I have a custom-built NVIDIA GPU/PC system, and have heard that NVIDIA GPUs ‘optimize’ Blender for a user. But I want to better understand, if anyone can explain, how exactly an NVIDIA GPU in a Windows system optimizes the program compared to an Apple system. As a mac user prior to building my first PC a few years ago (which I use exclusively for Blender/editing), and one who prefers Apple’s UI (still using my macbook pro for every use case other than Blender/editing), I am curious if the performance of an NVIDIA GPU in Blender significantly outperforms an Apple chip with comparable performance metrics (for example, comparing a 5090 card to the highest spec Mac Studio GPU), and if so, how specifically? Is it apples to apples? Or are there any variables - perhaps intangible - that make the comparison apples to oranges? Thanks!
I guess that NVIDIA GPUs (especially when it comes to high-end class) are always better in terms of performance compared to Apple solutions. Also, NVIDIA uses so-called “CUDA” cores, which can significantly boost performance and, thus, reduce time rendering. There’s also an OptiX, which is another NVIDIA technology that increases performance by using so-called RT cores (available on RTX cards only) which speeds up cycles renders and stuff.
Also, you can check benchmarks and see who wins who [here].
how exactly an NVIDIA GPU in a Windows system optimizes the program compared to an Apple system
The answer would be, I guess, — thanks to CUDA and OptiX cores. Also, I heard that Blender always performs better on GeForce chips rather than AMD ones, for example. I guess the same goes for Apple too.
On the other note, Apple Silicon solutions are always improving, but still they far from desktop ones.
Hope that helps,
Cheers
It seems RTX 5090 outperforms any Apple device very significantly regarding render speed. It is apples to oranges regarding render speed. I would guess that you can get a PC with RTX 5090 cheaper as well. That’s a guess though, I haven’t checked today.
You can probably get more memory with Apple though.
I suppose what I am wondering is if there are any intangible, qualitative superiorities outside of render speed. Say, like comparing two digital cinema cameras. Comparing Arri to Sony, the digital sensor in a camera produced by each company captures data in the same way, but the proprietary sensor of each company results in aesthetically different images. Does Blender’s codebase - interfacing with an NVIDIA chip compared to an Apple chip - result in any difference along the lines of my camera analogy?
Yes, of course, if you choose Apple, you get all the intangible quality superiorities in workflow or the flow of work and what not. 
For a price though. I would even say it’s without a doubt best at superiorities that are specifically intangible.
But the large amount of the unified memory, whatever that means does sound nice to be serious. …if you needed it. Which might be, or maybe not.
For the most part no, the devs try their very best to make sure that the output from Cycles, with all the same settings, will look the same no matter what it is rendered on, be that Nvidia, AMD, Intel, Apple GPU’s or AMD, Intel, Apple CPU.
HOWEVER, having said that, there are usually some very small/minor differences. In fact just within Nvidia, a CUDA render can be a very little different then an OptiX render. For a still image, it makes zero difference and in fact to see any difference, you actually need to comapre the two renders in an image editor to make them be highlighted.
For an animation, it could matter, as in rendering to frames and having each frame be randomly rendered on a different device. In that case, minor different noise patterns and hence denoising, etc could show up a lot more as flicker and the like. This is why render farms usually don’t mix and match and even if they do, if you go to render an animation, that stick to the same device type, so say all Intel CPU or all Nvidia OptiX GPU.