Page 5 of 5
Re: Komodo Syzygy evals
Posted: Sun May 25, 2014 12:00 am
by bob
I was responding to this:
I have sent you a private message to respect the forum rules
That reads, to me, expanded a bit:
I have sent you a private message telling you to respect the forum rules which you have violated...
It was my mistake, so my apologies...
Re: Komodo Syzygy evals
Posted: Sun May 25, 2014 12:20 am
by lkaufman
shrapnel wrote:lkaufman wrote:The setting of 2 is already the most aggressive; using 9 or 10 would mean fewer tb accesses. Those with very slow hard drives might want to use a higher value than 2, perhaps 6 would be reasonable. Values below 2 seem to just be bad.
Things are not so simple as you put it. It would be helpful if Ronald de Man posted his views on the matter here.
This topic has been discussed in relation to usage of Syzgy TBs by Stockfish on the Rybka Forum.
I should think that the same principles will apply to Komodo and Syzygy.
It is true that a Setting of 9 or 10 will result in fewer TB accesses, BUT, not by much !!..especially on fast SSDs !
This is because, no matter however fast the SSD there WILL be a definite, if slight, drop in kN/s speed with repeated ( and sometimes useless) TB Probes,especially early in the game !
In the end, one has to strike the right BALANCE, between the gain of ELO from TB usage and the DROP in ELO from kN/s drop.
Yes, I know that having a fast SSD will cause less drop in kN/s speed, BUT the drop in kN/s speed will definitely cause a drop in ELO
, over a period of time.
In the end, like I said, its all a matter of striking the right balance, especially in the case of SSD users !
Of course we tested this and concluded that 2 was best even without SSD. Maybe it's different for SF.
Re: Komodo Syzygy evals
Posted: Sun May 25, 2014 1:16 am
by NATIONAL12
lkaufman wrote:shrapnel wrote:lkaufman wrote:The setting of 2 is already the most aggressive; using 9 or 10 would mean fewer tb accesses. Those with very slow hard drives might want to use a higher value than 2, perhaps 6 would be reasonable. Values below 2 seem to just be bad.
Things are not so simple as you put it. It would be helpful if Ronald de Man posted his views on the matter here.
This topic has been discussed in relation to usage of Syzgy TBs by Stockfish on the Rybka Forum.
I should think that the same principles will apply to Komodo and Syzygy.
It is true that a Setting of 9 or 10 will result in fewer TB accesses, BUT, not by much !!..especially on fast SSDs !
This is because, no matter however fast the SSD there WILL be a definite, if slight, drop in kN/s speed with repeated ( and sometimes useless) TB Probes,especially early in the game !
In the end, one has to strike the right BALANCE, between the gain of ELO from TB usage and the DROP in ELO from kN/s drop.
Yes, I know that having a fast SSD will cause less drop in kN/s speed, BUT the drop in kN/s speed will definitely cause a drop in ELO
, over a period of time.
In the end, like I said, its all a matter of striking the right balance, especially in the case of SSD users !
Of course we tested this and concluded that 2 was best even without SSD. Maybe it's different for SF.
Wish I could buy it,but you do not recognize me as a returning customer.I bought TCEC.
Re: Komodo Syzygy evals
Posted: Sun May 25, 2014 5:02 am
by lkaufman
Let me know by pm if you don't get it by tomorrow.
Larry
Re: Komodo Syzygy evals
Posted: Mon May 26, 2014 11:23 am
by Hurnavich
No problem i have mild dyslexia
I should of wrote more clearly
I have sent you a private message (So that I respect the forum rules)
I was not sure that myself asking questions about a product in the forum rather than by private message would upset the mods and maybe cause me to break some rules possibly.
I have the greatest respect for Larry/Komodo team
I found the issues which now appear to be resolved,apologies to all for my impatience.
On another note Komodo 7a running in Aquarium IDEA on 6 instances is a fantastic studying engine.