3 ms·
Probably more accurate to multiply that by 4. Each SM is split into 4 partitions that can each execute different instructions but with shared L1 cache.
by kcb 2y ago
Probably more accurate to multiply that by 4. Each SM is split into 4 partitions that can each execute different instructions but with shared L1 cache.