Eh, Nvidia shares a lot of the blame on that one though, because the signed firmware they use post-Fermi can't be legally redistributed in any way, and is embed

Eh, Nvidia shares a lot of the blame on that one though, because the signed firmware they use post-Fermi can't be legally redistributed in any way, and is embedded in the proprietary graphics driver blobs, so it can't be extracted. That firmware is required to adjust GPU clocks, which start off very low. So the choice isn't "Good proprietary drivers" vs "Less featured open drivers", it's "proprietary drivers" vs "drivers that are legally forbidden from running above 0.5% their advertised clock speed, no matter if it's otherwise fully featured".[1] In contrast, AMD and Intel have open drivers in Mesa that work very well and fully supported and AMD allows its signed VBIOS to be redistributed, so it's clearly not just a matter of chip complexity, but the position of the vendor as well. In that vein I agree with your original point, though: I suspect we'll probably never see Linux on the M1 in anything but the most superficial form that lacks all the really good stuff. You might get the 8 core CPU, SSD/RAM, some peripherials. But no Neural Engine, no GPU, video decoders or image processors, power management, security features in the T2 like attested boot or encrypted key storage, etc. The only way that'll ever change is if Apple makes it happen. [1] For reference, the open source Nouveau driver for Fermi cards, nvc0, is the 4th most well supported Mesa driver in terms of features implemented (almost 90% of all features, surpassed only by radeonsi and i915). So it's not like there's no interest in FOSS Nvidia drivers... https://mesamatrix.net/ https://mesamatrix.net/