Daily Archives: 2021-08-19

Article note: The problem with all the BIG.little/Hybrid/Heterogeneous core designs is that OS schedulers that can use them effectively are somewhere between "experimental" and "active theoretical research topics." Microsoft seems to have something rolling out with Windows 11, and I'm _guessing_ it's the retarded child of the nifty convex optimization stuff they had going at Microsoft Research like a decade ago (RIP Burton Smith), but there is not a known-good solution in general for HMP/Energy Aware/etc. scheduling, and platforms that can do that are stuck in chicken-and-egg hell until there are better software solutions.

Enlarge / It's a bit too early for photos of Alder Lake-S CPUs, much less Raptor Lake-S—so here's a gorgeous photo of an alder tree on the shore of Llyn Gwynant, in North Wales' Snowdonia National Park. (credit: R A Kearton via Getty Images)

It looks like big.little CPU design—an architecture that includes both fast, power-hungry cores and slower, more power-efficient cores—is here to stay in the x86_64 world, according to unverified insider information leaked by wccftech and AdoredTV.

Intel’s big/little designs enter round two

At Intel's 2021 Architecture day, the company confirmed that its upcoming Alder Lake (12th generation) processors will use a mixture of performance and efficiency cores. This brings the company's discontinued 2020 Lakefield design concept firmly into the mainstream.

Big.little designs run time-sensitive tasks on bigger, hotter performance cores while running background tasks on slower but much less power-hungry cores. This architecture is near-universal in the ARM world—which now includes Apple M1 Macs as well as Android and iOS phones and tablets—but it's far less common in the x86_64 "traditional computing" world.

Article note: Online platforms that trade mostly on smut trying to pivot always goes well, with all those... no survivors... who have tried.

Article note: We really need to be careful about not promoting hygiene theater solutions that will wreck the credibility of real effective public policy. Also, as usual, the narrative has de-emphasized solutions that cost institutions money or power (more/better circulation, HEPA filters, keeping jobs that have no real reason to be in-person remote indefinitely, etc.) in favor of things that burden inviduals.

Article note: Nifty idea. DRASTICALLY inadequate stiffness. Hopefully it'll attract some CAM work on rotational axes. They're talking about custom CAM but don't seem to have the necessary skills onboard. It would be nice if they got attention onto FreeCAD Path workbench or something.