Is it possible to throttle back Houdini?

[Paul McKeown]

Thanks, Dave, but even setting the engine to run on one core only doesn't really solve the problem, as that core just becomes a hot spot.

The freeware program "Battle Encoder Shirase" does seem to be very satisfactory at the moment; it allows me to tune the rate at which the engine uses all the cores, so I can set it low whilst I am working on the PC and set it rather higher when I'm away, whilst completely avoiding the problem of the engine running the processor chip too hot.

[Dave Ewart]

Paul, good, glad you have a solution. For what it's worth, and I'm sure you realise this anyway, running at 100% all-cores CPU usage ought not to cause adverse effects: it does sound that, if this happens on your PC, there is a cooling fault of some sort or the cooling setup is simply not adequate. Agree that this would be a major faff to fix, though

[MikeTasker]

Don't know whether it's possible to burn a house down with a PC, but it's not an experiment I'm keen on trying out!

http://www.youtube.com/watch?v=i04p7nQZ ... re=related

[Paul McKeown]

Cleared out the dust from between the vanes of the processor chip cooling, seems to have done the trick. Now runs stable at 100% on all four cores without overheating. Seems to top out at a comfortable 68 degrees Celsius, PC noise a mere background hum, even after running Houdini for 3 hours. Glad I did the work on finding out how to limit the CPU quota, though, as I can still set a quota to 85% and leave the thing to run overnight without worrying.

[Nick Murphy]

If Houdini has 'UCI engine' configuration options, then you ought to be able to specify the maximum number of threads, i.e. CPU cores in this context. If you set that to be <4, then it won't use all your cores. Perhaps try setting it to 1?

Exactly. You can configure all multicore engines in their own settings to use as many cores as you like. Limit it to one core, and you should be fine.

You can set priority to "below normal" in the Fritz or ChessBase interface (though I think you said you did that already)

Also, you can set the priority to low in Windows task manager. This should have you running nice and cool while still getting GM level analysis

[Dave Ewart]

...you can set the priority to low ...

Technically, 'low priority' doesn't mean "use less CPU". If nothing else is running, low priority will still use 100% (or 400% for quad core), same as normal priority. Priority just affects behaviour relating to other demands on the CPU.

Using low priority is helpful if you don't want the analysis to take over the ability of other applications to get CPU time when they demand it. And as such, 'low priority' is usually a good idea in most circumstances.

[Paul McKeown]

You can set priority to "below normal" in the Fritz or ChessBase interface (though I think you said you did that already)

Also, you can set the priority to low in Windows task manager. This should have you running nice and cool while still getting GM level analysis

Setting the priority in the interface and setting it in the OS is exactly the same thing.

It pretty much misses the point altogether, though. It doesn't do anything at all, except set the level at which the process can be interrupted by other processes. What it certainly does not do is cap the utilisation of the CPU.

Sorry.

[Nick Murphy]

Upsettingly for me, you're right. I just re-read the thread and I did miss the point!

[Sean Hewitt]

Whenever I run Houdini (within Chessbase) it seems to crash everything else. I'm guessing it's eating memory - how do I rein it in?

According to ChessBase themselves, you give it a lower priority.

http://www.cruxis.com/chess/manual/inde ... estion.htm

Thanks. Worked perfectly.

[Mats Winther]

An average chess player can make do with a weaker engine than Houdini. When analyzing variants it is necessary to try out different alternatives, according to one's own preferences. One shouldn't just adopt the line that the engine shows. It must be tried out, together with other alternatives. So one should be able to find one's way with a weaker engine. I don't know why an amateur chess player would need Houdini. An amateur needs only learn the relevant variants in the ECO, plus some surprise lines which he has invented himself. When analyzing a game to understand why one lost, any engine will do. Concerning the unusual opening lines, not mentioned in the ECO, it is enough to see what title players have played before in these lines. Databases today contain several millions of games, so one can nearly always find solutions to one's problems there.

In fact, there are today online tree databases that do the job, like 365Chess.com and ChessTempo. However, these are commercial. Figsgames.com has a tree database of FICS games that is free, but rather slow. All games have an average rating above 2000, which are fairly high quality games. My point is that it's better to study games than inventing the wheel again.
M. Winther

[Paul McKeown]

If one selects standard games and average rating 2000-2200, one gets fairly good quality games.

Games played by players in that range are almost only interesting when they are getting ripped to shreds by much stronger players. The interest is in learning how to exploit their errors.

[Mats Winther]

If one selects standard games and average rating 2000-2200, one gets fairly good quality games.

Games played by players in that range are almost only interesting when they are getting ripped to shreds by much stronger players. The interest is in learning how to exploit their errors.

These are rapid chess ratings, typically 15 min per player, so many a 2100-2200 player could be an FM. Anyway, I gave the wrong interval. One can select "all" players, which means all over 2000 (lower rated games aren't stored), or one can select 2200-2400 or >2400.

It is useful for getting ideas about openings. Yesterday I met a player who positionally refuted the variant 1.e4 d6 2. d4 e5. I was convinced that this was playable, but now I know better.
M. Winther

[Dan O'Dowd]

If one selects standard games and average rating 2000-2200, one gets fairly good quality games.

Games played by players in that range are almost only interesting when they are getting ripped to shreds by much stronger players. The interest is in learning how to exploit their errors.

These are rapid chess ratings, typically 15 min per player, so many a 2100-2200 player could be an FM. Anyway, I gave the wrong interval. One can select "all" players, which means all over 2000 (lower rated games aren't stored), or one can select 2200-2400 or >2400.

It is useful for getting ideas about openings. Yesterday I met a player who positionally refuted the variant 1.e4 d6 2. d4 e5. I was convinced that this was playable, but now I know better.
M. Winther

Off-topic but would you post that game? I had a feeling myself that after 1. e4 e5 2. d4 d6 was unplayable and it'd be nice to see how it works.

[Roger de Coverly]

Off-topic but would you post that game? I had a feeling myself that after 1. e4 e5 2. d4 d6 was unplayable and it'd be nice to see how it works.

I cannot recall anyone writing a book on it, but there's over a thousand games where that position has been reached. Most transpose into lines of the Philidor as you would expect. The sequence 1. e4 d6 2. d4 e5 is more common as only a handful of players ever start 1. e4 e5 2. d4.

The sequence 3. Qxd8 Kxd8 is the most likely one to reach positions which cannot be reached from a normal Philidor.

[Matt Mackenzie]

ECO classifies 1e4 d6 2 d4 e5 3 de5 de5 4 Qd8 Kd8 and similar setups as B07 (which includes all Pirc lines without an early g6 and Bg7)