Fire 2.2 xTreme

Fire 2.2 xTreme

http://www.chesslogik.com/index.htm

any test ws houdini
?

not yet, just out....

prapqi wrote:

any test ws houdini
?

Guys-

this addresses a number of issues, but likely does not involve a substantial ELO gain.
perhaps a small one...but my limited testing was inconclusive.

i'm looking forward to your evaluations...

i have testet 8 min game houdini x
Fire 2.2 xTreme 1-0

samir70 wrote:

i have testet 8 min game houdini x
Fire 2.2 xTreme 1-0

well that's certainly encouraging!
thanks samir!

kranium wrote:

some might 'disapprove' of versions released so close together...to them i have a couple of things to say:
i will not 'conform' to the 'good-old-boy' establishment criteria of 1-2 releases per year. users and/or testers will not dictate how or when i release a new version.

if i have an improvement to the source code, no matter how insignificant, and i feel it will benefit the user, i may choose to release it. i do the work for free, i offer it for free, i have a right to release whenever and however i so choose. if you disagree with this, the answer is simple: don't download it.

best regards as always-

Keep the great work kranium and do your best to improve engine, no matter if new version so soon! :sm100:

Looks very good - an improvement IMO, thanks.

kranium wrote:

some might 'disapprove' of versions released so close together...to them i have a couple of things to say:
i will not 'conform' to the 'good-old-boy' establishment criteria of 1-2 releases per year. users and/or testers will not dictate how or when i release a new version.

if i have an improvement to the source code, no matter how insignificant, and i feel it will benefit the user, i may choose to release it. i do the work for free, i offer it for free, i have a right to release whenever and however i so choose. if you disagree with this, the answer is simple: don't download it.

Totally agreed. I think it's refreshing to see an engine-programmer (sorry, don't know the correct terminology) release updates/new versions regularly. It makes it so it's not a semi-annual one-hit-wonder, but a gradual improvement (which I would've thought allows for more accurate testing).

So keep up the good work man...

2.2 xTreme
PVHash now initializes correctly as LargePage memory TotalBaseHash size change issue fixed
TBHits also display in Fritz GUI
capability for non-LargePage compiles added
LowDepth razoring added -> UCI option Low Depth Margin
RobboInformatory removed, this was almost never used
Bulk Load Name, Bulk Unload Name, Bulk UnLoad Directory removed (unneeded and unused by most users)
this release includes Fire_22_xTreme_x64_NLP_PP...
NLP = No Large Pages, intended for systems (mostly Win XP) experiencing difficulty with Large Page memory configuration

http://www.chesslogik.com/fire.htm

A question for Norman:

We can try out so many settings..is great..but i find with Split depth you gain the most.
Everything depends who fast is your system.

I take Houdini as a example..Houdini has default split Depth=10
This is tested on a Q6600 4core system

I tried many games with different Split Depth(=SD) and on my Core i7 970 @4,155Ghz i get with SD=12
the best results!
I just have also a new laptop core i3 M380 @2,54Ghz 4Gb memory,Win 7 Professinal.
Played with the most strongest engine now each 120games..where default Houdini takes the lead before.
But i say SD=10 is to high for this laptop,he is much slower then the big system.
So,i reduced to SD=9 and he gained a nice Elo..and ofcourse the first place!

Now with Fire i have:
AN_Split_Depth=12
CN_Split_Depth=14
PV_Split_Depth=12

On what system is this tested as default and is it not possible the bring the minimum lower then 12!
Is the Split Depth in comparing with Houdini value?

When i know this,it gonne be easier to get more out off it..depending on what system people are testing is this SD important!

Ipman.

PS.I just checked the SD in Fire 2.0 there you can have min SD=1 !!
but in Fire 2.1 & 2.2 is min SD=12
Can you correct this,thanks!

Last edited by Ipman on Wed Jul 06, 2011 11:48 am; edited 1 time in total

hi Norman
please Fire 2.2 for 1 core (without SMP function) it is better and faster ? :D

Carlos wrote:

hi Norman
please Fire 2.2 for 1 core (without SMP function) it is better and faster ? :D

You have it if you simply set MaxThreads = 1

Ipman wrote:

A question for Norman:

We can try out so many settings..is great..but i find with Split depth you gain the most.
Everything depends who fast is your system.

I take Houdini as a example..Houdini has default split Depth=10
This is tested on a Q6600 4core system

I tried many games with different Split Depth(=SD) and on my Core i7 970 @4,155Ghz i get with SD=12
the best results!
I just have also a new laptop core i3 M380 @2,54Ghz 4Gb memory,Win 7 Professinal.
Played with the most strongest engine now each 120games..where default Houdini takes the lead before.
But i say SD=10 is to high for this laptop,he is much slower then the big system.
So,i reduced to SD=9 and he gained a nice Elo..and ofcourse the first place!

Now with Fire i have:
AN_Split_Depth=12
CN_Split_Depth=14
PV_Split_Depth=12

On what system is this tested as default and is it not possible the bring the minimum lower then 12!
Is the Split Depth in comparing with Houdini value?

When i know this,it gonne be easier to get more out off it..depending on what system people are testing is this SD important!

Ipman.

PS.I just checked the SD in Fire 2.0 there you can have min SD=1 !!
but in Fire 2.1 & 2.2 is min SD=12
Can you correct this,thanks!

Hi IPMan-
yes, good point...
i'll see what i can do in that regard.

Thx-
Norm

Dear Norman: a great news to meet Fire 2.2 xTreme fixed.
Thx a lot!!!

Ipman wrote:

A question for Norman:

We can try out so many settings..is great..but i find with Split depth you gain the most.
Everything depends who fast is your system.

I take Houdini as a example..Houdini has default split Depth=10
This is tested on a Q6600 4core system

I tried many games with different Split Depth(=SD) and on my Core i7 970 @4,155Ghz i get with SD=12
the best results!
I just have also a new laptop core i3 M380 @2,54Ghz 4Gb memory,Win 7 Professinal.
Played with the most strongest engine now each 120games..where default Houdini takes the lead before.
But i say SD=10 is to high for this laptop,he is much slower then the big system.
So,i reduced to SD=9 and he gained a nice Elo..and ofcourse the first place!

Now with Fire i have:
AN_Split_Depth=12
CN_Split_Depth=14
PV_Split_Depth=12

On what system is this tested as default and is it not possible the bring the minimum lower then 12!
Is the Split Depth in comparing with Houdini value?

When i know this,it gonne be easier to get more out off it..depending on what system people are testing is this SD important!

Ipman.

PS.I just checked the SD in Fire 2.0 there you can have min SD=1 !!
but in Fire 2.1 & 2.2 is min SD=12
Can you correct this,thanks!

I have discussed this question with Ipman.Maybe this explains the different scores on different computers.On Q9550 the score is the best and maybe SD is optimal for that speed and would have to be adjusted for other computers.
I will be going for a 10 day trip so I won't be able to test this problem with different SD but it is worth trying by other testers.

great engine!!!

robbolito wrote:

Ipman wrote:

A question for Norman:

We can try out so many settings..is great..but i find with Split depth you gain the most.
Everything depends who fast is your system.

I take Houdini as a example..Houdini has default split Depth=10
This is tested on a Q6600 4core system

I tried many games with different Split Depth(=SD) and on my Core i7 970 @4,155Ghz i get with SD=12
the best results!
I just have also a new laptop core i3 M380 @2,54Ghz 4Gb memory,Win 7 Professinal.
Played with the most strongest engine now each 120games..where default Houdini takes the lead before.
But i say SD=10 is to high for this laptop,he is much slower then the big system.
So,i reduced to SD=9 and he gained a nice Elo..and ofcourse the first place!

Now with Fire i have:
AN_Split_Depth=12
CN_Split_Depth=14
PV_Split_Depth=12

On what system is this tested as default and is it not possible the bring the minimum lower then 12!
Is the Split Depth in comparing with Houdini value?

When i know this,it gonne be easier to get more out off it..depending on what system people are testing is this SD important!

Ipman.

PS.I just checked the SD in Fire 2.0 there you can have min SD=1 !!
but in Fire 2.1 & 2.2 is min SD=12
Can you correct this,thanks!

I have discussed this question with Ipman.Maybe this explains the different scores on different computers.On Q9550 the score is the best and maybe SD is optimal for that speed and would have to be adjusted for other computers.
I will be going for a 10 day trip so I won't be able to test this problem with different SD but it is worth trying by other testers.

Yes, quite possible...Split Depth seems to be a very important parameter, especially on fast powerful systems.

cannot be implemented a test like in houdini to check optimal split depth?

kranium wrote:

Split Depth seems to be a very important parameter, especially on fast powerful systems.

+1

Tests are undergoing.
Thanks Kranium

very stable and same strength as Houdini1.5a
until now its draw between them,
long time controls (10+15) and both with TB hits.

Regards

Reda Menshawi

dear sir,

could you please elaborate on what Verbose_robbobases is?

thank you in adv.

stock