Ask HN: What’s the advantage of higher memory bandwidth in Apple Silicon chips?
3 comments
Short answer: graphics, machine learning, and file caching (which the OS does automatically for you).
Long(er) answer: For the most part, the entire system benefits from it. Across the stack you get a boost because of file caching, and the boost it gives to I/O dependent subsystems which make the entire system faster.
e.g. you don't need to use machine learning explicitly to benefit from it, macOS uses it heavily for some operations (e.g. text identification when you open an image on preview).
Sorry it's a vague answer, but the nuances of the benefit are highly dependent on the workload.
Long(er) answer: For the most part, the entire system benefits from it. Across the stack you get a boost because of file caching, and the boost it gives to I/O dependent subsystems which make the entire system faster.
e.g. you don't need to use machine learning explicitly to benefit from it, macOS uses it heavily for some operations (e.g. text identification when you open an image on preview).
Sorry it's a vague answer, but the nuances of the benefit are highly dependent on the workload.
It’s used for graphics as well which requires lots of bandwidth.
GPU needs high bandwidth. That's why you would notice x86 apu are very bandwidth starved because they cannot use the memory for the graphic cards
- M1: 66.67GB/s - M2: 100GB/s - M1 Pro: 200GB/s - M1 Max: 400GB/s - M1 Ultra: 800GB/s
The question is, which app or task will take advantage of higher unified memory bandwidth?