I would appreciate recommendations from the experts for recommended SF5 (with Syzygy bases) parameters to be used on the following system for infinite analysis, averaging 10-20 minutes/ply used for correspondence chess:
Intel i7-4930k with 6 cores O/C to 4.4 GHz with 32 GB of RAM on Win 7 Pro, using large pages.
Currently using for SF5: hash = 4.096 GB
minimum split depth = 0
Ponder is checked
Syzygy Probe depth = 3
Syzygy Probe limit = 6
Threads = 6
Thanks in advance to the experts.
Recommended SF5 parameters?
Moderators: hgm, Rebel, chrisw
-
- Posts: 4498
- Joined: Fri Apr 21, 2006 4:19 pm
- Location: IASI - the historical capital of MOLDOVA
- Full name: SilvianR
Re: Recommended SF5 parameters?
cma6 wrote:I would appreciate recommendations from the experts for recommended SF5 (with Syzygy bases) parameters to be used on the following system for infinite analysis, averaging 10-20 minutes/ply used for correspondence chess:
Intel i7-4930k with 6 cores O/C to 4.4 GHz with 32 GB of RAM on Win 7 Pro, using large pages.
Currently using for SF5:
hash = 4.096 GB
minimum split depth = 0
Ponder is checked
Syzygy Probe depth = 3
Syzygy Probe limit = 6
Threads = 6
Thanks in advance to the experts.
Silvian R
-
- Posts: 6442
- Joined: Tue Jan 09, 2007 12:31 am
- Location: PA USA
- Full name: Louis Zulli
Re: Recommended SF5 parameters?
I'm not an expert, but why not use a larger hash setting? 8192 MB at least. Minimum Split Depth = 0 is a placeholder, which will be replaced internally by 4 on your system, which is probably fine.cma6 wrote:I would appreciate recommendations from the experts for recommended SF5 (with Syzygy bases) parameters to be used on the following system for infinite analysis, averaging 10-20 minutes/ply used for correspondence chess:
Intel i7-4930k with 6 cores O/C to 4.4 GHz with 32 GB of RAM on Win 7 Pro, using large pages.
Currently using for SF5: hash = 4.096 GB
minimum split depth = 0
Ponder is checked
Syzygy Probe depth = 3
Syzygy Probe limit = 6
Threads = 6
Thanks in advance to the experts.
-
- Posts: 4498
- Joined: Fri Apr 21, 2006 4:19 pm
- Location: IASI - the historical capital of MOLDOVA
- Full name: SilvianR
Re: Recommended SF5 parameters?
Was just a joke, man ! I'm a fan of "exotic" chess engines. Pulse, Cave, Sunfish, MyTeacheR & co. are my great interests in computer chess. Monsters like Stockfish aren't on the first place on my agenda !cma6 wrote:Very useful!
Have a nice day !
Silvian R
-
- Posts: 658
- Joined: Wed Mar 08, 2006 8:58 pm
Re: Recommended SF5 parameters?
Have you tried your proposed setting? I doubt it.
What about learning by doing?
As an senior EXPERT I would recommand:
Do not change hash size ( well chosen by developers)
minimum split depth = 0 (zero is always good)
Ponder is checked (most experts use POFF)
Syzygy Probe depth = 12 (choose number of processors)
Syzygy Probe limit = 0 (let SF5 decide what do do)
Threads = 36
Consider bying a xeon system.
regards
Bernhard
What about learning by doing?
As an senior EXPERT I would recommand:
Do not change hash size ( well chosen by developers)
minimum split depth = 0 (zero is always good)
Ponder is checked (most experts use POFF)
Syzygy Probe depth = 12 (choose number of processors)
Syzygy Probe limit = 0 (let SF5 decide what do do)
Threads = 36
Consider bying a xeon system.
regards
Bernhard
-
- Posts: 1334
- Joined: Sun Jul 17, 2011 11:14 am
Re: Recommended SF5 parameters?
You are quite clearly not an expert. Any expert knows that adding more threads than physical CPUs ends up giving you less performance.BBauer wrote:Have you tried your proposed setting? I doubt it.
What about learning by doing?
As an senior EXPERT I would recommand:
Do not change hash size ( well chosen by developers)
minimum split depth = 0 (zero is always good)
Ponder is checked (most experts use POFF)
Syzygy Probe depth = 12 (choose number of processors)
Syzygy Probe limit = 0 (let SF5 decide what do do)
Threads = 36
Consider bying a xeon system.
regards
Bernhard
I'd give Stockfish 16 gigabytes of hash, though you probably won't use it all.
Minimum split depth at 0 will correspond to 4.
For infinite analysis Ponder makes zero difference. It's used for engine-engine games where Ponder is whether the engine will think on its opponent's move.
Not familiar with Syzygy options, so 0 is probably fine.
Threads should be 6. Any more increases NPS, but not time to depth.
Matthew:out
Some believe in the almighty dollar.
I believe in the almighty printf statement.
I believe in the almighty printf statement.
-
- Posts: 4498
- Joined: Fri Apr 21, 2006 4:19 pm
- Location: IASI - the historical capital of MOLDOVA
- Full name: SilvianR
Re: WRONG ! (my limited contribution -:))) !
BBauer wrote:Have you tried your proposed setting? I doubt it.
What about learning by doing?
1.-As an senior EXPERT I would recommand:
Do not change hash size ( well chosen by developers)
minimum split depth = 0 (zero is always good)
Ponder is checked (most experts use POFF)
Syzygy Probe depth = 12 (choose number of processors)
2.-Syzygy Probe limit = 0 (let SF5 decide what do do)
Threads = 36
Consider bying a xeon system.
regards
Bernhard
1.-sEnIoR eXpErT ? WOW !
That's a new joke in this thread ?
2.-If you have the 6-piece tables, set the value of "SyzygyProbeLimit" to 6 (the default). If you only have the 5-piece table, set it to 5. Set the value of this option to 0 if you want to temporarily disable tablebase probing.
Hoping I was (a bit) useful !
Have a nice day !
Silvian R
-
- Posts: 1813
- Joined: Thu Sep 18, 2008 10:24 pm
Re: Recommended SF5 parameters?
There are so many things wrong with this post...BBauer wrote:
As an senior EXPERT I would recommand:
Do not change hash size ( well chosen by developers)
minimum split depth = 0 (zero is always good)
Ponder is checked (most experts use POFF)
Syzygy Probe depth = 12 (choose number of processors)
Syzygy Probe limit = 0 (let SF5 decide what do do)
Threads = 36
Consider bying a xeon system.
regards
Bernhard
-
- Posts: 142
- Joined: Mon Apr 16, 2012 7:23 pm
- Location: Istanbul, TURKEY
Re: Recommended SF5 parameters?
There is no recommended settings. Just change the values and try. You can't expect values work exactly same on different computers.