3 ms·
I think for x86 you are (at least very close to) correct, but I've been seeing a serious shift towards ARM chips and (my pipe dream) soon RISCV. X86 may just ne
by brutal_chaos_ 5y ago
I think for x86 you are (at least very close to) correct, but I've been seeing a serious shift towards ARM chips and (my pipe dream) soon RISCV. X86 may just need to fall to the wayside at this point. Maybe PineBooks could be released with ARM chips too! (RISCV as well, but peripherals need to be a real thing(TM) first). Heck, the Pi400 is a good start in that direction (though obviously isn't enterprise ready yet).
- userbinator 5y agoThe openness situation for ARM is even worse. Case in point: the countless Android devices out there.
- swiley 5y agoIt's really a Qualcomm problem more than an Android problem. Everyone seems to think Qualcomm==Arm though because in the USA it pretty much does.
- userbinator 5y agoNot just Qualcomm but the vast majority of ARM SoCs used in Android devices, and in general, have no public documentation.
- SwimSwimHungry 5y agoHonestly, I vastly prefer OpenPOWER over ARM from a consistency in implementation perspective. Not to mention, at least with POWER9, no mandatory binary blobs or opaque management engine chips that can't be properly audited.
- selfhoster11 5y agoOpenPOWER is extremely expensive though. There is no universe in which I would pay several times the price for a high-powered x86 gaming PC/workstation just to have an open-source CPU and hardware like Talos. It has to be both open and affordable.