Houdini 2 tests

[img][You must be registered and logged in to see this link.][/img]

[img][You must be registered and logged in to see this link.][/img]
no books , no bases
games [You must be registered and logged in to see this link.]

[img][You must be registered and logged in to see this link.][/img]

[img][You must be registered and logged in to see this link.][/img]

games [You must be registered and logged in to see this link.]



thank you Venus

2 core zeemann 3.5final 1min blitzgames


houdini 2.0 prox64 +111 +4/-18/=33 65.50% 65.5/100

ivanhoe b47ex64 -111 +18/-49/=33 34,50% 34,5/100

with non pirated cracked engine you should get this sort of result



Pawn


Posts: 114
ELO: 125
Location: österreich
Warning Level:

[img][You must be registered and logged in to see this link.][/img][img][You must be registered and logged in to see this link.][/img]

This result from the "cracked" h2 pro , members are advised to test v h 1.5a at this time control

AMD Phenom(tm) II X6 1090T Processor x6 @3.80 GHz 8.00 MB Memory
Windows XP 64 Bit Professional Service Pack 2 (Build 3790)
Fritz Benchmark:
Speed: 28.33
KNS: 13599
GUI: CB Rybka 3
Book: Perfect 2011-10 moves
Hash: 256
GTB: ON
Ponder: OFF
Blitz:5' 0


1 Houdini 2.0 Pro x64 3195 +23/-15/=42 55.00% 44.0/80
2 Houdini 1.5a x64 3195 +15/-23/=42 45.00% 36.0/80






Intel(R) Core(TM)2 Quad Q9550 2.83GHz4x @3.5 GHz 4,096 MB Memory
Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)
Fritz Benchmark:
Speed: 20.90
KNS: 10032
GUI: CB Rybka 3
Book: Perfect 2011-10 moves
Hash: 256
TB and RB: ON
Ponder: OFF
Blitz:5' 0


1 Houdini 2.0 Pro x64 3195 +49/-36/=75 54.06% 86.5/160
2 Houdini 1.5a x64 3190 +36/-49/=75 45.94% 73.5/160







Intel(R) Core(TM) i7 4CPU 960 3.20 GHz x8 @ 4.00 GHz with 17.9 MB Memory
Microsoft Windows XP 64 Bit Professional Service Pack 2 (Build 3790)
Fritz Benchmark:
Speed: 24.98
KNS: 11988
GUI: CB Rybka 3
Book: Perfect 2011-10 moves
Hash: 256
GTB: ON
Ponder: OFF
Blitz:5' 0


1 Houdini 1.5a x64 3190 +19/-15/=46 52.50% 42.0/80
2 Houdini 2.0 Pro x64 3195 +15/-19/=46 47.50% 38.0/80






~~~

all tests above are by slankaman

seems even the "official " versions somewhat erratic

[img][You must be registered and logged in to see this link.][/img]

[img][You must be registered and logged in to see this link.][/img]
Awful result @7 minute t.c.


###########

but @ 10 minute t.c. ...


[img][You must be registered and logged in to see this link.][/img]

[img][You must be registered and logged in to see this link.][/img]

Both test on same hardware , 8 gig ram ,Fritz12 GUI windows 64
3.4 ghz quad , 512 engine hash, No books ,No bases , Ponder OFF

[img][You must be registered and logged in to see this link.]




[img][You must be registered and logged in to see this link.][/img]
games


[You must be registered and logged in to see this link.]

~~~~~

[You must be registered and logged in to see this link.]

games
[You must be registered and logged in to see this link.]



Huge difference here H1.5a v Dr4.1SSE42 H1.5a wins by average elo increase as found in most tests, whereas the h2.0Pro crushes it

for information re this this H20Pro please pm me

Intel(R) Core(TM) i7 Q- 960 3.20 GHz8x @ 4.00 GHz 18.00 MB Memory
Windows 7 Professional Professional (Build 7600)
Fritz Benchmarks:
Speed: 24.89
KNS: 11946
GUI: CB Rybka 3
Hash: 256
Book: Perfect 2011 -10 moves
GTB: ON
Ponder: OFF
Blitz:10' 0



1 Houdini 2.0 Pro x64 3195 +27/-14/=39 58.13% 46.5/80
2 Houdini 1.5a x64 3190 +14/-27/=39 41.88% 33.5/80



here slankaman tests at 10 minutes t.c. and gets similar results to ours

OK here is the test ive been running for days now.
I have a i7 980 cpu 6g ram. clocked at 4.3GHZ

hash tables set at 512
ponder off
Rybka 4 book
time control 5' +3"
6 cores vs 6 cores also with HT off

Rybka 4.1 special settings
White and Black pawns set at +10
White and Black Knights set at +25
No large pages enabled or NUMA

Results

[You must be registered and logged in to see this link.]

I plan on going all the way to 1200 games. Which will probably take me a week maybe more to get too as I only have the one pc to do this test with. But so far not really impressed with Houdini 2. and yes this is a bought engine not a hacked engine.

My free hosting link seems to keep expiring way too soon. Just in case

Houdini 2 pro +11 +82/-72/=174 51.67% 170/329 vs Rybka 4.1 special settings listed above.

329 games played thus far.

[You must be registered and logged in to see this link.] wrote:

OK here is the test ive been running for days now.
I have a i7 980 cpu 6g ram. clocked at 4.3GHZ

hash tables set at 512
ponder off
Rybka 4 book
time control 5' +3"
6 cores vs 6 cores also with HT off

Rybka 4.1 special settings
White and Black pawns set at +10
White and Black Knights set at +25
No large pages enabled or NUMA

Results



I plan on going all the way to 1200 games. Which will probably take me a week maybe more to get too as I only have the one pc to do this test with. But so far not really impressed with Houdini 2. and yes this is a bought engine not a hacked engine.

snob

you weren't foolish enough to pay for weep rybka as well were you

re run test using free Houdini 2 after please

Actually i find rybka performance surprisingly good as without those changes h2pro crushes it

good test

ps you should be getting Houdini 2.1 soon , h2 has a big bug in evaluation. and time management in h1.5a is better

Actually no i didnt buy Rybka 4. I did however buy Rybka 3 online. and they also sent me a CD by accident. So techincally i have an unopened Rybka 3 from a few years ago.

I think i got a better free imghosting to use. Looks like its working fine.

goldminer_999 wrote:

Interesting result reported on Rybka Forum (ironic, eh ?) this morning using the Silversuite:

[You must be registered and logged in to see this link.]

SilverSuite test Houdini 2.0 Pro vs. Houdini 1.5a with 40 min/game time controls on a decent hardware:

Houdini 2.0 Pro: 59.5
Houdini 1.5a: 40.5

This suggests that H20 is approx. 60 Elo stronger that H15a.

Test parameters:

CPU: Dual Xeon E5430 (8 x 2.66 GHz)
Hash: 512 MB hash for each engine
Operating system: 64 bit Vista
Large pages: no
Split depth: 14 (autotuned)
H20 speed: approx. 11,000 kN/s
Contempt: default (1)
Ponder: off
Test positions: Silversuite.pgn (50 positions, each played twice with colors reversed)
Book: Rybka 4 (limited to 15 moves)
Time controls: 40 min/game + 2 seconds/move
Tablebases: 5 men gaviota tables (both engines)
Attachment: SilverSuite_H2_vs_H15a.zip (368.7k)

on a very powerful pc that uses numaware architecture ....yes

very few have dual cpu motherboards with 2 Xeons , and why the fool used gvtbases for test beggars belief

on small , average or average high end pc's , in real tests ...no

interesting though

houdini 2 prox64 +107 +8/-2/=10 65.00% 13.0/20

fire2.2xtremex64 -1o7 +2/-8/=10 35.00% 7.0/20

houdini 2 pro-critter 1.2 2 core zeemann3.5 final 1 min games(20)


houdini 2.0 pro x64 +34 +5/-3/=12 55.00% 11.0/20

critter 1.2 x64 -34 +3/-5/=12 45.00% 9.0/20

Intel(R) Core(TM) i7CPU X980 3.33GHz 12x4250 GHz, 6GB Memory

Microsoft Windows Server 2008 R2

Fritz Benchmark:

Speed: 36

KNS: 17274

GUI: FRITZ12

Book: DeRyblmSt 5 and my own (only 16 moves) after 180 moves-award

Hash: 1024

GTB and RB: ON

PONDER:OFF

LARGE PAGE:ON

Blitz:8' 0



IvanhoeB47e (all triple bases-ON,totalbases-OFF, 6CORE, 12 threads, bulkloaddirectory--c:\robbotriplebases\5\Z)



Houdini 2 pro (gtb-ON, 6CORE)



Houdini 2 pro - IvanhoeB47e, Блиц 8м 0



1 Houdini 2.0 Pro x64.exe +4/=19/-1 56.25% 13.5/24

2 Ivanhoe B47e x64.exe +1/=19/-4 43.75% 10.5/24

Considering this test as on Huge 12 core numaware pc (IvanHoe does not support this architecture so is at big disadvantage) this is an extremely good result
For IvanHoe and those fortunate enough to have NEW RobBases of course

games [You must be registered and logged in to see this link.]

Nutzer-PC, Blitz:1' 0 2 core book zeemann3.5


1 Houdini 2.0 Pro x64 1½1½½½1½½1½½ 8.0/12
2 DeepSaros version 2.3h 0½0½½½0½½0½½ 4.0/12

houdini 2 pro x64 -firebird1.0 beta x64 1min 2 core zeemann 3.5


houdini2.0 pro x 64 +168 +11/-2/=7 72,5% 14.5 / 20

firebird 1.0 beta x64 -168 +2/-11/=7 27,5% 5.5/ 20

Testing is moving rather slowly. But here is my latest update from same testing as settings i have mentioned earlier in this post.

529 games played thus far.

Houdini 2 Pro. VS Rybka 4.1 SSE42 X64 +27 +150/-108/-269 53.98%


This is not very convicing of an engine that is supposed to be about 75 elo better overall. Houdini 2 is marginally better than Rybka 4.1 with special settings.

[You must be registered and logged in to see this link.] wrote:

Testing is moving rather slowly. But here is my latest update from same testing as settings i have mentioned earlier in this post.

529 games played thus far.

Houdini 2 Pro. VS Rybka 4.1 SSE42 X64 +27 +150/-108/-269 53.98%


This is not very convicing of an engine that is supposed to be about 75 elo better overall. Houdini 2 is marginally better than Rybka 4.1 with special settings.

houdini vs stockfish 2core 1min


houdini2.0 prox64 zeemann final3.5-------------+168 +11/-2/=7 72.5% 14.5/20

stockfish2.01 x64 Kevin update v. 1.0.ctg---------168 +2/-11/=7 27.5% 5,5/20

houdini -deepsaros



houdini 2.opro x64 +70 +9/-5/=6 60.00% 12.0/20

deepsaros 2.3se -70 +5/-9/=640.00% 8.0/ 20

Same test still on going. Same setup i mention in the first post. 7 980 cpu 6g ram. clocked at 4.3GHZ

hash tables set at 512
ponder off
Rybka 4 book
time control 5' +3"
6 cores vs 6 cores also with HT off

Rybka 4.1 special settings
White and Black pawns set at +10
White and Black Knights set at +25
No large pages enabled or NUMA

Current results

Houdini 2 Pro VS Rybka 4.1 SSE42 X64 +29elo +189/-135/=311 54.25% 635 games total


Not very impressive results against Rybka.

houdini 2 pro vs deep rypka 4 2core zeemann 3.5 final 1min games(25)



houdini 2.o pro x64 +99 +11/-4/=10 64.00% 16.0/25

deep rybka 4sse42 x64 -99 +4/-11/=10 36.00% 9.0/25

Venus sends


1196 games

Houdini 1.5a 64-bit 4CPU 3366
Houdini 2.0 64-bit 4CPU 3361


[You must be registered and logged in to see this link.]


It seems that there is no improvement on CCRL list.