You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 23: How GPUs work (15)
glinia

Given that it doesn't have a data cache, does the GPU access memory directly, or does it piggy-back off of the CPU cache hierarchy in some way?

moridin22

I think that other parts of the GPU mentioned in the "What's in a GPU?" slide probably take care of this, like the "Tex" block taking care of accessing textures.

Michael-hsiu

What is the limiting factor for the GPU? Is it speed of the ALU or fetch/decode speeds? I think both can be bottlenecks since per-fragment operations are happening in parallel, but I'm curious which one is more limiting.

zehric

I am also curious about whether arithmetic operations or memory operations are more expensive on the GPU. My gut says memory operations will still be slower. Also, I am pretty sure the GPU cannot access the CPU cache.

You must be enrolled in the course to comment