Question to Aart Bik

[hgm]

If you have two cores, you can already play ponder-on games. In theory the engines could affect each other by sharing higher-level caches and memory band-width, but if they are not actually designed to cheat by this, the effect they have on each other would hardly be noticeable. You could even do it on a single-core. (You would not get better play than in a ponder-off game, though, as this effectively reduces the thinking time by a factor 2.) It would not be worse then when you play multiple games concurrently (where engines playing in one game could also in theory affect engines in the other game, by sharing resources).

[Ovyron]

It would be worse because of the 40% of missed hits, and because in those instances, the hash will have been rewritten by useless info when compared to a hash of a ponder off engine.

So it'll always be better to use all your cores to play as many games as you can with Ponder OFF.

[Ras]

In theory the engines could affect each other by sharing higher-level caches and memory band-width.

There's also the turbo-boost feature in many CPUs which would kick in with singlethreaded engines, but maybe not if there are two heavy threads running.

If the computer is a laptop (or smartphone), there can be thermal issues in the long run which would force the CPU to throttle down. That would be particularly damaging with long time controls if only one of the engines is pondering.

[hgm]

So it'll always be better to use all your cores to play as many games as you can with Ponder OFF.

Indeed, like it is always 'better' to play several independent games in parallel than to run engines multi-threaded.

But what is better depends on your goals. Your goal seems to be to get the highest-level Chess for a given amount of CPU power. But that might not be other people's goal at all. For instance, measuring which engine has the better ponder implementation can also be a perfectly valid goal, and obviously you would have to play ponder-on games for that.

[Nordlandia]

Neat ⌛ device!

[hgm]

This is not for hourglass TC, but just a normal Chess clock (i.e. two timing devices, alternately running and stopped) based on hourglass technology. For hourglass TC you would only use a single hourglass.

Obvious problem is to reset the clocks for a new game...

[Nordlandia]

H.G.Muller: you're right. It's only based on hourglass TC but function like a normal clock.

If the device can be modified to use only one hourglass, now we're talking!

Here is the blueprint for that device ->

http://www.coroflot.com/ericchiang/Temp ... t-IDS-West

[hgm]

Well, you hardly need a device for that. An hourglass is the device. You just turn it over at the end of your move. That is how it was originally done. Making a device just to turn something upside-down seems like over-doing it.

[Nordlandia]

Well, you hardly need a device for that. An hourglass is the device. You just turn it over at the end of your move. That is how it was originally done. Making a device just to turn something upside-down seems like over-doing it.

I bet majority is so used to regular chess clocks that almost nobody want to bother turn it upside down during each move. A device has to rotate it by a lever, particularly because of simplicity. If you keep rotating it manually, at one point your hourglass will eventually fall to the floor and possible break. Therefore the HG need to be stationary or static etc...

[hgm]

On a device to flip an hourglass both players could press the same button!