The non beta version of FireBird 1.0 is now available from the following links. Will try to test it very soon!!
downloads: (Source: cyclonechess.dot.com)
FireBird 1.0 and 1.01 are available
FireBird 1.0 w32
FireBird 1.01 w32
(optimized windows 32-bit executable, logo, and readme files)
FireBird 1.0 x64
FireBird 1.01 x64
(optimized windows 64-bit executable, logo, and readme files)
FireBird 1.01 w32 no SSE2
(optimized windows 32-bit executable, logo, and readme files) for older CPUs that don't support Intel SSE2 instruction set
So here are some results for firebird Vs Igorrit Vs Rybka 3 (all 64 bit multicore versions)
ReplyDeleteTouerney Info ::::
30 games played / Tournament is finished
Time Control: Blitz 1 min + 2 sec/mov
Hardware: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz with 4,096 MB Memory
Operating system: Windows 7 Ultimate Professional (Build 7600)
Engine Score
1: FireBird_11_x64_ms_popcnt 13.5/20
2: Igorrit_0086v9_x64 9.5/20
3: Rybka 3 7.0/20
Result Table
-----------------FireBird_11_x64_ms_popcnt-----------------
FireBird_11_x64_ms_popcnt - Igorrit_0086v9_x64 : 6.0/10 2-0-8 (1========1) 60%
FireBird_11_x64_ms_popcnt - Rybka 3 : 7.5/10 6-1-3 (01111==1=1) 75%
-----------------Igorrit_0086v9_x64-----------------
Igorrit_0086v9_x64 - FireBird_11_x64_ms_popcnt : 4.0/10 0-2-8 (0========0) 40%
Igorrit_0086v9_x64 - Rybka 3 : 5.5/10 3-2-5 (=1==10=1=0) 55%
-----------------Rybka 3-----------------
Rybka 3 - FireBird_11_x64_ms_popcnt : 2.5/10 1-6-3 (10000==0=0) 25%
Rybka 3 - Igorrit_0086v9_x64 : 4.5/10 2-3-5 (=0==01=0=1) 45%
So overall firebird seems to be somewhat better in blitz time setting. I am expecting different result for longer time controls
Few more observations:::
ReplyDeleteRybka spawns 8 processes while other uses 8(or more) threads
Rybka's processor utilization is perfect to 100 percent, while rest both lacks in it somewhat(85-90%)
Midgame seems to be rather symmetric (may be cause of opening book of arena) and result only comes in endgames. Of course there were couple of exceptions but didnt see firebird doing any big mistake.
There werent any big blunders before endgames (sometimes firebird's pre-endgame analysis is crushing, it really is going upto 20-21 moves for me)
As the result shows there aint much difference in thinking for firebird and igorrit. Cant say firebird is much better but it has an edge over the other
Rybka castles too soon, firebird not so soon and igorrit castles mostly last.
I guess there are slightly different weights for bishops and nights. I doubt this but seemed to me like rybka prefers bishops.
Rybka took little bit more memory for hash than other. (I will configure it next time i run torney)
What remains is to see how firebird is doing it.
Hope I get source for it from somewhere.
Thanks for sharing. I totally agree the "midgame is rather systematic". Indeed this started from the opening. Now I started forcing engines to play from position in famous openings like sicilian najdorf, french advanced etc. Because chess engines tend to repeat the same evaluations, I run variety of engines matches (including weak engines) in long time control (normally 30 minutes) and just watch the games. This is very good and so far I learned a lot from this method.
ReplyDeleteYet have to see a miracle of rybka winning against firebird :P ... and I am talking about no blitz :P ... 40 mins per game .... and rybka blunders somewhere around move 60-70 when short on time. Firebird always keeps more time to itself.
ReplyDeleteAbout opening ... try olympiad book you have in arena, it gives many different openings (but not latest ones)
Branching factor of both rybka and firebird seems around 2.6. And if the numbers are not false, firebird has 10-15 times more processing power than rybka in terms of nodes visited. And in my opinion, thats outstanding ... i mean what the heck they did to have 10x improvement over rybka. THEY SURELY DESERVE THERE REWARD.
Another interesting obs is how firebird takes a lil more time to think in critical positions to look for a winning move. I didnt see such dynamic time setting for rybka (may be cause its playing against stronger player :P)
Lastly ... fire birds judgment of position seems consistent. I was guessing that they reduced branching factor to get more depth (which still might be true) but yet have to see firebird making a blunder cause of this.
Will contact you soon enuf. Keep up the good work ;) .. well clone or not ... i m ready to support firebird for any damn championship.
Dead download links!
ReplyDeleteClick on menu "DOWNLOAD ROBBOLITO ETC" at the top of this page for links
ReplyDeletewhats more powerful? 1.0 or 1.01?
ReplyDelete