And the king of tactics is .....

[Tony Thomas]

Not everyone got an 8 core Michael. Why dont you test it with 2 cpus Michael?

[MikeB]

Not everyone got an 8 core Michael. Why dont you test it with 2 cpus Michael?

I'm sure there's plenty of people who run it with 2 cpus, but I don't really have an interest in driving in 1st gear when i have a 5th gear I can go to...

I should just cool it. I'm just a little bit exuberant having 8 cores , that's all- it's awesome ...at my age I don't get exhuberant very often, but I can see how it might be annoying to people..sorry

[Dann Corbit]

Not everyone got an 8 core Michael. Why dont you test it with 2 cpus Michael?

I'm sure there's plenty of people who run it with 2 cpus, but I don't really have an interest in driving in 1st gear when i have a 5th gear I can go to...

I should just cool it. I'm just a little bit exuberant having 8 cores , that's all- it's awesome ...at my age I don't get exhuberant very often, but I can see how it might be annoying to people..sorry

You should continue to pound the universe into submission with the full 8 barrels blazing.

It might be nice to give us a step by step description of what it took to build your box. If we see things like "Reprogram the EPROM" and "Liquid Helium pump" then we can know we're just going to have to wait.

[Anabolic Karpov]

Hey Leto, I don't want to rain on your excitement with your 8 core but I think there is a fundamental flaw in your tests! You have overlooked the inherent variability from an mp engine and I guess if you ran all those tests again then you may well find you get a completely different set of results, except for the single CPU engines.

The only engines likely to give repeatable results are those running on just one core. So sorry, but you need to switch off seven of your cores and then you will get repeatable results.

[Marc MP]

On my Athlon 1.4gzh with 48M hash and 20 sec. per move, Chess Tiger 2007 Gambit scored 142/191 for Arasan6 and 152/183 for Ecmgcp good for a total of: 294/374 = 78,6%.

Is there a way to check if there are duplicates among epd test sets?

[Allard Siemelink]

I ran Crafty on my machine and updated the table.

Crafty 21.6 (8 cpus)
search time = 10 seconds , dual Xeon 5310, o/c 1.9 Ghz , 8 cores.
hash 768M

ecmgcp 10 sec max
1. Rybka WinFinder 2.2 mp scored 179 out of 183, 97.8%, 0.83s/5.34
2. Hiarcs 11.1 mp scored 165 out of 183, 90.1%, 1.32s/10.15
3. Crafty 21.6 scored 161 out of 183, 88.0%

Just ran bright-0.1d against the ecmgcp test: it also scores 161
(single core, 64Mb hash, AMD64 32-bit 2500Mhz)

[Leto]

Hey Leto, I don't want to rain on your excitement with your 8 core but I think there is a fundamental flaw in your tests! You have overlooked the inherent variability from an mp engine and I guess if you ran all those tests again then you may well find you get a completely different set of results, except for the single CPU engines.

The only engines likely to give repeatable results are those running on just one core. So sorry, but you need to switch off seven of your cores and then you will get repeatable results.

I wish I had 8 cores, I only have 2. I think you're confusing me with Michael.

btw Michael, enjoy it while you still can, the way technology is advancing, 32 cores will be the norm soon.

[kwlakey]

Hi there Leto.

I am a newbie and I have a couple of questions.

What is "CEGT Blitz testing?"

What does "mp" stand for?

What is "ecmgcp?"

What is "arasan6?"

Thanks for any explaination you can offer.

kerry

[Leto]

I decided to check out whether the new Rybka has an improved tactical ability, and it turns out that it does. It appears that it improved by about 4%.

Here's the new updated table:
1. Rybka WinFinder 2.2 mp scored 355 out of 374, 94.9%, 1.14s/6.16
2. Hiarcs 11.1 mp scored 329 out of 374, 87.9%, 1.72s/10.22
3. Deep Fritz 10 scored 320 out of 374, 85.5%, 1.57s/10.48
4. Zap!Chess Zanzibar x64 mp scored 312 out of 374, 83.4%, 1.78s/3.5
5. Fritz 10 scored 312 out of 374, 83.4%, 1.78s/10.22
6. Fritz 9 scored 310 out of 374, 82.8%, 1.77s/9.02
7. Rybka 2.3.2 mp scored 309 out of 374, 82.6%, 1.89s/9.01
8. Deep Junior 10 scored 299 out of 374, 79.9%, 1.53s/10.76
9. Rybka 2.2 mp scored 293 out of 374, 78.3%, 2.16s/8.63
10. Deep Shredder 10 x64 mp scored 289 out of 374, 77.2%, 3.35s/9.22
11. Arion 1.7 scored 218 out of 374, 58.2%, 2.74s/9.34

ecmgcp 10 sec max
1. Rybka WinFinder 2.2 mp scored 179 out of 183, 97.8%, 0.83s/5.34
2. Hiarcs 11.1 mp scored 165 out of 183, 90.1%, 1.32s/10.15
3. Deep Junior 10 scored 160 out of 183, 87.4%, 1.19s/10.65
4. Deep Fritz 10 scored 160 out of 183, 87.4%, 1.31s/10.21
5. Zap!Chess Zanzibar x64 mp scored 159 out of 183, 86.8%, 1.56s/2.77
6. Fritz 10 scored 156 out of 183, 85.2%, 1.56/10.05
7. Fritz 9 scored 155 out of 183, 84.6%, 1.49s/8.84
8. Rybka 2.3.2 mp scored 152 out of 183, 83%, 1.62s/8.86
9. Deep Shredder 10 x64 mp scored 151 out of 183, 82.5%, 3.16s/9.11
10. Rybka 2.2 mp scored 144 out of 183, 78.6%, 1.97s/8.61
11. Arion 1.7 scored 129 out of 183, 70.4%, 2.13s/8.99

arasan6 10 sec max
1. Rybka WinFinder 2.2 mp scored 176 out of 191, 92.1%, 1.46s/6.98
2. Hiarcs 11.1 mp scored 164 out of 191, 85.8%, 2.13s/10.29
3. Deep Fritz 10 scored 160 out of 191, 83.7%, 1.83s/10.76
4. Rybka 2.3.2 mp scored 157 out of 191, 82.1%, 2.15s/9.15
4. Fritz 10 scored 156 out of 191, 81.6%, 2.01s/10.39
5. Fritz 9 scored 155 out of 191, 81.8% 2.05s/9.21
6. Zap!Chess Zanzibar x64 mp scored 153 out of 191, 80.1%, 2.01s/4.26
7. Rybka 2.2 mp scored 149 out of 191, 78%, 2.35s/8.65
8. Deep Junior 10 scored 139 out of 191, 72.7%, 1.88s/10.87
9. Deep Shredder 10 x64 mp scored 138 out of 191, 72.2%, 3.54s/9.33
10. Arion 1.7 scored 89 out of 191, 46.5%, 3.72s/9.70

[Leto]

Hi there Leto.

I am a newbie and I have a couple of questions.

What is "CEGT Blitz testing?"

What does "mp" stand for?

What is "ecmgcp?"

What is "arasan6?"

Thanks for any explaination you can offer.

kerry

CEGT Blitz: http://www.husvankempen.de/nunn/40_4_Ra ... liste.html

mp=multiprocessor

ecmgcp and arasan6 are tactical test suites, you can download them here: http://www.arasanchess.org/tests.html

If you have a Fritz or other GUI from Chessbase you can go to Tools ----> Analysis -----> Process Test Set

The settings I used are 10 max secs, 1 extra ply