Not familiar with HeliumOS specifically, but for a generic atomic distro I would try layering Python temporarily, and then getting rid of it when you’re done.

Fusion used to work but autodesk changed the redirects in their login system, so it no longer does…

Tragic. Especially since there’s no reason Fusion couldn’t be a webapp or PWA, autodesk already made it annoyingly cloud-focused.

Can second this: direct heating of anything is always going to be more efficient. Also, only ~25% of incident energy on a PV cell is actually captured as electriciy (see here for theoretical backing), and the rest is lost in a lot of ways, but much of it is converted to heat at the PV cell, and if you’re capturing that you’re using direct heating anyhow.

Can confirm that the asusd and and asus-linux work fine on Bluefin and Ubuntu/mint; the Devs dont support X11 (which Mint is still on), but you can build it with the X11 flags on the GH repo and it works fine.

Is that true? I thought that pairs of USB-A ports shared the same PCIe lanes, and USB-C each got their own set?

Edit: thinking about it a bit more, I suppose it could depend on how the SOC/chipset allocates those lanes, but in my experience when writing a single USB I’m usually limited by the thermals of the USB, and writing well below the speed of the port. I suppose if you were writing many at once (or if your USBs were nice) that could bottleneck on the port speed.

Good point. Though, the vast majority of ML training and use is tensor math on floating points, so largely dot and cross products, among other matrix operations.

I think you’re thinking of the famous fast inverse square root algorithm from Quake.

With respect to the top comment, the only reason 3d graphics are possible (even at 850W of power consumption) is due to taking a bunch of shortcuts and approximations like culling of polygons. If its a reasonable shortcut it either has or will be taken.

Good question. Odd not to include one.

I mean you can do HTML -> TeX -> PDF with Pandoc, or to any other format pretty much. I would say writing markdown and passing it to TeX or directly to PDF is the most practical.

No worries, it’s pretty hard to keep track when their naming scheme is “it has a K in it”…

Can you unmount it? You may not be able to change the boot flag while it’s mounted.

If that doesn’t work, you likely can’t remove the boot flag while the system is booted. Try booting from installation media and changing the flag there.

I think most ereaders support rendering HTML, why not just save the HTML page? Thatd be a lot easier than converting formats.

Okular is pretty great, I can’t find a package that does good annotation of PDFs built on GTK.

Yeah, came here to recommend this. It’s basic, but the UI is great.

It definitely is, but I guess domestic demand could outstrip production.

Yeah if that’s not on-brand for Intel, I don’t know what is! I wonder what the max power draw for the 14900H is, it’s gotta be close 😂

I assumed that Linux was not really under the control of the US, but I guess the Foundation is incoporated in the US as a 501©(6) and the kernel org itself is a 501©(3), so that does give Congress more levers on the kernel than I expected.

Not to mention that most (all?) of the major corporate funders of the kernel are US-based…

I really hope the kernel doesnt get (geo)politicized.

Edit: based on @RobotToaster’s link, yeah it looks like every major “employer” contributor to the kernel other than Huawei, Linaro, Arm, and Suse are American. Arm is probably working mostly on support for its architecture, so I guess it’s Linaro (UK) and Suse(DE).

That’s not to downplay the role of independent contributors, but it seems like a good indicator of the “power of the purse strings”.

Edit 2: here’s a more recent set of development statistics from LWN. Looks like the ordering has changed quite a bit since 2022, or it varies a lot with each kernel version

The most confusing thing is that “200V” isnt a CPU, it’s the equivalent of “15th gen”.

The numbers before the V are un-parseable, but at least for the actual parts it’s a “Ultra 7 236.1425926V” or something

The original article [https://www.phoronix.com/review/core-ultra-7-lunar-lake-linux]

Yeah that was my reaction…chip design is the “easy” part of the creation of a new chip, and the fact that it the CPU was of course fabbed in Taiwan is a sign of just how much ground Chinese fabs still have to cover. And relying entirely on SK Hynix NAND and memory says a lot.