2 ms·
Also, modern xeons can do 32 dual precision floating point ops per cycle, per core. Since they have dozens of cores, that’s another factor of 1000, with your 1
by hedora 5y ago
Also, modern xeons can do 32 dual precision floating point ops per cycle, per core. Since they have dozens of cores, that’s another factor of 1000, with your 10x overhead estimate for the C64, that brings the speedup to ~ a billion. (96,000 * (10 to 150x) * 1000 ~= 1-10 billion)
- kragen 5y agoYes, SIMD operations on CPUs are in many ways similar to GPU computation.