Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Aquarium / new PC and Aqaurium 2018
- - By Boombaard (**) [nl] Date 2017-12-09 11:59
Hi,

Yesterday I got my new PC with Ryzen 1700X, with 8 cores and 16Gb RAM. Now I try to install my new Aquarium 2018 and encounter some problems:

1.
What are the best settings in IDeA while running the latest Stockfish? Which version of SF do I need (I installed 'for modern computers')? How many threads do I address to SF? Using 8 gives only 50% CPU usage. How much RAM do I address for hastables? Aquarium gives 32 MB as default. And what is the best time for the Analysis Quality Setting? I always used 60 seconds, but is it right?

2.
How do I copy my IDEA projects from the old PC to the new PC where I installed Aquarium 2018?

3.
I also want to install Aquarium 2018 over Aquarium 2016 on my old PC. How to do that? Could'nt find a proper answer on this forum.

Thanks in advance for an answer!
Parent - - By pawnslinger (****) [us] Date 2017-12-09 16:26 Edited 2017-12-09 16:33
Hello:

Your setup sounds very much the same as mine.  R7-1700X and 16gb of main memory.  But you may not like my answers.  <grin>  I use the "popcnt" version of Stockfish... it is probably not the fastest, but it avoids instructions that are more Intel proprietary.

1. There are no "best" settings for IDeA.  Mine are always a work in-progress.  I try to use depth limited searches, instead of time limited.  Generally I set the time to 999 seconds max and the depth to anything between 23 and 30+ depending on the position.  Set it higher for simpler end-game positions... I try to tune it so I get around 1k evaluations per hour.

2. When I got my new PC, I used a thumb dirve and just copied the whole of the Aquarium folder structure.  I think Aquarium is almost totally portable in this manner.  Once over on the new hardware, I imported the projects thru the new Aquarium, which I installed there.

3.  I never install "over" my old Aquarium.  I always do a side-by-side upgrade, so that the full setup is always available in case I need to fallback to the old stuff.

Hope this helps.  It is totally subjective... just how I do things, probably not the "best" way.
Parent - - By Chestman (*) [us] Date 2017-12-09 16:48
Hi. What would you recommend as setting for a Queen and Pawn endgame, with possibilities friggin queen checks all over the place on the side with more pawns, for the side with more pawns. I am allowing queries to TB server.

I am not sure whether to go for WIDER, just to try to cover every position in the near future hoping to find a safe spot for the king, or narrower, much narrower, to do so.
Parent - By pawnslinger (****) [us] Date 2017-12-09 19:48
Every end-game is different, but I would try for higher depth at each evaluation step.  With really difficult end-games I have used 35 or higher.  Also, if IDeA gets really bogged down in an end-game, I turn to my copy of Chessbase and use my local 6-man TB... sometimes it is much more efficient than IDeA, and I go for depths with Chessbase as high as I can... 50+ is not out of the question with infinite analysis and a good TB in Chessbase.  If I do that, I like to see well over 1 million hits in the TB before I get satisfied.
Parent - - By Boombaard (**) [nl] Date 2017-12-09 17:02
Thank you for your answer. I will try it that way.
I still wonder what 'count number' you use for the engine setting in IDeA? 8 threads because 8 cores Ryzen, or 16 (or less?) because the total number of Kn/s seems to be higher then:
around 8 * 1.270 = 10.060 or
16 * 950 = 15.200 ?
Parent - - By pawnslinger (****) [us] Date 2017-12-09 19:43
I use multiple single threaded instances of Stockfish.  How many?  It varies.  Depends on lots of different factors.  The temperature in the room where my computer is located has a lot to do with it.  I closely monitor my cpu temp and try to keep it under 60C, right now I have 13 Stockfish copies running while I type this.  It is Winter here, so the temp is low.  During the Summer it is much different... I live in a hot desert and the temps are hard to control even with good air conditioning.  So I balance a lot of factors and try to run as many copies of Stockfish as I can.  I try to get 1k kNps from each instance, sometimes I get more, sometimes less, it depends on the position and how aggressive I am with my depth setting.  It is an art trying to keep everything balanced.  The other day I came in and found my cpu overheating, near 80C, because my bios had decided to turn off the cpu cooler fan.  So I have to keep a close reign on things to keep it flowing smoothly.
Parent - - By Boombaard (**) [nl] Date 2017-12-10 07:31
Ok, thank you . I have no heatproblems here, so I think 15 instances will do?
And can you tell how many MB I have to choose for hash memory (default it is 32 Mb in the engine settings)?
Parent - - By pawnslinger (****) [us] Date 2017-12-10 16:06 Edited 2017-12-10 16:14
Setting the hash value has not changed.  It is no different now than before.  You just have more memory available.  Generally speaking, the hash is used to give some historical memory to the engine, i.e. so that it does not need to recalculate the evaluation for the same position over and over again.  So you should make the hash as big as possible without running out of memory.  When I first got my Ryzen system I set it very high and soon saw that everything slowed to a snail's pace.  I had run out of physical memory due to my over eagerness.  So I adjusted trying to find some happy value that would give maximum effect from the hash memory, but not too much and things come to a crawl.

Hash has always been like this, you want it to be large enough to give the historical memory, but not sooooo large than things actually slow down, instead of speeding up.

I think I am currently using 128mb right now, but I am not sure that is the correct value (even for my setup).  Setting the proper value is a matter of experiment. 

Using 15 instances of Stockfish may be a little too high, you may not leave much compute power for Aquarium itself and Windows also needs some compute available.  If you leave only 1 core for those then again things may slow.  As things must be fit into a scarce resource.  I have actually run my system at 100 percent and Windows/Aquarium stop being responsive in a quick manner.  The heat in that case was not the factor that limited things, but how fast Windows responded to my keyboard/mouse, etc.  I had accidentally used 18 Stockfish cores.  this maxed out my system.  To maintain everything at the optimal flow, I rarely go over 13 Stockfish's.  It depends on how much needs to be left for these other processes, so things move efficiently.  Also, I frequently want to play other games while Aquarium is working in the background... things like Hoyle card games, etc.  In that case, I usually go with 12 Stockfish's, this leaves enough for less demanding things like card games or YouTube watching.  So everything needs a little slice of computer power or all must suffer.

I have my system setup so I can quickly and efficiently adjust my Stockfish count to allow for other things.  So it is always a continual balancing act.  I never deliberately go over 14 Stockfish.  But 12 and 13 are common settings.  Sometimes I use 14, sometimes less than 12 ... it all depends on other factors, like heat and what else needs compute power in my system.
Parent - By Boombaard (**) [nl] Date 2017-12-10 18:44
Thank you for your explanation.
Up Topic Rybka Support & Discussion / Aquarium / new PC and Aqaurium 2018

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill