Komodo 9 Announcement

[mjlef]

The main reason we have not added large pages is the whole memory fragmentation issue. Large pages requires contiguous memory, and people would have to reboot often to even run the program. Perhaps in the future if Windows uses a better memory management scheme to free up more contiguous memory we can look at doing this. But it is a bit hard to explain to people the need to reboot their computer just to run a chess program.

If anyone knows a solution to the fragmentation issue, let me know.

[mjlef]

Great you still keep working in Komodo, but I cannot avoid to tell you that at this level of playing force the 40 elo extra points are less desirable or important, at least for a player like me, than a new GUI.
There is something tired and even bothersome in just loading an engine in the very same gui, whatever it be. The field to gui expansion is huge. I am not just talking of the presentation in the screen, but of new, innovative services to make of a program a real chess partner, which today they are not.
I know I have said this same thing many times, without any effect, but I am a man of hope-...

Fern

Writing a GUI is a very different job than writing a chess engine.
Usually, a very different sort of programmer works on GUIs.
Typically, they need to be rather artistic.

That is very true. To give you some sense of my artistic skills, here is a screen shot of my old chess program NOW: http://i46.tinypic.com/2u5azvp.jpg

[syzygy]

Great news to see FRC and persistent hash. Are there plans to eventually offer a Komodo that supports large pages?

I understand that for engine tournaments, this might not be very useful since LP can't always be guaranteed, but the extra speedup of about 15% is useful for squeezing every drop out of hardware when doing analysis.

Running Komodo on Linux (which has transparent LP support) should do the trick.

[rcmaddox]

Slight disappointment in not packaging android with windows purchase.

The issue we started having with Android version is trying to support all the different versions of Android and different hardware in a single file. This coupled with the difficulty for users to transfer and install has made me decide to let the Google Play Store handle things. Plus it is under $5.

I can't get the bundled Android K8 version to even work on my Nexus 7, but the Google Play version works fine.

[lkaufman]

Komodo 9 price is set at $59.96, or $39.99 if you bought Komodo 8. If you want the one year subscription it will be $40 more in each case, the same as you would pay for one upgrade separately. So the total cost for the year will be the same as if you bought Komodo 9 plus one version six months later. The in-between versions will thus be free.

You're effectively promising that the next three upgrade prices will be fixed. I don't see why you would reduce your options that much. One reason people might have to buy the subscription is fear of a price rise, but you've removed that.

Sorry, I don't follow you. A subscription will cost only as much as Komodo 9 plus one upgrade, but you get three. That seems like plenty of incentive to me. Maybe we priced the subscription a bit lower than it really should be, but keeping the price below $100 seemed important.

[rcmaddox]

We have looked into doing an Amazon Android store version. It is definitely on the list. The programmer who did the Android version of Komodo is a bit under the weather, but I am sure he will do both as time permits. I am not sure when, but once we lock down Komodo 9 code and feels better he will work on it.

Maybe you could sell the .apk from your site and Kindle users could then sideload it.

Edit: Never mind, bad idea.

[lkaufman]

Can release any games that Komodo 9 has played?

Just look here for matches played by the Banks and by Clemens Keck with recent versions up thru 1415.00. 1416.00, which may become Komodo 9, is almost the same.

[lkaufman]

Great news to see FRC and persistent hash. Are there plans to eventually offer a Komodo that supports large pages?

I understand that for engine tournaments, this might not be very useful since LP can't always be guaranteed, but the extra speedup of about 15% is useful for squeezing every drop out of hardware when doing analysis.

There have been hardly any requests for this, and Mark doesn't favor it.

[Frank Quisinsky]

Hi Adam,

I hope you visit from time to time my result (TalkChess tourney selection) or on my boring site if you like.

I am using two normely i7 4th generation processors with Special cool Systems (works fine now all the time). Have a look in my selection (conditions), you can find pics from my systems. Using the fastest available RAM and a Special ASUS Board. All is installed on SSD ... I think the qualitiy of the games I produced and my opening book is very high. Not perfect for sure but very high.

Komodo 9 vs. Stockfish dev. from the same day ... both are sync for future testing ... this I will have and next week I start it. Very slow my Rating list will be better and better ... perhaps today I have the same high Level I produced with SWCR in the past. I am happy that many of the programmers have fun with this work. Not the great work CEGT or CCRL do but an additional good work I hope.

Thanks for your interest and have fun with my small work. Hope I can make the Komodo 9 release a very small bit more interesting with my test-runs.

Best
Frank

[Robert Flesher]

We plan to release Komodo 9 on Tuesday, April 28 if nothing goes wrong. It will be the current version, number 1416.00, unless we find a meaningful improvement or make UCI option changes in the next couple days. We are satisfied that it is as large an improvement over Komodo 8 as Komodo 8 was over Komodo 7.
Two new major features have been added. One is Persistent Hash, which is desirable for analysis. The other is that Komodo can now play Chess 960 (a.k.a. Fischerrandom). Maximum hash size has been raised to 64 Gb, and UCI options have been added to control the amount of reduction (lmr) and pruning (lmp). Although we believe the defaults are near-optimum for bullet chess, it is our hope that users will discover that for long time control games some other settings are better.
We will have more accurate estimates of elo gain as release date gets closer, but as of now we think the gain is more than 40 elo in single-core mode and 50 in multi-processor mode. As usual, gains come from all areas, including search, evaluation, and time management. Most noticeable should be the increase in search depth reached in a given amount of time, which is due to better extension, reduction, and pruning rules. We have done quite a bit to improve performance at blitz time controls, but we still anticipate that Komodo will earn its highest ratings against other engines at longer time controls.
We have decided to offer would-be purchasers the choice to simply buy Komodo 9, or to purchase a year's subscription to upgrades. We expect, but cannot promise, that upgrades will be quarterly, but we won't put out an upgrade unless we estimate we have gained at least ten elo and/or added a new feature. If we make a breakthru we might put out an upgrade sooner. The pricing is such that if you would normally buy new versions twice a year (roughly our past frequency of release) the extra upgrades will be free. In either case, those who bought Komodo 8 will get about a $20 discount. Subscription purchasers will also get older versions for free. All prices will be slid down by one version, so Komodo 6 will now be free.
We will no longer include the Android version with the Windows package, but it will be available soon from Google PlayStore for less than $5.

Larry and Mark for KomodoChess

Great news! I will purchase your very fine engine. Still missing Don, regards, a loyal customer forever.

Rob