4 ms·
The way you make ST faster is by moving more to hardware. Instead of computing CRC32 or AES one instruction at a time according to their algorithm, you just ha
by dist-epoch 2y ago
The way you make ST faster is by moving more to hardware.
Instead of computing CRC32 or AES one instruction at a time according to their algorithm, you just have a hardware CRC/AES instruction which does it in one go.
Same for decoding movies, or running neural-networks.
AMD/Intel still deliver 10-15% ST improvement with each generation. Compound that over 20 years and it's way more than 3-4 times faster.