Komodo 5.1r1 reporting K/ns info

[Hurnavich]

Hi, I notice the new updated Komodo 5.1r1 running in aquarium has a strange problem. The previous 5.1 showed constant speed output in k/ns eg 6000+ k/ns from start ,the updated 5.1r1 keeps displaying 0 kns to 6000+ then back to 0 then back to 6000+ as an eg in other words keeps jumping from 0 to actual k/ns. The previous 5.1 does not have this problem,but I have noticed the file is much bigger I ran task manager while the output was changing and it was running at full 100% all cores.

In Fritz 12 gui this is not a problem k/ns are normal not 0 to what ever

So has something changed in reporting K/ns in 5.1r1 causing a stall in output info?

Many thanks

[FWCC]

Now everyone back to the drawing board this NEW UPDATED BUG fix has to be tested on IPON and all other test sites.This update should do slightly better and maybe solidifying No.1 at LTC.








FWCC
CHESS FOR LIFE

[lkaufman]

Now everyone back to the drawing board this NEW UPDATED BUG fix has to be tested on IPON and all other test sites.This update should do slightly better and maybe solidifying No.1 at LTC.

Regarding whether to retest the bugfix version, here are my comments on the matter:
The biggest differences will affect two categories of testers (talking only about Windows version):

1. Those who experienced the memory leak problem. If you restarted the engine between games this would not have been a problem. It is my understanding that IPON was not affected by this, I don't know about other testing groups.
2. Those who test at very fast speeds. The new version is much better/faster at I/O, so anyone who tested at game in two minutes (or maybe 40 moves in 2 minutes) or faster is encouraged to retest. Tests run at super-fast speeds (like game in a few seconds) may show large gains for the bugfix version, but these gains have almost nothing to do with normal chess.
3. There are a couple fixes to specific rook endgames in the new version. Probably these are worth something like one elo

So in conclusion, we don't feel it is necessary or worth the time/effort to retest the new version at normal blitz levels (like 40/4 or game/5') or longer unless you observed or suspected the memory leak problem, but it is worth retesting at very fast levels.








FWCC
CHESS FOR LIFE

[FWCC]

Hello Larry,so you are saying that at faster levels,lets say 3+0 or 5+0 Komodo 5.1r1 will do better?Clemens is running an interesting match now Komodo 5.1 vs Houdini 3 at LTC of 60+10 and so far the results look promising for Komodo,it stands to reason that 5.1r1 should do even better than 5.1 and stand alone at LTCs.This is on a 4 cores.

http://www.clemens-keck.de//screen/






FWCC

[lkaufman]

Hello Larry,so you are saying that at faster levels,lets say 3+0 or 5+0 Komodo 5.1r1 will do better?Clemens is running an interesting match now Komodo 5.1 vs Houdini 3 at LTC of 60+10 and so far the results look promising for Komodo,it stands to reason that 5.1r1 should do even better than 5.1 and stand alone at LTCs.This is on a 4 cores.

http://www.clemens-keck.de//screen/






FWCC

Yes, I know about the Clemens match. Personally I have to believe that Komodo may be a bit lucky in that match so far, as I would be surprised if Komodo 5.1 were actually stronger than Houdini 3 on equal cores at any practical time limit. I'm actually running similar matches (at 60' plus 30" in my case) with the bugfix version against Houdini 3 on 4 and 12 cores, I'll post results when I have a decent number of games. It's possible that we've caught Houdini 3, but I think we need one major upgrade before I would bet on this. The bugfix version is probably only about one elo stronger than the original one at non-blitz time limits, which is of course insignificant.
My best guess is that at game/5' the improvement is in the 2 to 3 elo range, still pretty insignificant, and at game/3' maybe 4 or 5 elo. Hopefully we'll have some real data from people on this forum soon.

Larry

[CatPower]

I expect Robert to have enough elo improvement lying on the shelf and to release a new version once you get close.

[velmarin]

I expect Robert to have enough elo improvement lying on the shelf and to release a new version once you get close.

Not a bad idea,
when Komodo approaches, Houdini publishes new version.

Two months later, Komodo publishes new version.
Two months later, ect, ect.

There cake for two, or not ....

[Don]

I expect Robert to have enough elo improvement lying on the shelf and to release a new version once you get close.

It amazes me that you even feel threatened enough to make such a comment.

[Don]

I expect Robert to have enough elo improvement lying on the shelf and to release a new version once you get close.

Not a bad idea,
when Komodo approaches, Houdini publishes new version.

Two months later, Komodo publishes new version.
Two months later, ect, ect.

There cake for two, or not ....

My personal expectation is that Komodo will end up on top and it will last for some fairly limited amount of time since Larry and I are both aging. To me it seems like a sure thing.

But if it doesn't happen that way I will not lose any sleep over it. Life is too short to worry about that kind of stuff. We have people who will do our worrying for us and who get upset if some program passes another.

[CatPower]

It amazes me that you even feel threatened enough to make such a comment.

I don't feel threatened at all, I am the customer remember.
Seems to me you are projecting.