Are global memory loads/stores going through caches in devices with compute capability 1.x?

Question

I'm using an old Tesla GPU C1060, G80 architecture. I'm wondering if data requested by threads from the global memory address space is cached, as it happens in more recent architectures.

Answer 1

CUDA Programming Guide, Section G.3.1:

The local and global memory spaces reside in device memory and are not cached.