Does anyone know related information about L2 cache in Fermi? I heard that it is as slow as global memory, and using L2 is just to increase memory bandwidth. But I can not find an official source confirming this. Has anyone measured L2 hit latency? What about size, row size and other parameters?
Essentially, how does L2 read misses affect performance? In my sense, L2 only makes sense in memory related applications. Please feel free to give your opinion.
thank
source
share