And to be clear, when you look at benchmarks that are showing a 6x or 10x or whatever performance advantage, that’s almost entirely the result of parallelism. The round reduction from 10 to 7 rounds provides a straightforward 1.3~1.4x performance bump, which is all well and good, but it’s not the source of the big numbers.
Why is BLAKE3 so much faster than blake2b? two reasons:
What is the security justification for it having fewer rounds?
And to be clear, when you look at benchmarks that are showing a 6x or 10x or whatever performance advantage, that’s almost entirely the result of parallelism. The round reduction from 10 to 7 rounds provides a straightforward 1.3~1.4x performance bump, which is all well and good, but it’s not the source of the big numbers.
Related: https://lobste.rs/s/daw1oa/reasons_prefer_blake3_over_sha256