Post #1396184

2026-03-31 09:03 UTC

@lproven There’s decisions that are worth the inconvenience like unified memory, and some that could have been more open. The display architecture could have been designed more like it is on UEFI systems, or they could have shared a spec confidentially with AMD. I don’t think it’s malicious, they can iterate faster when the only 3rd-party they might break is Asahi. They got what they deserved for letting the Apple Silicon Mac Pro be limited though - nobody wanted it. Then again, see the big debacle with Nvidia and High Sierra. They have their beef with Nvidia, but making Mac Pro customers collateral damage was a choice.

Replies (1)

@lproven@social.vivaldi.net 2026-03-31 11:27

@kirb I am not so sure. As the benefits (_and_ costs) to Apple customers of its design decisions affect its market and sales, then other companies will work out which bits work, which bits they can imitate, which bits they can replace, and which bits don't work for anyone else. My thesis is just that we'll see comparable levels of integration from other vendors. Not _the same_ but similar at least viewed from some level. Current GPUs for x86 machines sit on their own board or even on the end of a cable. They have their own RAM. They are nearly separate dedicated computers. My guess is that the factors driving the huge size and complexity of these GPUs will go away with greater integration. Either they'll cease to exist, replaced by much simpler GPUs that get their speed by being packaged with the CPU. _Or_ they will evolve into actually fully-independent computers with their own general-purpose CPUs, like the remarkable Bolt Graphics units: https://www.theregister.com/2025/10/29/bolt_graphics_zeus_gpu/

Open ##1396185