Two GPUs, three decoders, one stride bug, and what it took to reach 697 tokens/sec on one short request.
Blogs
Longer notes on things I’ve been learning.
Core DRAM dies, the base die, TSVs, the interposer, and how the data moves through all of it.
Longer notes on things I’ve been learning.
Two GPUs, three decoders, one stride bug, and what it took to reach 697 tokens/sec on one short request.
Core DRAM dies, the base die, TSVs, the interposer, and how the data moves through all of it.