Building PC

[Rapakei]

Hello everyone,


Is it possible to merge 6-piece WDL and 7-piece WDL 1 folder?

Do you also want to combine DTZ 6 items and 7 items DTZ into 1 folder?

Which parts do I need to make the engine work faster in a short time.
My current PC calculates around 41 depth in approximately 4 to 5 hours.


I think the working memory has an important task for this, the cores less or just as important?

Regards, Raph

[Dann Corbit]

Hello everyone,


Is it possible to merge 6-piece WDL and 7-piece WDL 1 folder?

Do you also want to combine DTZ 6 items and 7 items DTZ into 1 folder?

Which parts do I need to make the engine work faster in a short time.
My current PC calculates around 41 depth in approximately 4 to 5 hours.


I think the working memory has an important task for this, the cores less or just as important?

Regards, Raph

All the tablebase files can be put into one folder.
I spread them across different folders and disk drives simply for organization, but even Nalimov with Syzygy with Scorpio bitbase in one folder is fine.
The speed will not be affected unless you have raid with striping or something. Even then, it won't be significant.
More memory means a bigger hash table. Very little change in strength.

One of my machines can get between 40 and 55 plies in an hour in opening positions.
It only cost $700.
But the soundproof box for it was about $800.
And it costs more than $200/month in electricity to run it.
It also generates so much heat, I cannot run it in the summer (air conditioner is not strong enough to compensate, and the family room becomes a furnace).

So I guess that every solution has drawbacks.

I strongly suggest studying the Ipman benchmarks:
http://www.ipmanchess.yolasite.com/amd- ... -bench.php

I am planning to buy a nice 7nm Epyc machine when they come out this summer, but then again, I'm criminally insane.

[Dann Corbit]

Here is a sample of analyzed positions from a run I am doing right now:

Code: Select all

r1bqkb1r/pp3ppp/2np1n2/2P1p3/8/B1P1P3/P2P1PPP/RN1QKBNR w KQkq - acd 47; acs 3600; bm cxd6; c3 "cxd6"; cce 89; ce 62; id "Orangutan.0875"; pm cxd6; pv cxd6; 
r1b1kb1r/pp1p1ppp/2n1pn2/3q4/2N5/B3PN2/P1PP1PPP/R2QKB1R w KQkq - acd 50; acs 3600; bm Bxf8; c3 "Bxf8"; cce 444; ce 62; id "Orangutan.0836"; pm Bxf8; pv Bxf8 Kxf8; 
r1bqkb1r/pp1p1ppp/2n2n2/2P1p3/8/B3P3/P1PP1PPP/RN1QKBNR w KQkq - acd 46; acs 3600; bm c3; c3 "c3"; cce 355; ce 59; id "Orangutan.0828"; pm c3; pv c3 Qa5; 
rnb1kbnr/pp1p1ppp/4p3/q1P5/8/4P3/P1PP1PPP/RNBQKBNR w KQkq - acd 46; acs 3600; bm Bb2; c3 "Nf3"; cce 0; ce 58; id "Orangutan.0520"; pm Bb2; pv Bb2 Qxc5 c4 Qb6 Qc2 Nc6 Be2 Qd8 Nf3 Nf6 O-O Be7 d4 d5 Nbd2 O-O Ne5 Bd7 Rfc1 Rc8 Qd3 Nxe5 dxe5 dxc4 Nxc4 Nd5 Nd6 Rxc1+ Rxc1 Bc6 Qd4 Qb6 Qd2 f6 Bg4 f5 Bf3 Qb4 Qxb4 Nxb4 Bxc6 Nxc6 Nxb7 Rb8 Rxc6 Rxb7 Bc1 g5 f3 Kg7 Kf2 Kf7 Ke2 g4 fxg4 fxg4; 
r1b1kb1r/pp1p1ppp/2n1pn2/q1P5/8/N3PN2/P1PP1PPP/R1BQKB1R w KQkq - acd 50; acs 3600; bm Nc4; c3 "Nc4"; cce 242; ce 58; id "Orangutan.0435"; pm Nc4; pv Nc4; 
1nb1kb1r/1pq1pp1p/2p2p2/2Pp4/1P6/4P3/3P1PPP/1N1QKBNR w Kk - acd 44; acs 3600; bm d4; c3 "d4"; cce 0; ce 56; id "Orangutan.0605"; pm d4; pv d4; 
rnb1k1nr/pp1p1ppp/8/q1b1p3/8/2N5/P1PPPPPP/1RBQKBNR w Kkq - acd 45; acs 3600; bm e3; c3 "Rb5"; cce 355; ce 56; id "Orangutan.0887"; pm e3; pv e3 d6; 

I'm holding back the really good stuff, though. I'm greedy like that.

[Leo]

Hello everyone,


Is it possible to merge 6-piece WDL and 7-piece WDL 1 folder?

Do you also want to combine DTZ 6 items and 7 items DTZ into 1 folder?

Which parts do I need to make the engine work faster in a short time.
My current PC calculates around 41 depth in approximately 4 to 5 hours.


I think the working memory has an important task for this, the cores less or just as important?

Regards, Raph

All the tablebase files can be put into one folder.
I spread them across different folders and disk drives simply for organization, but even Nalimov with Syzygy with Scorpio bitbase in one folder is fine.
The speed will not be affected unless you have raid with striping or something. Even then, it won't be significant.
More memory means a bigger hash table. Very little change in strength.

One of my machines can get between 40 and 55 plies in an hour in opening positions.
It only cost $700.
But the soundproof box for it was about $800.
And it costs more than $200/month in electricity to run it.
It also generates so much heat, I cannot run it in the summer (air conditioner is not strong enough to compensate, and the family room becomes a furnace).

So I guess that every solution has drawbacks.

I strongly suggest studying the Ipman benchmarks:
http://www.ipmanchess.yolasite.com/amd- ... -bench.php

I am planning to buy a nice 7nm Epyc machine when they come out this summer, but then again, I'm criminally insane.

Interesting. The truth comes out. I wondered what your electricity bill was. I am wondering what your going to spend on the epyc and how that will decrease your electric bill. Don't mean to be nosy.

[Dann Corbit]

Interesting. The truth comes out. I wondered what your electricity bill was. I am wondering what your going to spend on the epyc and how that will decrease your electric bill. Don't mean to be nosy.

I guess that the Epyc machine will be about $25000, but until I see prices of real systems, it's just a guess. (They price is Epyc, too).
I guess that the electricity bill will still be easily measurable, but a fraction of the cost of the current monstrosities.
I will probably replace the whole pile of machines with a single one.
On the other hand, I plan to put a stack of GPUs in it, so it might go right back up to "barn-burner" electricity hog.
But one of my sons does HVAC, so I might just upgrade my air conditioning so that I have a real server room.

[Leo]

Interesting. The truth comes out. I wondered what your electricity bill was. I am wondering what your going to spend on the epyc and how that will decrease your electric bill. Don't mean to be nosy.

I guess that the Epyc machine will be about $25000, but until I see prices of real systems, it's just a guess. (They price is Epyc, too).
I guess that the electricity bill will still be easily measurable, but a fraction of the cost of the current monstrosities.
I will probably replace the whole pile of machines with a single one.
On the other hand, I plan to put a stack of GPUs in it, so it might go right back up to "barn-burner" electricity hog.
But one of my sons does HVAC, so I might just upgrade my air conditioning so that I have a real server room.

Thanks. I am surprised your looking at buying a $20,000+ machine. Do what you have to do.

[Rapakei]

Hello everyone,


Is it possible to merge 6-piece WDL and 7-piece WDL 1 folder?

Do you also want to combine DTZ 6 items and 7 items DTZ into 1 folder?

Which parts do I need to make the engine work faster in a short time.
My current PC calculates around 41 depth in approximately 4 to 5 hours.


I think the working memory has an important task for this, the cores less or just as important?

Regards, Raph

All the tablebase files can be put into one folder.
I spread them across different folders and disk drives simply for organization, but even Nalimov with Syzygy with Scorpio bitbase in one folder is fine.
The speed will not be affected unless you have raid with striping or something. Even then, it won't be significant.
More memory means a bigger hash table. Very little change in strength.

One of my machines can get between 40 and 55 plies in an hour in opening positions.
It only cost $700.
But the soundproof box for it was about $800.
And it costs more than $200/month in electricity to run it.
It also generates so much heat, I cannot run it in the summer (air conditioner is not strong enough to compensate, and the family room becomes a furnace).

So I guess that every solution has drawbacks.

I strongly suggest studying the Ipman benchmarks:
http://www.ipmanchess.yolasite.com/amd- ... -bench.php

I am planning to buy a nice 7nm Epyc machine when they come out this summer, but then again, I'm criminally insane.

How do you do that, files tablebases in one folder?

[Dann Corbit]

it does not matter where you put them very much, when you use them, you jut tell the program where they are. i use separate folders, but that is just for organization,

[ouachita]

"I strongly suggest studying the Ipman benchmarks:
http://www.ipmanchess.yolasite.com/amd- ... -bench.php

I am planning to buy a nice 7nm Epyc machine when they come out this summer, but then again, I'm criminally insane.

How do you do that, files tablebases in one folder?"

For single CPU systems, the obvious winner for me at Ipman Chess is the Intel Xeon W-3175X, a 28-core overclockable beast

[Dann Corbit]

Dollars per node, the new Epyc chips win. And I expect the new threadrippers to really shine. I hope for October. I may just go Epyc, but it would be nice to have a quiet machine for once.