing from their side, it's less than desirable with regards to Intel Dynamic Platform and Thermal Framework. That's where many Linux distributions don't even ship Intel Thermald and it's not in an optimal configuration unless running the binary-only dptfxtract utility, etc. DPTF isn't properly documented and the out-of-the-box support on Linux as a result is less than optimal. Weird DPTF tables on various laptops have caused performance regressions in the past under Linux as we've spotted as well as other performance issues.
Take for example the recent issue of some HP laptops have been performing less than optimally on AC power on some Linux kernel versions due to DPTF handling issues. There have also been thermal throttling issues under Linux with Lenovo acknowledging is due to closed-source Intel DPTF settings. Earlier this year some work got underway outside of Intel on reverse engineering Intel's DPTF adaptive policy but that didn't seem to bear any dramatic improvements yet.
So now with Intel seemingly being all the more interested in battery performance and believing they have an advantage over AMD in this area, I inquired if they are doing anything regarding improving DPTF on Linux... At the briefing they had no answer but would get back to me. I'm still waiting, but will be ecstatic and pass it along should they actually have any real improvements they are working on in this area to improve the out-of-the-box Linux laptop experience and/or better open-source DPTF support or public documentation.
Add A Comment