A problem with that is, CUDA is used for a bunch of scientific computing applications (and probably bitcoin miners), besides just the expected graphics use. I know a few scientists who really wish Apple and Nvidia got along, for precisely this reason.
I don’t know that they are hurting even in their present situation. Cloud gaming is the next big push, so they are tossing their hat into that ring with a completely different take, where you control the library. With a solid connection, this service is probably a lot cheaper than building a dedicated gaming rig, as NVIDIA is in a place where they can price their GPUs at insanity and sell every one of them. A decent 1080p GPU is around $300-400 these days. That’s at least a couple years worth of GeForce Now gaming.
Uh what? Nvidia makes an ARM cpu. And they still lead the gaming and 3D rending market. What are you smoking?
Can't use it on most Blizzard games, specifically WoW
They're referring to desktop/laptop CPU akin to AMD's Ryzen, Intel's Core and Apple's M1 series. There is no Nvidia version of these CPUs.
That’s not a problem, though. Those that need CUDA still have CUDA, likely in a stable non-macOS system. Apple and others are focusing on the huge market of people that may need some performance beyond the CPU (for science, machine learning or other purposes), but not necessarily CUDA. As those solutions come on-line, the libraries they use will be created and refined over time. There are things that we think of today as being CUDA only solutions that will arise on these non-Nvidia systems and may show surprising performance on devices where we’d never expect such performance previously.
Both Intel and AMD have a path to evolving systems whereby they’re both broadly compatible with the current Windows/Linux OS’s while also providing the benefits of tightly coupled CPU/GPU’s. Nvidia’s on the outside of this and would have to work within either Intel’s or AMD’s constraints to achieve similar performance/efficiencies. It’s possible that Nvidia will necessarily need to cede the embedded market altogether eventually.
They do not have to leave embedded market at all, they still have a good but aging solution with Tegra for their Nvidia Shield and Switch consoles. They also are quickly expending their networking tech stack; which they can sell to routers, network switches, etc.
Nvidia needed to acquire ARM to take control of the direction of the platform and the licenses for integration with their GPU IP. That's why there were willing to pay $40 Billion for the company. Nvidia leads the market today. I'm talking about the future. If they lose the "socket" on future generation systems then they'll be out of the game.
It's diffidently a sign that AS based Macs are being recognized by game service companies looking for more subscribers.
OOPS! You are correct. By “embedded” what I really meant was “integrated” i.e. those devices becoming a larger and larger chunk of the computing device market. If future mobile systems are able to have impressive battery life AND performance based on having CPU/GPU working in concert, there wouldn’t be much opening left for Nvidia to provide a solution that won’t be as integrated, won’t work as well, and won’t be as efficient.
Lots of licensees make ARM CPUs. Want Nvidia wanted was to control the direction of the ARM architecture, and to design-in and integrate future GPUs into the ARM ecosystem to sell those designs at scale to the market. They need to own ARM to have that control. Again, that's why they were willing to spend $40B for the company.
