L2 on the GPU: where it sits

Accelerated Computing CUDA CUDA Programming and Performance
1

In the Nsight “Profiling Guide” I ran into this sentence:

The L2 cache sits between on-chip memory clients and the framebuffer.

My understanding of this statement is that L2 is then off chip. Moreover, I imagine the term “framebuffer” in this context refers to the global memory. Is that correct?

2

L2 is on-chip. Any of the GPU whitepapers will be a useful resource. For example here p9

framebuffer refers to global memory or the GPU DRAM that is off-chip.

1 Like