Apple dedicated GPU

[diamond.g]

Could the Nvidia RTX 4000 SFF be more efficient than Apple's GPU?

NVIDIA RTX 4000 SFF Ada Generation Graphics Card

Powered by the NVIDIA Ada Lovelace Architecture.

www.nvidia.com

Nah

 

[Xiao_Xi]

Nah

What made you believe that? That GPU has a similar computational capability to the M1 Ultra GPU using only 70W.

 

[leman]

Could the Nvidia RTX 4000 SFF be more efficient than Apple's GPU?

NVIDIA RTX 4000 SFF Ada Generation Graphics Card

Powered by the NVIDIA Ada Lovelace Architecture.

www.nvidia.com

For some applications, certainly. For example, I’d expect it to be much more efficient in matrix multiplication workloads. For general purpose compute, I don’t think there will be much difference this it other way. For games Apple will probably be more efficient.

 

[diamond.g]

What made you believe that? That GPU has a similar computational capability to the M1 Ultra GPU using only 70W.

Isn’t that about the same amount of power the Ultra uses?

Edit: I was also trying to be cheeky. Really would expect them to perform about the same with Ada being better at RT loads that use the hardware.

 

[Sydde]

I’d expect it to be much more efficient in matrix multiplication workloads.

But most of that is done in the neural engine on Apple SoCs, not the GPU.

 

[leman]

But most of that is done in the neural engine on Apple SoCs, not the GPU.

Not f you actually want to do matrix multiplication. Neural engine can do some ML stuff, sure, if you use Apples frameworks and your problem statement fits what the NPU can actually do.

But even then I’d expect the Nvidia GPU to have much better performance with at least comparable efficiency.