You know what CUDA is, right?
There are two aspects of CUDA. One is the computational tool. Second, it is often used by Nvidia as a hook build a bigger moat around their hardware. Sections 3.2 and 3.3 of
their programming guide outlines interoperatbiliity with OpenGL , DirectX / Direct3D (up through 12) , and Vulkan ( but no Metal ... which has been out for as long as Vulkan . Nvidia has put lot of efforts into telling folks that they can sometimes skip OpenGL SL , MS computational API , and OpenCL to get to best performance by way of CUDA.
There two levels to CUDA. One, is the library and app level compiler language level (CUDA Runtime). However, there is also a low level aspect to CUDA (CUDA Driver). It is the latter were pragmatically GL SL , OpenCL , or some other computational code is going to come in and interact. The "closer to general purpose" computation code and shader code are all going to share the same base level resources so there are going to be some common control and compose(compile) elements there. CUDA's Runtime model is a bit more broader, general than Metal ( which started out more shader language replacement focused ).
It isn't a completely clean silo with zero interactions with other parts of the graphics stack.
That's the only thing that NVIDIA said they'd stop supporting. So it either means 1) They're dropping support for NVIDIA cards altogether in macOS,
Pragmatically, 1) is already true. How releases have they done on Mojave? On Catalina? It's been over 12 months since Mojave launched and nothing. How is that is that in the active supported status? At the point of being
two versions back on the OS , that is pretty much dropping the ball on support.
or 2) They're moving to Metal for their cards on macOS
They haven't moved at al in terms of the contemporary versions of macOS. That would more so be a restart because they haven't been moving at all for long while now. [ It is akin to a Windows driver developer saying "Oh, guess should start moving to supporting DirectX." as if that was an option for a committed, dutiful partner of Windows development. ]
I suppose it is possible that Nvidia would leave their protective moat behind. The amount of money they throw into constructing that on multiple platforms though makes that unlikely. But yeah, it isn't completely impossible.
We shouldn't jump to conclusions. But unfortunately, yeah, it might be true that they're just dropping Apple support altogether.
We can
speculate that it's true, but shouldn't jump to the conclusion yet.
Observations (that they aren't in a support status now) isn't really a conclusion (as in some long chain of inference or speculation ). This new CUDA stuff is going to old instances of macOS ( which are now in at best major security bug fix only status. ). From a Nvidia first view of the world, that is "on last" macOS support. From a macOS perspective though that is pretty far off.
