M1 GPU Cores - Why so few vs PC GPUs ?

A PC GPU (AMD, Nvidia) has thousands of Cores, when the M1 Ultra has around 48 cores. How come such a huge difference in amount of cores can anyway make the performance at par ?

Posted on Apr 2, 2022 12:27 AM

Reply
Question marked as Top-ranking reply

Posted on Apr 2, 2022 8:08 AM

What Nvidea calls a 'Cuda Core' is what AMD calls a "Unified Shader" and what Intel calls a "Shading Unit". Each is too primitive to be called a full CPU.


A better comparison is the Nvidea "Streaming MultiProcessor" and the AMD "Compute Unit". The fastest RTX 3090 only has 82 of those, while the AMD RX 6900XT only has 80 respectively. Thats just 25-28% more than an Apple M1-Ultra 64.


With architectural differences, clock rate differences, memory bandwidth differences, etc performance can be extremely close depending on workload.

4 replies
Question marked as Top-ranking reply

Apr 2, 2022 8:08 AM in response to EarthHobbit2

What Nvidea calls a 'Cuda Core' is what AMD calls a "Unified Shader" and what Intel calls a "Shading Unit". Each is too primitive to be called a full CPU.


A better comparison is the Nvidea "Streaming MultiProcessor" and the AMD "Compute Unit". The fastest RTX 3090 only has 82 of those, while the AMD RX 6900XT only has 80 respectively. Thats just 25-28% more than an Apple M1-Ultra 64.


With architectural differences, clock rate differences, memory bandwidth differences, etc performance can be extremely close depending on workload.

This thread has been closed by the system or the community team. You may vote for any posts you find helpful, or search the Community for additional answers.

M1 GPU Cores - Why so few vs PC GPUs ?

Welcome to Apple Support Community
A forum where Apple customers help each other with their products. Get started with your Apple Account.