TalkChess.com

Hi, Harum.

Each new version shows relevant improvements compared with the previous one. At this moment I have my two computes (one with 4 threads limit w32, the other with 8 threads limit w64) testing the versions 6 of Igorrit.

When Igorrit v6 uses 8 threads ir runs about 55/60% faster than Igorrit V5 using 4 threads. A very nice scaling proportion.

These guys are doing a great job!. (The issue clone or not is useless for me, I simply enjoy what a free engine is offering to computer chess lovers).

Regards,

Tom.

Tomcass wrote:Hi, Harum.

Each new version shows relevant improvements compared with the previous one. At this moment I have my two computes (one with 4 threads limit w32, the other with 8 threads limit w64) testing the versions 6 of Igorrit.

When Igorrit v6 uses 8 threads ir runs about 55/60% faster than Igorrit V5 using 4 threads. A very nice scaling proportion.

These guys are doing a great job!. (The issue clone or not is useless for me, I simply enjoy what a free engine is offering to computer chess lovers).

Regards,

Tom.

Thanks Tom.

I'll download and test v6. You say v6 is much more improved than v5?

Best,

I don't feel that V6 has any improvement over V5 other than the potential to use up to 8 threads. In one of my computers, where only 4 threads are available, I expect a very similar result.

In my faster computer I expected that the substantial increase in speed (almost 60%) would imply a better result, but by now it is roughly the same if not slightly worse. I will post the results of both tests here in about four hours.

Regards from Barcelona.

Tom.

Tomcass wrote:Hi, Harum.

Each new version shows relevant improvements compared with the previous one. At this moment I have my two computes (one with 4 threads limit w32, the other with 8 threads limit w64) testing the versions 6 of Igorrit.

When Igorrit v6 uses 8 threads ir runs about 55/60% faster than Igorrit V5 using 4 threads. A very nice scaling proportion.

These guys are doing a great job!. (The issue clone or not is useless for me, I simply enjoy what a free engine is offering to computer chess lovers).

Regards,

Tom.

I agree with you on that, they are very clever guys who ever they are !

Tomcass wrote:I don't feel that V6 has any improvement over V5 other than the potential to use up to 8 threads. In one of my computers, where only 4 threads are available, I expect a very similar result.

In my faster computer I expected that the substantial increase in speed (almost 60%) would imply a better result, but by now it is roughly the same if not slightly worse. I will post the results of both tests here in about four hours.

Regards from Barcelona.

Tom.

Hello Tom,
The time management on v6 is a major disaster with 4 and 8 cores
Have you a link to download the v5 ?

Florian

Florian wrote:

Tomcass wrote:I don't feel that V6 has any improvement over V5 other than the potential to use up to 8 threads. In one of my computers, where only 4 threads are available, I expect a very similar result.

In my faster computer I expected that the substantial increase in speed (almost 60%) would imply a better result, but by now it is roughly the same if not slightly worse. I will post the results of both tests here in about four hours.

Regards from Barcelona.

Tom.

Hello Tom,
The time management on v6 is a major disaster with 4 and 8 cores
Have you a link to download the v5 ?

Florian

Yes, I have found some bugs in time management in V6. But not a disaster. See my next post, please. Only two relevant bugs in 40 games. In my opinion, not a disaster. Anyway, V6 is 60% faster than V5 but ... I have doubts about if it is stronger or not.

Tom.

Quad 2,67 4 Gz Ram 4 threads.
10 minutes per game.
Paradigm book.
All engines with 4 cores.


2010-01-IgorritV6-1 2010

Igorrit 0.086v6 w32 - Naum 4(x4) 7.0 - 3.0 +5/=4/-1 70.00%
Igorrit 0.086v6 w32 - Rybka 3 32-bit (x4) 6.5 - 3.5 +4/=5/-1 65.00%
Igorrit 0.086v6 w32 - Stockfish 1.6s 4.5 - 5.5 +0/=9/-1 45.00%
Igorrit 0.086v6 w32 - Deep Shredder 12 UCI 3.5 - 6.5 +0/=7/-3 35.00%

One loss against Rybka and another one against DS12 was because a crash in Igor. (Games 3997 and 4027). Without these bugs, the results would have been:

Igorrit 0.086v6 w32 - Naum 4(x4) 7.0 - 3.0 +5/=4/-1 70.00%
Igorrit 0.086v6 w32 - Rybka 3 32-bit (x4) 6.5 - 2.5 +4/=5/-0 71,50%
Igorrit 0.086v6 w32 - Stockfish 1.6s 4.5 - 5.5 +0/=9/-1 45.00%
Igorrit 0.086v6 w32 - Deep Shredder 12 UCI 3.5 - 5.5 +0/=7/-2 38,50%

In summary, brilliant results of Igor against the two strongest commercial engines: Rybka and Naum. Igor seems to have been tuned specially well against Rybka 3. But Deep Shredder seems to be the best anti-Igor engine in the market. Great performance of Stockfish, also. No one single loss agaist Igor!.

Games:

http://www.megaupload.com/?d=HIXEU9KD

Regards,

Tom.
[/b]

Quad i7 975 12 Gz. RAM
10 minutes game
Fritz 11 GUI
HS Book
Robbo : 1 thread. All other engines: 8 threads.

2010-01IgorritV6 (64w) (x8) 2010

Igorrit 0.086v6 x64 - Deep Shredder 12 x64 4.0 - 6.0 +1/-3/=6 40.00%
Igorrit 0.086v6 x64 - Stockfish 1.6s JA 64bit (x8) 6.0 - 4.0 +4/-2/=4 60.00%
Igorrit 0.086v6 x64 - RobboLito 0.09 x64 5.5 - 4.5 +2/-1/=7 55.00%
Igorrit 0.086v6 x64 - Rybka 3 (x8) 5.5 - 4.5 +2/-1/=7 55.00%

Four decisive time bugs for Igor. Games 3739, 3757, 3770 and 3773. 2 against Deep Shredder, 1 against Stockfish and 1 against Robbo. Results without these bugs:

2010-01IgorritV6 (64w) (x8) 2010

Igorrit 0.086v6 x64 - Deep Shredder 12 x64 4.0 - 4.0 +1/-1/=6 50.00%
Igorrit 0.086v6 x64 - Stockfish 1.6s JA 64bit (x8) 6.0 - 3.0 +4/-1/=4 67.00%
Igorrit 0.086v6 x64 - RobboLito 0.09 x64 5.5 - 3.5 +2/-0/=7 61,00%
Igorrit 0.086v6 x64 - Rybka 3 (x8) 5.5 - 4.5 +2/-1/=7 55.00%

My feeling is that besides the time bugs, v6 is not stronger than v5 using the same number of threads. The increase in the speed is, however, substantial, around 60%!.

Games:

http://www.megaupload.com/?d=ZSPFASJ6

Regards from Barcelona.

Tom

to clarify, v6 is faster on 8 threads but not on 4 or less? this would make sense as 8 thread support was added for v6. i hope that it is not significantly faster otherwise as i have found it is unreliable (hangs) in eng-eng matches, so i am using v5.

yanquis1972 wrote:to clarify, v6 is faster on 8 threads but not on 4 or less? this would make sense as 8 thread support was added for v6. i hope that it is not significantly faster otherwise as i have found it is unreliable (hangs) in eng-eng matches, so i am using v5.

You are right. In my computer V6 -8 threads- is about 60% faster than V5 -4 threads-. But V5 is not only more reliable -bugs free- but also slightly stronger, IMHO, than V6. I don't know the reason. I hope than V7 will overcome these problems...

Tom.