Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
smatovic
Posts: 903
Joined: Wed Mar 10, 2010 9:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by smatovic » Tue Jun 25, 2019 11:25 am

Dokterchen wrote:
Tue Jun 25, 2019 11:01 am
Hello Srdja,

I have a RTX 2070 and maybe I did someting wrong but Zeta crashed during the benchmark:
...
KR
Torsten
Hello Torsten,

thanks for testing, not 100% sure, it may be a bug in my program, but the search
is running only for a second+, so this could be a sign for OS or driver depended
gpu timeout.

The .reg file from my repository solved on my Windows 7 Pro a 2 second gpu
timeout issue for me.

https://raw.githubusercontent.com/smato ... tTo20s.reg

Download, double-click and reboot OS. Maybe this is a solution.

Afaik LC0 works differently, it does many small batches per second, so it is no
subject to timeouts like Zeta is.

--
Srdja

User avatar
Guenther
Posts: 3040
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by Guenther » Tue Jun 25, 2019 11:32 am

smatovic wrote:
Tue Jun 25, 2019 10:35 am
Guenther wrote:
Tue Jun 25, 2019 10:32 am
If you like me to change the country entry to Montenegro it is up to you - I think that flag is still missing :)
Okay, then go for it :-)

--
Srdja
Done (found 2017-04-11 for 0.99a which is now the start entry).
Also added the direct link as workaround for browser problems with the google sheets display.
Current foe list count : [97]
http://rwbc-chess.de/chronology.htm

smatovic
Posts: 903
Joined: Wed Mar 10, 2010 9:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by smatovic » Tue Jun 25, 2019 11:39 am

smatovic wrote:
Tue Jun 25, 2019 11:25 am
Dokterchen wrote:
Tue Jun 25, 2019 11:01 am
Hello Srdja,

I have a RTX 2070 and maybe I did someting wrong but Zeta crashed during the benchmark:
...
KR
Torsten
Hello Torsten,

thanks for testing, not 100% sure, it may be a bug in my program, but the search
is running only for a second+, so this could be a sign for OS or driver depended
gpu timeout.

The .reg file from my repository solved on my Windows 7 Pro a 2 second gpu
timeout issue for me.

https://raw.githubusercontent.com/smato ... tTo20s.reg

Download, double-click and reboot OS. Maybe this is a solution.

Afaik LC0 works differently, it does many small batches per second, so it is no
subject to timeouts like Zeta is.

--
Srdja
Just realized that even an increased 20 seconds timeout will not be sufficient
for the benchsmp command to succeed...

here an mod with 200 seconds timeout:

https://zeta-chess.app26.de/downloads/S ... To200s.reg

--
Srdja

Dokterchen
Posts: 104
Joined: Wed Aug 15, 2007 10:18 am
Location: Munich

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by Dokterchen » Tue Jun 25, 2019 12:06 pm

smatovic wrote:
Tue Jun 25, 2019 11:39 am
smatovic wrote:
Tue Jun 25, 2019 11:25 am
Dokterchen wrote:
Tue Jun 25, 2019 11:01 am
Hello Srdja,

I have a RTX 2070 and maybe I did someting wrong but Zeta crashed during the benchmark:
...
KR
Torsten
Hello Torsten,

thanks for testing, not 100% sure, it may be a bug in my program, but the search
is running only for a second+, so this could be a sign for OS or driver depended
gpu timeout.

The .reg file from my repository solved on my Windows 7 Pro a 2 second gpu
timeout issue for me.

https://raw.githubusercontent.com/smato ... tTo20s.reg

Download, double-click and reboot OS. Maybe this is a solution.

Afaik LC0 works differently, it does many small batches per second, so it is no
subject to timeouts like Zeta is.

--
Srdja
Just realized that even an increased 20 seconds timeout will not be sufficient
for the benchsmp command to succeed...

here an mod with 200 seconds timeout:

https://zeta-chess.app26.de/downloads/S ... To200s.reg

--
Srdja
OK, that fixed the problem:

Code: Select all

[2019-06-25 13:56:15] zeta.exe -l 
[2019-06-25 13:56:15] #> Zeta 099l
[2019-06-25 13:56:15] #> Experimental chess engine written in OpenCL.
[2019-06-25 13:56:15] #> Copyright (C) 2011-2019 Srdja Matovic, Montenegro
[2019-06-25 13:56:15] #> This is free software, licensed under GPL >= v2
[2019-06-25 13:56:15] #> eninge is initialising...
[2019-06-25 13:56:15] feature done=0
[2019-06-25 13:56:16] #> ...finished basic inits.
[2019-06-25 13:56:22] >> new
[2019-06-25 13:56:24] #fen: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
[2019-06-25 13:56:24] ###ABCDEFGH###
[2019-06-25 13:56:24] #8 rnbqkbnr
[2019-06-25 13:56:24] #7 pppppppp
[2019-06-25 13:56:24] #6 --------
[2019-06-25 13:56:24] #5 --------
[2019-06-25 13:56:24] #4 --------
[2019-06-25 13:56:24] #3 --------
[2019-06-25 13:56:24] #2 PPPPPPPP
[2019-06-25 13:56:24] #1 RNBQKBNR
[2019-06-25 13:56:24] ###ABCDEFGH###
[2019-06-25 13:56:31] >> sd 12
[2019-06-25 13:56:38] >> st 2000
[2019-06-25 13:56:43] >> benchsmp
[2019-06-25 13:56:43] ### doing inits for benchsmp depth 12: ###
[2019-06-25 13:56:44] ### computing benchsmp depth 12: ###
[2019-06-25 13:56:44] ### work-groups: 1 ###
[2019-06-25 13:56:44] depth score time nodes pv 
[2019-06-25 13:56:45] 1 60 56 20 b1c3 
[2019-06-25 13:56:45] 2 0 56 79 b1c3 b8c6 
[2019-06-25 13:56:45] 3 60 56 604 b1c3 b8c6 g1f3 
[2019-06-25 13:56:45] 4 0 57 1222 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 13:56:45] 5 27 61 3652 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 13:56:45] 6 0 82 20044 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 13:56:45] 7 22 104 40822 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 13:56:46] 8 12 197 123932 b1c3 b8c6 d2d4 g8f6 d4d5 c6e5 g1f3 e5f3 
[2019-06-25 13:56:49] 9 16 473 378077 b1c3 g8f6 g1f3 g7g6 d2d4 d7d6 c1f4 b8c6 e2e3 
[2019-06-25 13:57:07] 10 15 2286 2051730 b1c3 g8f6 g1f3 g7g6 e2e4 b8c6 d2d3 d7d5 c1g5 c8e6 
[2019-06-25 13:57:33] 11 12 4870 4444785 b1c3 g8f6 g1f3 b8c6 e2e3 g7g6 b2b3 e7e6 g2g3 f8b4 f1d3 
[2019-06-25 13:58:48] 12 7 12315 11355568 b1c3 b8c6 g1f3 e7e6 e2e3 g7g6 f1b5 g8f6 g2g3 f8d6 b2b3 d6b4 
[2019-06-25 13:58:48] #11355568 searched nodes in 123.153000 seconds, with 231876 ttmovehits, and 168514 ttscorehits, 1776 iidhits, ebf: 3.489059, nps: 92206  
[2019-06-25 13:58:48] ### doing inits for benchsmp depth 12: ###
[2019-06-25 13:58:49] ### computing benchsmp depth 12: ###
[2019-06-25 13:58:49] ### work-groups: 2 ###
[2019-06-25 13:58:49] depth score time nodes pv 
[2019-06-25 13:58:50] 1 60 57 40 b1c3 
[2019-06-25 13:58:50] 2 0 57 157 b1c3 b8c6 
[2019-06-25 13:58:50] 3 60 57 780 b1c3 b8c6 g1f3 
[2019-06-25 13:58:50] 4 0 59 1544 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 13:58:50] 5 27 61 4253 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 13:58:50] 6 0 68 19539 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 13:58:50] 7 22 90 56697 g1f3 b8c6 d2d3 g8f6 b1c3 d7d6 b2b3 
[2019-06-25 13:58:50] 8 3 142 151140 g1f3 b8c6 d2d4 g8f6 d4d5 c6b4 b1c3 d7d6 
[2019-06-25 13:58:52] 9 20 283 409732 b1c3 b8c6 g1f3 g8f6 d2d3 e7e6 c1e3 b7b6 g2g3 
[2019-06-25 13:59:01] 10 9 1206 2124285 g1f3 b8c6 b1c3 g8f6 d2d4 d7d5 e2e3 e7e6 b2b3 b7b6 
[2019-06-25 13:59:16] 11 12 2656 4819600 e2e3 b8c6 g1f3 g8f6 b1c3 e7e6 b2b3 f8b4 g2g3 g7g6 f1c4 
[2019-06-25 13:59:48] 12 7 5908 10860490 e2e3 b8c6 g1f3 g8f6 b1c3 e7e6 g2g3 g7g6 f1b5 f8d6 b2b3 d6b4 
[2019-06-25 13:59:48] #10860490 searched nodes in 59.082000 seconds, with 222534 ttmovehits, and 159635 ttscorehits, 1478 iidhits, ebf: 3.477116, nps: 183820  
[2019-06-25 13:59:48] ### doing inits for benchsmp depth 12: ###
[2019-06-25 13:59:49] ### computing benchsmp depth 12: ###
[2019-06-25 13:59:49] ### work-groups: 4 ###
[2019-06-25 13:59:49] depth score time nodes pv 
[2019-06-25 13:59:50] 1 60 53 80 b1c3 
[2019-06-25 13:59:50] 2 0 53 315 b1c3 b8c6 
[2019-06-25 13:59:50] 3 60 54 1148 b1c3 b8c6 g1f3 
[2019-06-25 13:59:50] 4 0 54 2243 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 13:59:50] 5 27 56 5307 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 13:59:50] 6 0 61 24805 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 13:59:50] 7 22 75 75237 d2d3 b8c6 g1f3 g8f6 b1c3 d7d6 b2b3 
[2019-06-25 13:59:51] 8 10 118 232802 b1c3 b8c6 d2d4 e7e6 d4d5 e6d5 g1f3 g8f6 
[2019-06-25 13:59:51] 9 16 178 453134 b1c3 e7e6 g1f3 g8f6 e2e3 g7g6 f1b5 b8c6 b2b3 
[2019-06-25 13:59:56] 10 9 687 2354386 b1c3 b8c6 d2d4 g8f6 g1f3 e7e6 e2e4 b7b6 e4e5 f6g4 
[2019-06-25 14:00:04] 11 12 1431 5159048 b1c3 d7d5 d2d4 g8f6 c1e3 b8c6 g1f3 e7e6 g2g3 f8b4 d1d3 
[2019-06-25 14:00:26] 12 12 3676 13630325 e2e3 b8c6 g1f3 g8f6 b1c3 e7e6 g2g3 f8b4 f1d3 g7g6 a2a3 b4d6 
[2019-06-25 14:00:26] #13630325 searched nodes in 36.762000 seconds, with 288459 ttmovehits, and 203160 ttscorehits, 1965 iidhits, ebf: 3.538410, nps: 370772  
[2019-06-25 14:00:26] ### doing inits for benchsmp depth 12: ###
[2019-06-25 14:00:28] ### computing benchsmp depth 12: ###
[2019-06-25 14:00:28] ### work-groups: 8 ###
[2019-06-25 14:00:28] depth score time nodes pv 
[2019-06-25 14:00:28] 1 60 56 160 b1c3 
[2019-06-25 14:00:28] 2 0 56 616 b1c3 b8c6 
[2019-06-25 14:00:28] 3 60 56 1842 b1c3 b8c6 g1f3 
[2019-06-25 14:00:28] 4 0 57 3485 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 14:00:28] 5 27 57 7579 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 14:00:28] 6 0 61 30042 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 14:00:28] 7 22 65 65766 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 14:00:28] 8 12 84 209256 b1c3 b8c6 d2d4 g8f6 d4d5 c6e5 g1f3 e5f3 
[2019-06-25 14:00:29] 9 18 145 685881 g1f3 g8f6 b1c3 d7d5 d2d3 d5d4 c3e4 b8c6 e4f6 
[2019-06-25 14:00:31] 10 12 359 2374218 g1f3 g8f6 b1c3 d7d5 e2e3 e7e6 f3d4 b8c6 f1b5 c8d7 
[2019-06-25 14:00:34] 11 12 656 4749810 g1f3 g8f6 b1c3 b8c6 e2e4 d7d6 d2d3 c8e6 c1e3 g7g6 b2b3 
[2019-06-25 14:00:50] 12 13 2279 17704831 g1f3 g8f6 b1c3 b8c6 e2e4 d7d5 e4e5 d5d4 c3b5 f6e4 f1d3 c8f5 
[2019-06-25 14:00:50] #17704831 searched nodes in 22.797000 seconds, with 371228 ttmovehits, and 276751 ttscorehits, 2821 iidhits, ebf: 3.610319, nps: 776629  
[2019-06-25 14:00:50] ### doing inits for benchsmp depth 12: ###
[2019-06-25 14:00:52] ### computing benchsmp depth 12: ###
[2019-06-25 14:00:52] ### work-groups: 16 ###
[2019-06-25 14:00:52] depth score time nodes pv 
[2019-06-25 14:00:52] 1 60 54 320 b1c3 
[2019-06-25 14:00:52] 2 0 54 1257 b1c3 b8c6 
[2019-06-25 14:00:52] 3 60 54 3267 b1c3 b8c6 g1f3 
[2019-06-25 14:00:52] 4 0 54 5936 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 14:00:52] 5 27 56 12096 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 14:00:52] 6 0 57 44531 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 14:00:53] 7 22 61 82806 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 14:00:53] 8 12 73 279021 b1c3 b8c6 d2d4 g8f6 d4d5 c6e5 g1f3 e5f3 
[2019-06-25 14:00:53] 9 18 101 730035 b1c3 g8f6 g1f3 b8c6 d2d3 d7d6 b2b3 b7b6 g2g3 
[2019-06-25 14:00:54] 10 14 237 2947693 b1c3 g8f6 g1f3 b8c6 e2e4 d7d5 f1d3 e7e6 b2b3 b7b6 
[2019-06-25 14:00:57] 11 12 528 7699709 b1c3 g8f6 g1f3 d7d5 d2d4 b8c6 e2e3 g7g6 f1d3 e7e6 b2b3 
[2019-06-25 14:01:07] 12 13 1558 24640447 e2e4 b8c6 b1c3 e7e6 g1f3 d7d5 d2d4 d5e4 c3e4 g8f6 e4f6 d8f6 
[2019-06-25 14:01:07] #24640447 searched nodes in 15.588000 seconds, with 604116 ttmovehits, and 414969 ttscorehits, 3207 iidhits, ebf: 3.703295, nps: 1580731  
[2019-06-25 14:01:07] ### doing inits for benchsmp depth 12: ###
[2019-06-25 14:01:09] ### computing benchsmp depth 12: ###
[2019-06-25 14:01:09] ### work-groups: 32 ###
[2019-06-25 14:01:09] depth score time nodes pv 
[2019-06-25 14:01:09] 1 60 53 640 b1c3 
[2019-06-25 14:01:09] 2 0 53 2537 b1c3 b8c6 
[2019-06-25 14:01:09] 3 60 53 5777 b1c3 b8c6 g1f3 
[2019-06-25 14:01:09] 4 0 53 10653 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 14:01:09] 5 27 54 21306 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 14:01:10] 6 0 56 72073 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 14:01:10] 7 22 57 136133 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 14:01:10] 8 12 67 409714 b1c3 b8c6 d2d4 d7d5 e2e3 g8f6 g1f3 e7e6 
[2019-06-25 14:01:10] 9 18 90 1150406 b1c3 d7d5 g1f3 g8f6 b2b3 d5d4 c3b5 b8c6 d2d3 
[2019-06-25 14:01:11] 10 12 190 4430758 b1c3 d7d5 g1f3 g8f6 e2e3 e7e6 f3d4 b8c6 f1b5 c8d7 
[2019-06-25 14:01:13] 11 11 390 11046634 g1f3 b8c6 e2e3 g8f6 b1c3 g7g6 f1b5 e7e6 b5c6 b7c6 f3e5 
[2019-06-25 14:01:18] 12 10 952 29639266 b1c3 b8c6 e2e4 e7e6 d2d4 f8b4 g1f3 g8f6 f1d3 b7b6 g2g3 g7g6 
[2019-06-25 14:01:18] #29639266 searched nodes in 9.525000 seconds, with 678031 ttmovehits, and 592958 ttscorehits, 3903 iidhits, ebf: 3.756289, nps: 3111733  
[2019-06-25 14:01:18] ### doing inits for benchsmp depth 12: ###
[2019-06-25 14:01:20] ### computing benchsmp depth 12: ###
[2019-06-25 14:01:20] ### work-groups: 36 ###
[2019-06-25 14:01:20] depth score time nodes pv 
[2019-06-25 14:01:20] 1 60 53 720 b1c3 
[2019-06-25 14:01:20] 2 0 53 2830 b1c3 b8c6 
[2019-06-25 14:01:21] 3 60 53 6924 b1c3 b8c6 g1f3 
[2019-06-25 14:01:21] 4 0 54 11905 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 14:01:21] 5 27 54 23750 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 14:01:21] 6 0 56 73917 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 14:01:21] 7 22 59 155093 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 14:01:21] 8 12 67 449836 b1c3 b8c6 d2d4 d7d5 g1f3 g8f6 e2e3 e7e6 
[2019-06-25 14:01:21] 9 15 87 1195232 b1c3 b8c6 d2d4 g8f6 g1f3 d7d6 e2e3 g7g6 b2b3 
[2019-06-25 14:01:22] 10 13 221 6135299 b1c3 b8c6 g1f3 g8f6 e2e3 g7g6 b2b3 d7d6 g2g3 c8g4 
[2019-06-25 14:01:24] 11 12 392 12475967 b1c3 b8c6 g1f3 g8f6 e2e3 e7e6 g2g3 g7g6 b2b3 f8b4 f1c4 
[2019-06-25 14:01:28] 12 10 821 28587863 b1c3 b8c6 g1f3 g8f6 e2e3 e7e6 d2d4 g7g6 g2g3 b7b6 f1b5 f8d6 
[2019-06-25 14:01:28] #28587863 searched nodes in 8.219000 seconds, with 814605 ttmovehits, and 587883 ttscorehits, 2627 iidhits, ebf: 3.745868, nps: 3478265  
[2019-06-25 14:01:28] ### workers	#nps		#nps speedup	#time in s	#ttd speedup	#relative ttd speedup ###
[2019-06-25 14:01:28] ### 1		92195		1.000000	123.169000	1.000000	1.000000 
[2019-06-25 14:01:28] ### 2		183770		1.993275	59.098000	2.084148	2.084148 
[2019-06-25 14:01:28] ### 4		370610		4.019849	36.778000	3.348986	1.606885 
[2019-06-25 14:01:28] ### 8		776629		8.423765	22.797000	5.402860	1.613282 
[2019-06-25 14:01:28] ### 16		1580731		17.145518	15.588000	7.901527	1.462471 
[2019-06-25 14:01:28] ### 32		3111733		33.751646	9.525000	12.931129	1.636535 
[2019-06-25 14:01:28] ### 36		3478265		37.727263	8.219000	14.985886	1.158900 
Torsten

smatovic
Posts: 903
Joined: Wed Mar 10, 2010 9:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by smatovic » Tue Jun 25, 2019 12:12 pm

Dokterchen wrote:
Tue Jun 25, 2019 12:06 pm
smatovic wrote:
Tue Jun 25, 2019 11:39 am
smatovic wrote:
Tue Jun 25, 2019 11:25 am
Dokterchen wrote:
Tue Jun 25, 2019 11:01 am
Hello Srdja,

I have a RTX 2070 and maybe I did someting wrong but Zeta crashed during the benchmark:
...
KR
Torsten
Hello Torsten,

thanks for testing, not 100% sure, it may be a bug in my program, but the search
is running only for a second+, so this could be a sign for OS or driver depended
gpu timeout.

The .reg file from my repository solved on my Windows 7 Pro a 2 second gpu
timeout issue for me.

https://raw.githubusercontent.com/smato ... tTo20s.reg

Download, double-click and reboot OS. Maybe this is a solution.

Afaik LC0 works differently, it does many small batches per second, so it is no
subject to timeouts like Zeta is.

--
Srdja
Just realized that even an increased 20 seconds timeout will not be sufficient
for the benchsmp command to succeed...

here an mod with 200 seconds timeout:

https://zeta-chess.app26.de/downloads/S ... To200s.reg

--
Srdja
OK, that fixed the problem:
...
Hmm, I don't want to bother, but did you rerun the --guessconfigx command after applying the .reg file?
It looks like already during the guessconfigx command the timeout occurred, so the RTX 2070 will not be
utilized fully during benchsmp...

--
Srdja

Dokterchen
Posts: 104
Joined: Wed Aug 15, 2007 10:18 am
Location: Munich

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by Dokterchen » Tue Jun 25, 2019 1:04 pm

smatovic wrote:
Tue Jun 25, 2019 12:12 pm
Dokterchen wrote:
Tue Jun 25, 2019 12:06 pm
smatovic wrote:
Tue Jun 25, 2019 11:39 am
smatovic wrote:
Tue Jun 25, 2019 11:25 am
Dokterchen wrote:
Tue Jun 25, 2019 11:01 am
Hello Srdja,

I have a RTX 2070 and maybe I did someting wrong but Zeta crashed during the benchmark:
...
KR
Torsten
Hello Torsten,

thanks for testing, not 100% sure, it may be a bug in my program, but the search
is running only for a second+, so this could be a sign for OS or driver depended
gpu timeout.

The .reg file from my repository solved on my Windows 7 Pro a 2 second gpu
timeout issue for me.

https://raw.githubusercontent.com/smato ... tTo20s.reg

Download, double-click and reboot OS. Maybe this is a solution.

Afaik LC0 works differently, it does many small batches per second, so it is no
subject to timeouts like Zeta is.

--
Srdja
Just realized that even an increased 20 seconds timeout will not be sufficient
for the benchsmp command to succeed...

here an mod with 200 seconds timeout:

https://zeta-chess.app26.de/downloads/S ... To200s.reg

--
Srdja
OK, that fixed the problem:
...
Hmm, I don't want to bother, but did you rerun the --guessconfigx command after applying the .reg file?
It looks like already during the guessconfigx command the timeout occurred, so the RTX 2070 will not be
utilized fully during benchsmp...

--
Srdja
No problem, here we are again:

Code: Select all

[2019-06-25 14:56:01] zeta.exe -l -p 0 -d 0 --guessconfigx 
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> ### Query the OpenCL Platforms on Host...
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> Number of OpenCL Platforms found: 1 
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> Platform: 0,  Vendor:  NVIDIA Corporation 
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> ### Query the OpenCL Devices on Platform...
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> Number of OpenCL Devices found: 1 
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> ### Query and check the OpenCL Device...
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> Device: 0, Device name: GeForce RTX 2070 
[2019-06-25 14:56:01] #>
[2019-06-25 14:56:01] #> OK, Device Endianness is little
[2019-06-25 14:56:01] #> OK, CL_DEVICE_MAX_COMPUTE_UNITS: 36 
[2019-06-25 14:56:01] #> OK, CL_DEVICE_MAX_MEM_ALLOC_SIZE: 2048 MB >= 128 MB 
[2019-06-25 14:56:01] #> OK, CL_DEVICE_GLOBAL_MEM_SIZE: 8192 MB
[2019-06-25 14:56:01] #> OK, Device extension cl_khr_global_int32_base_atomics is supported.
[2019-06-25 14:56:01] #> OK, Device extension cl_khr_local_int32_base_atomics is supported.
[2019-06-25 14:56:01] #> OK: Device extension cl_khr_int64_extended_atomics not supported.
[2019-06-25 14:56:01] #> OK, CL_DEVICE_MAX_WORK_GROUP_SIZE: 1024 >= 64
[2019-06-25 14:56:01] #> OK, CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS: 3 >= 3
[2019-06-25 14:56:01] #> OK, CL_DEVICE_MAX_WORK_ITEM_SIZES [3]: 64 >= 64
[2019-06-25 14:56:01] #> OK, CL_DEVICE_AVAILABLE: CL_TRUE 
[2019-06-25 14:56:01] #
[2019-06-25 14:56:01] #> ### Running NPS-Benchmark for minimal config on device,
[2019-06-25 14:56:01] #> ### this can last about 4 seconds... 
[2019-06-25 14:56:01] #> ### threadsX: 1 
[2019-06-25 14:56:01] #> ### threadsY: 1 
[2019-06-25 14:56:01] #> ### total work-groups: 1 
[2019-06-25 14:56:01] #> ### total threads: 64 
[2019-06-25 14:56:01] #
[2019-06-25 14:56:03] #fen: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
[2019-06-25 14:56:03] ###ABCDEFGH###
[2019-06-25 14:56:03] #8 rnbqkbnr
[2019-06-25 14:56:03] #7 pppppppp
[2019-06-25 14:56:03] #6 --------
[2019-06-25 14:56:03] #5 --------
[2019-06-25 14:56:03] #4 --------
[2019-06-25 14:56:03] #3 --------
[2019-06-25 14:56:03] #2 PPPPPPPP
[2019-06-25 14:56:03] #1 RNBQKBNR
[2019-06-25 14:56:03] ###ABCDEFGH###
[2019-06-25 14:56:03] depth: 1, nodes 20, nps: 36, time: 0.547000 sec, score: 60  move b1c3
[2019-06-25 14:56:03] depth: 2, nodes 59, nps: 59000, time: 0.001000 sec, score: 0  move b1c3
[2019-06-25 14:56:03] depth: 3, nodes 525, nps: 30882, time: 0.017000 sec, score: 60  move b1c3
[2019-06-25 14:56:03] depth: 4, nodes 618, nps: 36352, time: 0.017000 sec, score: 0  move b1c3
[2019-06-25 14:56:03] depth: 5, nodes 2430, nps: 75937, time: 0.032000 sec, score: 27  move b1c3
[2019-06-25 14:56:03] depth: 6, nodes 16392, nps: 74509, time: 0.220000 sec, score: 0  move b1c3
[2019-06-25 14:56:04] depth: 7, nodes 20778, nps: 94876, time: 0.219000 sec, score: 22  move b1c3
[2019-06-25 14:56:05] depth: 8, nodes 83110, nps: 90043, time: 0.923000 sec, score: 12  move b1c3
[2019-06-25 14:56:07] depth: 9, nodes 254196, nps: 92941, time: 2.735000 sec, score: 16  move b1c3
[2019-06-25 14:56:08] #
[2019-06-25 14:56:08] #> ### Running NPS-Benchmark for best config,
[2019-06-25 14:56:08] #> ### this can last about some minutes... 
[2019-06-25 14:56:08] #
[2019-06-25 14:56:08] #
[2019-06-25 14:56:08] #> ### Running NPS-Benchmark for threadsY on device,
[2019-06-25 14:56:08] #> ### this can last about 4 seconds... 
[2019-06-25 14:56:08] #> ### threadsX: 36 
[2019-06-25 14:56:08] #> ### threadsY: 1 
[2019-06-25 14:56:08] #> ### total work-groups: 36 
[2019-06-25 14:56:08] #> ### total threads: 2304 
[2019-06-25 14:56:08] #
[2019-06-25 14:56:09] #fen: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
[2019-06-25 14:56:09] ###ABCDEFGH###
[2019-06-25 14:56:09] #8 rnbqkbnr
[2019-06-25 14:56:09] #7 pppppppp
[2019-06-25 14:56:09] #6 --------
[2019-06-25 14:56:09] #5 --------
[2019-06-25 14:56:09] #4 --------
[2019-06-25 14:56:09] #3 --------
[2019-06-25 14:56:09] #2 PPPPPPPP
[2019-06-25 14:56:09] #1 RNBQKBNR
[2019-06-25 14:56:09] ###ABCDEFGH###
[2019-06-25 14:56:09] depth: 1, nodes 720, nps: 1210, time: 0.595000 sec, score: 60  move b1c3
[2019-06-25 14:56:09] depth: 2, nodes 2028, nps: 2028000, time: 0.001000 sec, score: 0  move b1c3
[2019-06-25 14:56:09] depth: 3, nodes 3700, nps: 231250, time: 0.016000 sec, score: 60  move b1c3
[2019-06-25 14:56:09] depth: 4, nodes 5431, nps: 5431000, time: 0.001000 sec, score: 0  move b1c3
[2019-06-25 14:56:09] depth: 5, nodes 11444, nps: 11444000, time: 0.001000 sec, score: 27  move b1c3
[2019-06-25 14:56:09] depth: 6, nodes 51803, nps: 3047235, time: 0.017000 sec, score: 0  move b1c3
[2019-06-25 14:56:09] depth: 7, nodes 75287, nps: 4705437, time: 0.016000 sec, score: 22  move b1c3
[2019-06-25 14:56:10] depth: 8, nodes 398190, nps: 3160238, time: 0.126000 sec, score: 13  move b1c3
[2019-06-25 14:56:10] depth: 9, nodes 1717173, nps: 3782319, time: 0.454000 sec, score: 18  move g1f3
[2019-06-25 14:56:11] depth: 10, nodes 4142087, nps: 3728251, time: 1.111000 sec, score: 15  move g1f3
[2019-06-25 14:56:11] #
[2019-06-25 14:56:11] #> ### Running NPS-Benchmark for threadsY on device,
[2019-06-25 14:56:11] #> ### this can last about 4 seconds... 
[2019-06-25 14:56:11] #> ### threadsX: 36 
[2019-06-25 14:56:11] #> ### threadsY: 2 
[2019-06-25 14:56:11] #> ### total work-groups: 72 
[2019-06-25 14:56:11] #> ### total threads: 4608 
[2019-06-25 14:56:11] #
[2019-06-25 14:56:13] #fen: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
[2019-06-25 14:56:13] ###ABCDEFGH###
[2019-06-25 14:56:13] #8 rnbqkbnr
[2019-06-25 14:56:13] #7 pppppppp
[2019-06-25 14:56:13] #6 --------
[2019-06-25 14:56:13] #5 --------
[2019-06-25 14:56:13] #4 --------
[2019-06-25 14:56:13] #3 --------
[2019-06-25 14:56:13] #2 PPPPPPPP
[2019-06-25 14:56:13] #1 RNBQKBNR
[2019-06-25 14:56:13] ###ABCDEFGH###
[2019-06-25 14:56:13] depth: 1, nodes 1440, nps: 2706, time: 0.532000 sec, score: 60  move b1c3
[2019-06-25 14:56:13] depth: 2, nodes 2975, nps: 2975000, time: 0.001000 sec, score: 0  move b1c3
[2019-06-25 14:56:13] depth: 3, nodes 7054, nps: 7054000, time: 0.001000 sec, score: 60  move b1c3
[2019-06-25 14:56:13] depth: 4, nodes 10348, nps: 10348000, time: 0.001000 sec, score: 0  move b1c3
[2019-06-25 14:56:13] depth: 5, nodes 22084, nps: 22084000, time: 0.001000 sec, score: 27  move b1c3
[2019-06-25 14:56:13] depth: 6, nodes 87911, nps: 5171235, time: 0.017000 sec, score: 0  move b1c3
[2019-06-25 14:56:13] depth: 7, nodes 136610, nps: 8035882, time: 0.017000 sec, score: 22  move b1c3
[2019-06-25 14:56:13] depth: 8, nodes 442765, nps: 6918203, time: 0.064000 sec, score: 12  move b1c3
[2019-06-25 14:56:14] depth: 9, nodes 1280427, nps: 6276602, time: 0.204000 sec, score: 18  move b1c3
[2019-06-25 14:56:15] depth: 10, nodes 6935339, nps: 6720289, time: 1.032000 sec, score: 10  move b1c3
[2019-06-25 14:56:15] #
[2019-06-25 14:56:15] #> ### Running NPS-Benchmark for threadsY on device,
[2019-06-25 14:56:15] #> ### this can last about 4 seconds... 
[2019-06-25 14:56:15] #> ### threadsX: 36 
[2019-06-25 14:56:15] #> ### threadsY: 4 
[2019-06-25 14:56:15] #> ### total work-groups: 144 
[2019-06-25 14:56:15] #> ### total threads: 9216 
[2019-06-25 14:56:15] #
[2019-06-25 14:56:16] #fen: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
[2019-06-25 14:56:16] ###ABCDEFGH###
[2019-06-25 14:56:16] #8 rnbqkbnr
[2019-06-25 14:56:16] #7 pppppppp
[2019-06-25 14:56:16] #6 --------
[2019-06-25 14:56:16] #5 --------
[2019-06-25 14:56:16] #4 --------
[2019-06-25 14:56:16] #3 --------
[2019-06-25 14:56:16] #2 PPPPPPPP
[2019-06-25 14:56:16] #1 RNBQKBNR
[2019-06-25 14:56:16] ###ABCDEFGH###
[2019-06-25 14:56:17] depth: 1, nodes 2880, nps: 5403, time: 0.533000 sec, score: 60  move b1c3
[2019-06-25 14:56:17] depth: 2, nodes 5929, nps: 5929000, time: 0.001000 sec, score: 0  move b1c3
[2019-06-25 14:56:17] depth: 3, nodes 13054, nps: 13054000, time: 0.001000 sec, score: 60  move b1c3
[2019-06-25 14:56:17] depth: 4, nodes 19588, nps: 1152235, time: 0.017000 sec, score: 0  move b1c3
[2019-06-25 14:56:17] depth: 5, nodes 41628, nps: 41628000, time: 0.001000 sec, score: 27  move b1c3
[2019-06-25 14:56:17] depth: 6, nodes 136544, nps: 8534000, time: 0.016000 sec, score: 0  move b1c3
[2019-06-25 14:56:17] depth: 7, nodes 227772, nps: 6902181, time: 0.033000 sec, score: 22  move b1c3
[2019-06-25 14:56:17] depth: 8, nodes 680905, nps: 8619050, time: 0.079000 sec, score: 3  move b1c3
[2019-06-25 14:56:17] depth: 9, nodes 2296412, nps: 9149051, time: 0.251000 sec, score: 18  move b1c3
[2019-06-25 14:56:18] depth: 10, nodes 8266673, nps: 9436841, time: 0.876000 sec, score: 14  move b1c3
[2019-06-25 14:56:19] depth: 11, nodes 12763190, nps: 9489360, time: 1.345000 sec, score: 12  move b1c3
[2019-06-25 14:56:20] #
#
################################################################################
// Zeta OpenCL Chess config file for device: GeForce RTX 2070 
################################################################################
threadsX: 36;
threadsY: 2;
nodes_per_second: 6720289;
tt1_memory: 1024; // in MB
tt2_memory: 768; // in MB
opencl_platform_id: 0;
opencl_device_id: 0;
opencl_gpugen: 2;
################################################################################
[2019-06-25 14:56:20] ##### Above output was saved in file config_0_0_.txt 
[2019-06-25 14:56:20] ##### rename it to config.txt to let engine use it
[2019-06-25 14:56:20] #
[2019-06-25 14:56:55] zeta.exe -l 
[2019-06-25 14:56:55] #> Zeta 099l
[2019-06-25 14:56:55] #> Experimental chess engine written in OpenCL.
[2019-06-25 14:56:55] #> Copyright (C) 2011-2019 Srdja Matovic, Montenegro
[2019-06-25 14:56:55] #> This is free software, licensed under GPL >= v2
[2019-06-25 14:56:55] #> eninge is initialising...
[2019-06-25 14:56:55] feature done=0
[2019-06-25 14:56:57] #> ...finished basic inits.
[2019-06-25 14:57:00] >> new
[2019-06-25 14:57:02] #fen: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
[2019-06-25 14:57:02] ###ABCDEFGH###
[2019-06-25 14:57:02] #8 rnbqkbnr
[2019-06-25 14:57:02] #7 pppppppp
[2019-06-25 14:57:02] #6 --------
[2019-06-25 14:57:02] #5 --------
[2019-06-25 14:57:02] #4 --------
[2019-06-25 14:57:02] #3 --------
[2019-06-25 14:57:02] #2 PPPPPPPP
[2019-06-25 14:57:02] #1 RNBQKBNR
[2019-06-25 14:57:02] ###ABCDEFGH###
[2019-06-25 14:57:15] >> sd 12
[2019-06-25 14:57:19] >> st 2000
[2019-06-25 14:57:28] >> benchsmp
[2019-06-25 14:57:28] ### doing inits for benchsmp depth 12: ###
[2019-06-25 14:57:30] ### computing benchsmp depth 12: ###
[2019-06-25 14:57:30] ### work-groups: 1 ###
[2019-06-25 14:57:30] depth score time nodes pv 
[2019-06-25 14:57:30] 1 60 52 20 b1c3 
[2019-06-25 14:57:30] 2 0 52 79 b1c3 b8c6 
[2019-06-25 14:57:30] 3 60 53 604 b1c3 b8c6 g1f3 
[2019-06-25 14:57:30] 4 0 53 1222 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 14:57:30] 5 27 58 3652 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 14:57:30] 6 0 80 20044 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 14:57:31] 7 22 102 40822 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 14:57:32] 8 12 195 123908 b1c3 b8c6 d2d4 g8f6 d4d5 c6e5 g1f3 e5f3 
[2019-06-25 14:57:34] 9 16 470 378072 b1c3 g8f6 g1f3 g7g6 d2d4 d7d6 c1f4 b8c6 e2e3 
[2019-06-25 14:57:53] 10 15 2288 2051845 b1c3 g8f6 g1f3 g7g6 e2e4 b8c6 d2d3 d7d5 c1g5 c8e6 
[2019-06-25 14:58:18] 11 12 4877 4445440 b1c3 g8f6 g1f3 b8c6 e2e3 g7g6 b2b3 e7e6 g2g3 f8b4 f1d3 
[2019-06-25 14:59:33] 12 7 12331 11355889 b1c3 b8c6 g1f3 e7e6 e2e3 g7g6 f1b5 g8f6 g2g3 f8d6 b2b3 d6b4 
[2019-06-25 14:59:33] #11355889 searched nodes in 123.318000 seconds, with 231876 ttmovehits, and 168515 ttscorehits, 1776 iidhits, ebf: 3.489067, nps: 92086  
[2019-06-25 14:59:33] ### doing inits for benchsmp depth 12: ###
[2019-06-25 14:59:34] ### computing benchsmp depth 12: ###
[2019-06-25 14:59:34] ### work-groups: 2 ###
[2019-06-25 14:59:34] depth score time nodes pv 
[2019-06-25 14:59:35] 1 60 50 40 b1c3 
[2019-06-25 14:59:35] 2 0 50 157 b1c3 b8c6 
[2019-06-25 14:59:35] 3 60 51 805 b1c3 b8c6 g1f3 
[2019-06-25 14:59:35] 4 0 51 1572 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 14:59:35] 5 27 53 4275 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 14:59:35] 6 0 64 21072 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 14:59:35] 7 22 75 40832 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 14:59:36] 8 10 153 186133 b1c3 d7d6 e2e4 e7e5 d2d4 g8f6 f1b5 c7c6 
[2019-06-25 14:59:38] 9 17 381 609451 g1f3 g8f6 d2d4 d7d5 e2e3 b8c6 b1d2 e7e6 c2c3 
[2019-06-25 14:59:49] 10 12 1506 2686691 g1f3 g8f6 b1c3 d7d5 e2e3 e7e6 f3d4 b8c6 f1b5 c8d7 
[2019-06-25 15:00:13] 11 12 3906 7131639 g2g3 b8c6 g1f3 g8f6 b2b3 d7d6 b1c3 c8g4 d2d3 b7b6 c1f4 
[2019-06-25 15:00:44] 12 8 6978 12837078 g2g3 b8c6 g1f3 d7d5 e2e3 g8f6 b1c3 a7a6 d2d4 e7e6 f1d3 g7g6 
[2019-06-25 15:00:44] #12837078 searched nodes in 69.789000 seconds, with 243795 ttmovehits, and 203450 ttscorehits, 2081 iidhits, ebf: 3.522128, nps: 183941  
[2019-06-25 15:00:44] ### doing inits for benchsmp depth 12: ###
[2019-06-25 15:00:46] ### computing benchsmp depth 12: ###
[2019-06-25 15:00:46] ### work-groups: 4 ###
[2019-06-25 15:00:46] depth score time nodes pv 
[2019-06-25 15:00:46] 1 60 50 80 b1c3 
[2019-06-25 15:00:46] 2 0 50 313 b1c3 b8c6 
[2019-06-25 15:00:46] 3 60 50 1156 b1c3 b8c6 g1f3 
[2019-06-25 15:00:46] 4 0 51 2218 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 15:00:46] 5 27 51 5368 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 15:00:46] 6 0 57 25130 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 15:00:46] 7 22 65 54707 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 15:00:47] 8 12 108 209724 b1c3 g8f6 g1f3 b8c6 d2d3 d7d6 b2b3 b7b6 
[2019-06-25 15:00:47] 9 18 175 460303 b1c3 g8f6 g1f3 e7e6 d2d3 b7b6 f3e5 b8c6 c1f4 
[2019-06-25 15:00:53] 10 9 715 2486886 b1c3 g8f6 g1f3 b8c6 d2d4 d7d5 c1f4 c8e6 e2e3 g7g6 
[2019-06-25 15:01:00] 11 12 1465 5315593 b1c3 g8f6 g1f3 b8c6 e2e4 d7d5 e4d5 f6d5 c3d5 d8d5 c2c3 
[2019-06-25 15:01:21] 12 13 3513 13060132 b1c3 b8c6 g1f3 d7d5 d2d4 e7e6 e2e4 d5e4 c3e4 g8e7 c2c3 e7d5 
[2019-06-25 15:01:21] #13060132 searched nodes in 35.136000 seconds, with 282392 ttmovehits, and 213412 ttscorehits, 1862 iidhits, ebf: 3.526798, nps: 371702  
[2019-06-25 15:01:21] ### doing inits for benchsmp depth 12: ###
[2019-06-25 15:01:22] ### computing benchsmp depth 12: ###
[2019-06-25 15:01:22] ### work-groups: 8 ###
[2019-06-25 15:01:22] depth score time nodes pv 
[2019-06-25 15:01:23] 1 60 50 160 b1c3 
[2019-06-25 15:01:23] 2 0 50 615 b1c3 b8c6 
[2019-06-25 15:01:23] 3 60 51 1871 b1c3 b8c6 g1f3 
[2019-06-25 15:01:23] 4 0 51 3468 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 15:01:23] 5 27 51 7465 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 15:01:23] 6 0 56 29883 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 15:01:23] 7 22 61 62889 b1c3 b8c6 g1f3 e7e6 d2d4 g8f6 e2e3 
[2019-06-25 15:01:23] 8 1 82 233950 d2d3 b8c6 g1f3 g8f6 g2g3 d7d5 b1c3 c8g4 
[2019-06-25 15:01:24] 9 18 143 707165 g1f3 g8f6 d2d3 b8c6 e2e4 d7d5 b1c3 d5e4 d3e4 
[2019-06-25 15:01:26] 10 15 367 2472950 g1f3 g8f6 b1c3 e7e6 e2e4 b8c6 d2d4 d7d5 e4e5 f6d7 
[2019-06-25 15:01:31] 11 10 895 6692891 g1f3 g8f6 d2d4 e7e6 e2e3 b8c6 b1d2 g7g6 c2c3 f6d5 g2g3 
[2019-06-25 15:01:39] 12 11 1703 13134509 g1f3 g8f6 d2d4 d7d5 b1c3 e7e6 e2e3 f8d6 b2b3 b8c6 g2g3 g7g6 
[2019-06-25 15:01:39] #13134509 searched nodes in 17.032000 seconds, with 300778 ttmovehits, and 223973 ttscorehits, 1674 iidhits, ebf: 3.528339, nps: 771166  
[2019-06-25 15:01:39] ### doing inits for benchsmp depth 12: ###
[2019-06-25 15:01:41] ### computing benchsmp depth 12: ###
[2019-06-25 15:01:41] ### work-groups: 16 ###
[2019-06-25 15:01:41] depth score time nodes pv 
[2019-06-25 15:01:41] 1 60 51 320 b1c3 
[2019-06-25 15:01:41] 2 0 51 1257 b1c3 b8c6 
[2019-06-25 15:01:41] 3 60 53 3241 b1c3 b8c6 g1f3 
[2019-06-25 15:01:41] 4 0 53 5830 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 15:01:41] 5 27 53 11783 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 15:01:41] 6 0 56 45427 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 15:01:41] 7 22 59 91281 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 15:01:42] 8 12 70 264428 b1c3 b8c6 d2d4 d7d5 g1f3 g8f6 e2e3 e7e6 
[2019-06-25 15:01:42] 9 15 118 1043971 b1c3 b8c6 g1f3 e7e6 d2d3 g8e7 b2b3 b7b6 c1f4 
[2019-06-25 15:01:44] 10 9 292 3851477 g1f3 g8f6 b1c3 e7e6 e2e4 b8c6 e4e5 f6g4 d2d4 b7b6 
[2019-06-25 15:01:48] 11 13 670 10024316 b1c3 b8c6 g1f3 g8f6 e2e4 e7e5 f1b5 d7d6 b5c6 b7c6 d2d3 
[2019-06-25 15:01:55] 12 10 1414 22283865 b1c3 b8c6 g1f3 g8f6 e2e3 e7e6 g2g3 g7g6 f1b5 a7a6 b5c6 b7c6 
[2019-06-25 15:01:55] #22283865 searched nodes in 14.149000 seconds, with 479538 ttmovehits, and 387416 ttscorehits, 3091 iidhits, ebf: 3.674769, nps: 1574942  
[2019-06-25 15:01:55] ### doing inits for benchsmp depth 12: ###
[2019-06-25 15:01:56] ### computing benchsmp depth 12: ###
[2019-06-25 15:01:56] ### work-groups: 32 ###
[2019-06-25 15:01:56] depth score time nodes pv 
[2019-06-25 15:01:57] 1 60 51 640 b1c3 
[2019-06-25 15:01:57] 2 0 51 2526 b1c3 b8c6 
[2019-06-25 15:01:57] 3 60 51 6297 b1c3 b8c6 g1f3 
[2019-06-25 15:01:57] 4 0 51 11099 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 15:01:57] 5 27 53 21751 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 15:01:57] 6 0 54 69803 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 15:01:57] 7 22 56 136524 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 15:01:57] 8 12 64 390695 b1c3 b8c6 d2d4 d7d5 c1f4 g8f6 g1f3 e7e6 
[2019-06-25 15:01:57] 9 18 84 1039940 b1c3 g8f6 g1f3 e7e6 d2d3 b8c6 g2g3 b7b6 b2b3 
[2019-06-25 15:01:58] 10 13 197 4707769 b1c3 g8f6 g1f3 b8c6 d2d4 d7d5 e2e3 e7e6 g2g3 b7b6 
[2019-06-25 15:02:00] 11 12 381 10807653 b1c3 g8f6 e2e4 e7e5 d2d3 b8c6 g1f3 d7d5 e4d5 f6d5 c3e4 
[2019-06-25 15:02:05] 12 10 893 27817597 b1c3 g8f6 e2e3 b8c6 g1f3 g7g6 d2d4 e7e6 g2g3 f8d6 f1b5 a7a6 
[2019-06-25 15:02:05] #27817597 searched nodes in 8.939000 seconds, with 870224 ttmovehits, and 496186 ttscorehits, 2212 iidhits, ebf: 3.738006, nps: 3111936  
[2019-06-25 15:02:05] ### doing inits for benchsmp depth 12: ###
[2019-06-25 15:02:07] ### computing benchsmp depth 12: ###
[2019-06-25 15:02:07] ### work-groups: 36 ###
[2019-06-25 15:02:07] depth score time nodes pv 
[2019-06-25 15:02:07] 1 60 50 720 b1c3 
[2019-06-25 15:02:07] 2 0 51 2830 b1c3 b8c6 
[2019-06-25 15:02:07] 3 60 51 7121 b1c3 b8c6 g1f3 
[2019-06-25 15:02:07] 4 0 51 12591 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 15:02:07] 5 27 51 24224 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 15:02:07] 6 0 53 75054 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 15:02:07] 7 22 56 137035 b1c3 g8f6 g1f3 e7e6 d2d3 b8c6 b2b3 
[2019-06-25 15:02:07] 8 1 62 395112 b1c3 g8f6 g2g3 d7d6 g1f3 b8c6 d2d3 c8g4 
[2019-06-25 15:02:08] 9 18 89 1334115 b1c3 g8f6 g1f3 d7d5 d2d4 g7g6 c1g5 b8c6 e2e3 
[2019-06-25 15:02:09] 10 13 217 6077563 d2d4 g8f6 b1c3 e7e6 e2e4 f8b4 e4e5 f6d5 g1e2 b8c6 
[2019-06-25 15:02:11] 11 12 381 12184827 b1c3 g8f6 g1f3 d7d5 e2e3 b8c6 d2d4 e7e6 g2g3 g7g6 b2b3 
[2019-06-25 15:02:15] 12 10 801 27942599 b1c3 g8f6 e2e4 b8c6 d2d4 d7d5 e4e5 f6e4 c3e4 d5e4 g1e2 c8e6 
[2019-06-25 15:02:15] #27942599 searched nodes in 8.016000 seconds, with 768132 ttmovehits, and 527252 ttscorehits, 2967 iidhits, ebf: 3.739295, nps: 3485853  
[2019-06-25 15:02:15] ### doing inits for benchsmp depth 12: ###
[2019-06-25 15:02:16] ### computing benchsmp depth 12: ###
[2019-06-25 15:02:16] ### work-groups: 72 ###
[2019-06-25 15:02:16] depth score time nodes pv 
[2019-06-25 15:02:17] 1 60 50 1440 b1c3 
[2019-06-25 15:02:17] 2 0 51 4448 b1c3 b8c6 
[2019-06-25 15:02:17] 3 60 51 11740 b1c3 b8c6 g1f3 
[2019-06-25 15:02:17] 4 0 51 22109 b1c3 b8c6 g1f3 g8f6 
[2019-06-25 15:02:17] 5 27 51 43850 b1c3 b8c6 g1f3 g8f6 d2d3 
[2019-06-25 15:02:17] 6 0 53 137021 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 
[2019-06-25 15:02:17] 7 22 56 250879 b1c3 b8c6 g1f3 g8f6 d2d3 d7d6 b2b3 
[2019-06-25 15:02:17] 8 4 64 782334 b1c3 b8c6 d2d4 d7d5 g1f3 g8f6 e2e3 e7e6 
[2019-06-25 15:02:17] 9 15 97 2880714 g1f3 b8c6 d2d3 d7d5 b2b3 g8f6 b1c3 e7e6 c1f4 
[2019-06-25 15:02:18] 10 12 207 10229051 g1f3 b8c6 d2d4 e7e6 e2e4 b7b6 b1c3 g8f6 e4e5 f6g4 
[2019-06-25 15:02:20] 11 12 414 23754610 e2e3 g8f6 g1f3 g7g6 b2b3 b8c6 b1c3 d7d6 g2g3 c8e6 d2d4 
[2019-06-25 15:02:26] 12 7 936 58389404 e2e3 g8f6 b1c3 d7d5 d2d4 e7e6 g1f3 g7g6 f1d3 b8c6 g2g3 b7b6 
[2019-06-25 15:02:26] #58389404 searched nodes in 9.360000 seconds, with 1575647 ttmovehits, and 1504839 ttscorehits, 5682 iidhits, ebf: 3.957403, nps: 6238184  
[2019-06-25 15:02:26] ### workers	#nps		#nps speedup	#time in s	#ttd speedup	#relative ttd speedup ###
[2019-06-25 15:02:26] ### 1		92086		1.000000	123.318000	1.000000	1.000000 
[2019-06-25 15:02:26] ### 2		183912		1.997177	69.800000	1.766734	1.766734 
[2019-06-25 15:02:26] ### 4		371659		4.035999	35.140000	3.509334	1.986340 
[2019-06-25 15:02:26] ### 8		771166		8.374411	17.032000	7.240371	2.063175 
[2019-06-25 15:02:26] ### 16		1572164		17.072780	14.174000	8.700296	1.201637 
[2019-06-25 15:02:26] ### 32		3111936		33.793801	8.939000	13.795503	1.585636 
[2019-06-25 15:02:26] ### 36		3478909		37.778913	8.032000	15.353337	1.112923 
[2019-06-25 15:02:26] ### 72		6238184		67.743023	9.360000	13.175000	0.858120 
Torsten

smatovic
Posts: 903
Joined: Wed Mar 10, 2010 9:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by smatovic » Tue Jun 25, 2019 1:24 pm

Okay, thank you.

According to the results, the RTX 2070 runs during guessconfigx only 1 or 2
workers per Compute Unit efficiently, 36 CUs * 2 workers = 72 workers in total.
Contrary Pascal architecture was able to run 4 workers per Compute Unit.

So, thank you all, the patch was partly successful, at least for Pascal, and
the half way for Turing.

Zeta v099l release is now available on my GitHub page:

https://github.com/smatovic/Zeta/releases

https://github.com/smatovic/Zeta/releases/tag/v099l

For further Turing tuning I really need an hardware upgrade, or maybe I will
rent something via Amazon AWS or Google Cloud...

--
Srdja

Dann Corbit
Posts: 10096
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by Dann Corbit » Tue Jun 25, 2019 4:47 pm

Your project is showing that alpha beta search can be successful on GPU cards (I assume here that you have not switched to MCTS, is that right?)
I recall at the start of your project, some well wishers explained that it could not be done.
You seem to have beautiful NPS scaling.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

smatovic
Posts: 903
Joined: Wed Mar 10, 2010 9:18 pm
Location: Hamburg, Germany
Full name: Srdja Matovic
Contact:

Re: Looking for someone to test Zeta v099l on RTX 2080 TI, or similar, gpu

Post by smatovic » Tue Jun 25, 2019 5:38 pm

Dann Corbit wrote:
Tue Jun 25, 2019 4:47 pm
Your project is showing that alpha beta search can be successful on GPU cards
Yea :-)

If only someone could somehow increase the nps throughput per worker 10 fold...
Dann Corbit wrote:
Tue Jun 25, 2019 4:47 pm
(I assume here that you have not switched to MCTS, is that right?)
Yes, the v099 series is classic parallel AlphaBeta, v097 and v098 were BestFirstMiniMax,
similar to MCTS but with AlphaBeta playouts at leaf nodes.
Dann Corbit wrote:
Tue Jun 25, 2019 4:47 pm
You seem to have beautiful NPS scaling.
It is a bit too shiny, the ID loop is on CPU host, so with each search depth
iteration there is some fix overhead for calling the gpu kernel, therefore the
partly superlinear speedups.

And the time to depth speedup is imo not that bad either, ignore the step from 32 to 64 workers,
my ABDADA implementation scales only up to 32 workers, and the RMO parallel search
steps in from 64 and above workers, but I keep ABDADA running till 64 cos in heavy positions
there might be something to gain, and in future there will be plenty of workers to fill with work.

--
Srdja

Post Reply