mcdiniz wrote:still no exe .. this revision (661)?,, best
Fauzi wrote:does the JA compile get stronger than normal compile??
Vladimir wrote:Protector 1.5.0 x64 vs. Protector 1.4.0 x64
Time Control: 3 Minutes
1000 GamesCode:
1 Protector 1.5.0 x64 +92 +401/=457/-142 62.95% 629.5/1000
2 Protector 1.4.0 x64 -92 +142/=457/-401 37.05% 370.5/1000
1000 out of 1000 games played
PGN: depositfiles.com mgp66jksy
deefree49 wrote:So far it isn't looking like an improvement over 1.7.0. I'm running some 3' games, no bonus, 16 games no book, 10 games with book and here are the results;
Protector 1.7.0 beta r938 x64; +10 =15 -1 over Protector 1.8.0 x64 (17.5 to 8.5)
i7 with 8 logical processors. No ponder.
Seems like some tweaking on the new release might be needed.
Nevertheless, I appreciate the work and who knows, this may be an anomaly. If not, Raimund will figure it out.
deefree49 wrote:Hi Tomfl,
Interesting that you asked that. I did 2 little matches and combined the results. The last 10 games were set at 4 CPUs instead of 8. The results still favored 1.7.0. That result was +3 =7 -0. So the only win on my computer by 1.8.0 was when I was running both at 8 CPUs.
I'm not sure what to set it at to tell you the truth. Some engines automatically set at 8 on my computer. The benchmark says "8 logical processors found".
I did notice the KN/s on 1.8.0 were a lot lower than 1.7.0 during the games but I've read in discussion groups that the number isn't all that important. Honestly I really don't know myself what the truth is. Anyway, 1.7.0 was showing about 3 times as many KN/s as 1.8.0. It was approximately 7500 to 2500. I thought that seemed like a big difference and maybe that something is amiss in the program.
mcdiniz wrote: