Chess.com 2018 computer chess championship

[AndrewGrant]

Thanks. I don't understand the choice to enable pondering. The server has 48 physical cores. Each engine will run 46 threads. So with pondering on, 92 threads will be running on only 48 physical cores. Seems like a bad idea to me.

It's called 'hyperthreading' Louis, an Intel technology...each physical core = 2 logical cores
https://en.wikipedia.org/wiki/Hyper-threading

I'm quite familiar with hyperthreading. My concerns remain.

Your concerns are justified. I've lost two games in the last five years. I used HT for a few moves in both, and am convinced HT resulted in weak moves resulting is loss. I otherwise never use HT.

No, you are mistaken. The concerns in this thread are about the mixing of hyperthread and pondering.

For engines with proper CPU scaling support, hyperthreads will be an elo gain over non hyperthreads.

The following link shows this is the case when using Stockfish: http://tests.stockfishchess.org/tests/v ... 02b2e60162

[Uri Blass]

every time I try to write something in the CCCC chat I get the following messgae:

"This room is in 15 minutes followers-only mode. Follow computerchess to join the community!"

I tried typing Follow computerchess and got the same message and I guess nobody see my messages.

[Guenther]

every time I try to write something in the CCCC chat I get the following messgae:

"This room is in 15 minutes followers-only mode. Follow computerchess to join the community!"

I tried typing Follow computerchess and got the same message and I guess nobody see my messages.

You must follow the channel at twitch and after that you still have to wait 15 minutes.
How to follow a channel at twitch is explained on various sites, if you do a google search.

e.g. here
https://www.computerhope.com/issues/ch001866.htm#follow

[ouachita]

Thanks. I don't understand the choice to enable pondering. The server has 48 physical cores. Each engine will run 46 threads. So with pondering on, 92 threads will be running on only 48 physical cores. Seems like a bad idea to me.

It's called 'hyperthreading' Louis, an Intel technology...each physical core = 2 logical cores
https://en.wikipedia.org/wiki/Hyper-threading

I'm quite familiar with hyperthreading. My concerns remain.

Your concerns are justified. I've lost two games in the last five years. I used HT for a few moves in both, and am convinced HT resulted in weak moves resulting is loss. I otherwise never use HT.

No, you are mistaken. The concerns in this thread are about the mixing of hyperthread and pondering.

For engines with proper CPU scaling support, hyperthreads will be an elo gain over non hyperthreads.

The following link shows this is the case when using Stockfish: http://tests.stockfishchess.org/tests/v ... 02b2e60162

LOL, I'm not mistaken about why I lost these two games - HT.

[zullil]

Thanks. I don't understand the choice to enable pondering. The server has 48 physical cores. Each engine will run 46 threads. So with pondering on, 92 threads will be running on only 48 physical cores. Seems like a bad idea to me.

It's called 'hyperthreading' Louis, an Intel technology...each physical core = 2 logical cores
https://en.wikipedia.org/wiki/Hyper-threading

I'm quite familiar with hyperthreading. My concerns remain.

Your concerns are justified. I've lost two games in the last five years. I used HT for a few moves in both, and am convinced HT resulted in weak moves resulting is loss. I otherwise never use HT.

No, you are mistaken. The concerns in this thread are about the mixing of hyperthread and pondering.

For engines with proper CPU scaling support, hyperthreads will be an elo gain over non hyperthreads.

The following link shows this is the case when using Stockfish: http://tests.stockfishchess.org/tests/v ... 02b2e60162

LOL, I'm not mistaken about why I lost these two games - HT.

Further discussion would require additional information.

Operating system? Engine? Number of physical cores available? Number of virtual cores? Number of threads set for engine?

[AndrewGrant]

Thanks. I don't understand the choice to enable pondering. The server has 48 physical cores. Each engine will run 46 threads. So with pondering on, 92 threads will be running on only 48 physical cores. Seems like a bad idea to me.

It's called 'hyperthreading' Louis, an Intel technology...each physical core = 2 logical cores
https://en.wikipedia.org/wiki/Hyper-threading

I'm quite familiar with hyperthreading. My concerns remain.

Your concerns are justified. I've lost two games in the last five years. I used HT for a few moves in both, and am convinced HT resulted in weak moves resulting is loss. I otherwise never use HT.

No, you are mistaken. The concerns in this thread are about the mixing of hyperthread and pondering.

For engines with proper CPU scaling support, hyperthreads will be an elo gain over non hyperthreads.

The following link shows this is the case when using Stockfish: http://tests.stockfishchess.org/tests/v ... 02b2e60162

LOL, I'm not mistaken about why I lost these two games - HT.

There is a 100% chance I know more on the subject than you do.

[corres]

Hyper threads can use only the unused resources of the processor.
So it yields plus Elo only if the processor is not overclocked.
I think it would be good policy if they would be used dedicated threads: main threads for thinking and hyper threads for pondering only. Or one half of the whole threads for one of engines and the other half of the whole threads for the other engine.
But the best solution is using two computers that is they should play computer-computer matches instead of engine-engine matches. This happens when Leela plays against AB engines.
If engines together use the whole threads during search and pondering they can disturb the work of each other.

[Milos]

Hyper threads can use only the unused resources of the processor.
So it yields plus Elo only if the processor is not overclocked.

Lol, again you with those ridiculous claims. You don't even understand what overclocking actually means.
Overclocking and hyperthreading are two completely orthogonal ways to speed up CPU operations, if you even know what orthogonal means.

[corres]

Hyper threads can use only the unused resources of the processor.
So it yields plus Elo only if the processor is not overclocked.

Lol, again you with those ridiculous claims. You don't even understand what overclocking actually means.
Overclocking and hyperthreading are two completely orthogonal ways to speed up CPU operations, if you even know what orthogonal means.

Please. make some experiments about that instead of thinking on idealized situations.
Do you understand CPU with unlimited resources?
If yes, please divide that information with us.

[Milos]

Hyper threads can use only the unused resources of the processor.
So it yields plus Elo only if the processor is not overclocked.

Lol, again you with those ridiculous claims. You don't even understand what overclocking actually means.
Overclocking and hyperthreading are two completely orthogonal ways to speed up CPU operations, if you even know what orthogonal means.

Please. make some experiments about that instead of thinking on idealized situations.
Do you understand CPU with unlimited resources?
If yes, please divide that information with us.

To understand what I am talking about you need at least basic knowledge of modern computer architecture and how thermal management works.
Simply speaking HT by duplicating instruction fetching, caching and branching enables more efficient pipelining of execution units and more instructions being executed per clock cycle.
When you overclock you cpu, you simply increase the frequency of those clock cycles so they take less time to execute.
Having HT on, increases power consumption and heating of CPU therefore you might see that when CPU is overclocked you don't get the same benefit from HT than what you get when CPU is not overclocked. For example, while HT brings 30% more performance it increases power dissipation for 10% and increases throttling for 10% yielding positive net effect of lets say 20%. With overclocking HT still brings 30% but increases power dissipation for 20% yielding smaller positive net effect of lets say 10%. These numbers are just for illustration purposes so you can understand what I am talking about.