HIARCS evaluation speed and tuning

[Rob Appleby]

I'm running deep HIARCS 14 on a 13" retina mbp (late 2013), with 4 threads and a variety of hash table sizes, from 128Mb (the default) to 2Gb. I'm quite surprised to see that HIARCS evaluates at around 900K nps, while stockfish, for instance, evaluates at 3Mnps and upwards (in opening positions and the early middle game with most pieces still on the board). Since I'm using the engines for analysis rather than play, this is quite a bottle neck in the use of HIARCS for deep searches. Is there any way to tune it to run faster, or this just normal?

[herO]

I'm running deep HIARCS 14 on a 13" retina mbp (late 2013), with 4 threads and a variety of hash table sizes, from 128Mb (the default) to 2Gb. I'm quite surprised to see that HIARCS evaluates at around 900K nps, while stockfish, for instance, evaluates at 3Mnps and upwards (in opening positions and the early middle game with most pieces still on the board). Since I'm using the engines for analysis rather than play, this is quite a bottle neck in the use of HIARCS for deep searches. Is there any way to tune it to run faster, or this just normal?

It is absolutely normal. Stockfish is uses mainly brute force approach, Hiarcs has more chess knowledge. This is the reason why Hiarcs search "slowly" than Stockfish. An engine search slower if its evalution function has more chess knowledge.

[Rob Appleby]

Thanks. Having searched the forum a bit more thoroughly, this was what I have understood to be the situation.