Hi people, I want say if I can chose the number of GPU core to use or if is automatic scalable. Thank you :)
only if u have SLI and u don’t want to use first or second cards cuda cores.
maybe u can turn of some of cores by remaking bios, but i never tryed that.
Then if I have a Tesla S2050 with 448 core for GPU, the CUDA program to exploits all the core??
Automatically scalable. It will fill the gpu if it is needed.
Hi,
i am new to cuda , just read this topic mentioned here…hence thought to ask my doubt…hope you people don’t mind.
if in a code ,i have max number of threads and blocks that my card can support then is all my cores utilized…or it depends on the operation?
Depends on the operation. There is the theoretical occupancy which depends on the resources used (registers, shared memory, etc…). For me the only criteria to how I choose the grid of blocks and threads is how fast does it finish.