SP vs. Quad

[AdminX]

I'm starting some tests of Rybka 3 on one core vs. various other leading programs on four. All at 40/2', then 40/1' repeating. Here are the early returns:

http://rybkaforum.net/cgi-bin/rybkaforu ... l?tid=5343

How far is Rybka going to go to prove it's the best already!?

[bob]

I'm starting some tests of Rybka 3 on one core vs. various other leading programs on four. All at 40/2', then 40/1' repeating. Here are the early returns:

http://rybkaforum.net/cgi-bin/rybkaforu ... l?tid=5343

How far is Rybka going to go to prove it's the best already!?

A much more interesting test would be to play Rybka vs your group using 1 cpu for all, then Rybka using 4 vs the group with 1, to get an idea of the SMP improvement, which is something that seems to be an unknown and the subject of discussion.

[AdminX]

First match: Rybka 3 vs. Naum 3.1: +25=18-8 for +120 Elo.
update: +48=48-32 for +43 Elo.
final: +59=52-37 for +52 Elo.

Second match: Rybka 3 vs. Deep Fritz 10.1: +33=11-10 for +158 Elo.
update: +87=38-28 for +141 Elo.
final: +101=49-32 for +139 Elo.

Third match: Rybka 3 one core vs. Naum 3.1 on octal: +46=42-19 for +89 Elo.
final: +58=50-22 for +98 Elo.

[Tony Thomas]

First match: Rybka 3 vs. Naum 3.1: +25=18-8 for +120 Elo.
update: +48=48-32 for +43 Elo.
final: +59=52-37 for +52 Elo.

Second match: Rybka 3 vs. Deep Fritz 10.1: +33=11-10 for +158 Elo.
update: +87=38-28 for +141 Elo.
final: +101=49-32 for +139 Elo.

Third match: Rybka 3 one core vs. Naum 3.1 on octal: +46=42-19 for +89 Elo.
final: +58=50-22 for +98 Elo.

Weird result for Naum, hopefully the fluke result is just a statistical anomaly. I would have expected Naum on 8 cores to equal the result of Rybka by looking at the initial quad test.