4 ms·
Nah, still ~300GB/s memory bandwidth. That will be slower than the M5 max, by a wide margin for LLM inference.
by aenis 4mo ago
Nah, still ~300GB/s memory bandwidth. That will be slower than the M5 max, by a wide margin for LLM inference.
loading thread…