With the latest speedup of Triton VM, intoduced in Triton VM v9.0.0, proof generation saw a massive speedup on powerful CPUs. On a Threadripper 7995wx, we observed a 6.8x speedup, seeing the time it took to produce a proof of padded height 2^{21} (biggest size the composers need today) drop from over 100 seconds to less than 15 seconds.
This is of course a very welcome development for the network as a whole. But we must warn the block composers, that compete on building the fastest block proposal, to not rest on their laurels! “Why?”, you may ask? “Because of succinctness”, we say! Succinctness will require the block composer to produce more proofs than today. So getting a faster Triton VM prover than we have today will still be a worthwhile effort, as the network is going to increase the workload that it puts on the shoulders of block composers.