Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Rybka Support / Large Pages in Rybka 4
1 2 3 4 5 Previous Next  
Parent - By Lukas Cimiotti (Bronze) [de] Date 2010-06-21 06:08

>Would you seriously recommend 24 gig of RAM


No, that was a special advice for Alan :evil:
Parent - By Banned for Life (Gold) Date 2010-06-21 07:34
Actually, you need 24 or 32 GB of DRAM and not using other programs or unnecessary services on this computer. This describes 0.001% of normal people! :lol:
Parent - By mindbreaker (****) [us] Date 2010-06-21 02:33
Hmm the engine indeed appears to be running with it checked but it does not appear to be any faster...I wonder if the OS is just ignoring the request for large pages and using normal pages.
Parent - - By mindbreaker (****) [us] Date 2010-06-21 02:56
Just verified the speed it is exactly the same.  I had it sit on the same position either way for 8 minutes and I got 272kn/s in both cases in that position.  I am doubting XP 64-bit is actually using large pages without some other tweak.
Parent - By Banned for Life (Gold) Date 2010-06-21 07:40
The other tweak is to start Rybka 4 up immediately after booting up, and to not have lots of services or other programs running at startup.

Ensuring there are no other programs or services running at startup is hard to specify in a few sentences here, but it's easy to test. If you computer takes longer than 20 seconds to boot up, you have other stuff going on that will fragment your memory before you start.

By the way, I've already posted an example of a log when I tested this myself. Worked fine and gave the expected 10-15% speedup.
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-06-21 07:40

> You should also consider adding a memory defragmentation tool.


If there are some suggestions I will add them to Nick's document.

Vas
Parent - - By Banned for Life (Gold) Date 2010-06-21 07:45
7-max's memory defragmentation option works well with XP64 IF, AND ONLY IF YOU DON"T ZERO OUT YOUR PAGING FILE (it needs the paging file to do the defrag). It's free and I think Nick has used 7-max already. I suspect 7-max won't work with Vista or W7, but I haven't tried it.

One advantage is that you can do the defrag at any time, so you wouldn't have to restrict yourself to starting up Rybka right after bootup. The disadvantage is that the defrag operation takes some time (about 30 seconds on my 4 GB machines).
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-06-21 07:55
Ok, thanks.

Vas
Parent - - By Nick (****) [gb] Date 2010-06-13 19:20
Vas an update which talks about 32-bit support, how to set the required privilege on Windows Home editions and some minor stuff.
Attachment: Rybkalargepages-draftFAQ0.2a.pdf (211k)
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-06-14 05:40
Thanks! Actually, mega-thanks!! :smile:

Vas
- - By urodets (**) [ua] Date 2010-05-27 04:51
Not sure is it right place to post, here is a quick comparison results rybka4 on linux and windows, with and without huge pages.I now that 1-CPU test is more accurate but anyway:
Windows Server 2008 R2 (1024 Mb hash, quad6600@3.2Ghz, huge pages off):

d14  nps   223298
d15  nps   231802
d16  nps   236433
d17  nps   238045
d18  nps   242555
d19  nps   254799

GentooLinux , GUI-less (1024 Mb hash, huge pages off), surprisingly!

d14  nps   228487
d15  nps   236245
d16  nps   241351
d17  nps   241094
d18  nps   248463
d19  nps   260848

Windows (1024 Mb, huge pages on)

d14  nps   251424
d15  nps   261728
d16  nps   273551
d17  nps   271371
d18  nps   278336
d19  nps   283390

Linux (1024 Mb, huge pages on)

d14  nps   248598
d15  nps   271579
d16  nps   275632
d17  nps   277491
d18  nps   287419
d19  nps   286487
Parent - - By M ANSARI (*****) [kw] Date 2010-05-27 06:54
Another way to check the performance of Large Pages is to run a determenistic test on a position using one core.  Here is a result I got with and without LP.  As you can see in this case it took 6 seconds less to do the same thing.

9442: Deep Rybka 4 x64 - Naum 4.2_8core, R4rc1_8core_1_1_gauntlet 2010
3N4/3K1P2/8/5kPb/8/b7/8/8 b - - 0 1

8                 
7                 
6                 
5                 
4                 
3                 
2                 
1                 
a  b  c  d  e  f  g  h 
Start<>EndLookup

Analysis by Deep Rybka 4 x64:

94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  +/=  (0.30 !)   Depth: 6   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.15 !)   Depth: 6   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 6   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 7   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 8   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 9   00:00:00  0kN, tb=1
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 10   00:00:00  1kN, tb=2
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 11   00:00:00  1kN, tb=3
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 12   00:00:00  7kN, tb=20
94...Kg6 95.Ke6 Kxg5 96.Nc6 Bg4+ 97.Ke5 Kg6 98.Kf4 Bh5 99.Ne5+ Kf6
  =  (-0.12 !)   Depth: 12   00:00:00  11kN, tb=41
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Bf3+ 98.Ke6 Kf8 99.g6 Be4 100.Ke5 Bxg6 101.Kf6
  =/+  (-0.27 !)   Depth: 12   00:00:00  13kN, tb=54
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bf5 100.Nf3 Be7 101.Nd4 Bg6 102.Nc6 Bxf7+ 103.Kd4 Bf6+
  =/+  (-0.67 !)   Depth: 12   00:00:00  23kN, tb=151
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg8 97.Ke6 Kg7[] 98.Kf5 Bc5 99.Kf4 Be8 100.Kf5 Ba4 101.Kf4 Bc2 102.Kg4 Bd6 103.Kf4 Bg6
  -/+  (-0.93)   Depth: 12   00:00:00  35kN, tb=237
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg8 97.Ke6 Kg7 98.Kf5 Bc1 99.Nc6 Be8 100.Ne5 Ba4 101.Kg4 Bd1+ 102.Kf5 Bc2+ 103.Kg4 Be4 104.Nd7 Bg6 105.Nc5 Be3 106.Ne6+ Kf7
  -/+  (-1.05 !)   Depth: 13   00:00:00  57kN, tb=479
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg7 97.Ke6 Bd1 98.Kf5 Bc2+ 99.Kg4 Bd6 100.Nc4 Bc7 101.Ne3 Be4 102.Nf5+ Kg6 103.Nh4+ Kf7 104.Nf5
  -/+  (-0.95)   Depth: 13   00:00:00  58kN, tb=480
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg7 97.Ke6 Bd1 98.Kf5 Bc2+ 99.Kg4 Bd6 100.Nc4 Bc7 101.Ne3 Be4 102.Nf5+ Kg6 103.Nh4+ Kf7 104.Nf5 Bd5 105.Ne3
  -/+  (-1.07 !)   Depth: 14   00:00:00  80kN, tb=797
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg7 97.Ke6 Bd1 98.Kf5 Bc2+ 99.Kg4 Bd6 100.Nc4 Bc7 101.Ne3 Be4 102.Nf5+ Kg6 103.Nh4+ Kf7 104.Nf5 Bd3 105.Nh4 Be2+
  -/+  (-1.11)   Depth: 14   00:00:00  101kN, tb=1152
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kd5 Bf5 103.Nc6 Bf6 104.Nd4 Bg6
  -/+  (-1.12)   Depth: 15   00:00:01  139kN, tb=1651
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kd5 Bf5 103.Nc6 Bf6 104.Ne5 Ke7 105.Kd4 Be6 106.Ke4 Kf8 107.Kf4 Bg7 108.Ke4 Bh6 109.Kd4 Bc1
  -/+  (-1.12)   Depth: 16   00:00:01  173kN, tb=2116
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Bf6 105.Nd6 Bg4 106.Kc5 Bg7 107.Ne8 Bh6 108.Nd6 Be6
  -/+  (-1.12)   Depth: 17   00:00:01  229kN, tb=2971
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Bf6 105.Nd6 Bg4 106.Kc5 Bg7 107.Ne8 Bh6 108.Nd6 Be6
  -/+  (-1.12)   Depth: 18   00:00:02  294kN, tb=4033
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4
  -/+  (-1.12)   Depth: 19   00:00:02  378kN, tb=5307
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4
  -/+  (-1.12)   Depth: 20   00:00:03  481kN, tb=7051
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4 107.Ne4 Bxf7 108.Nc5
  -/+  (-1.12)   Depth: 21   00:00:05  668kN, tb=10719
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4 107.Ne4 Bxf7 108.Nc5
  -/+  (-1.12)   Depth: 22   00:00:05  755kN, tb=12417
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Ba3 103.Kd5 Ke7 104.Kc6 Bg2+ 105.Kb6 Bd5
  -/+  (-1.24 !)   Depth: 23   00:00:08  1194kN, tb=21985
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Ba3 103.Kd5 Ke7 104.Ke4 Be6 105.Nc6+ Kf8 106.Ne5 Bc5 107.Kf4 Bd6 108.Ke4 Ba3 109.Kd4 Bd6
  -/+  (-1.34)   Depth: 23   00:00:12  1677kN, tb=34473
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Be7 101.Kd5 Kf6 102.Kd4 Ke6 103.Ke4 Bg2+ 104.Kd4 Bd6 105.Nc4 Bf8 106.Ne5 Be7 107.Nd3 Kxf7 108.Nf4 Bf6+ 109.Ke3 Bc6
  -+  (-1.46 !)   Depth: 24   00:00:19  2783kN, tb=63766
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Be7 101.Kd5 Kf6 102.Kd4 Ke6 103.Ke4 Bg2+
  -+  (-1.61 !)   Depth: 24   00:00:26  3849kN, tb=96356
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Ne6+ Kxf7[] 98.Nf4 Be7[] 99.g6+ Kg7[] 100.Nd5 Ba3 101.Nf4 Bb2+ 102.Ke4 Kf6 103.Ke3 Bc1+ 104.Ke4[] Bf5+ 105.Kf3[] Ba3 106.Kf2 Bc5+ 107.Kg3 Bc8 108.Kf3 Bb7+ 109.Kg4 Bd6
  -+  (-1.83)   Depth: 24   00:00:37  5191kN, tb=141824
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kd6 Kf8 101.Kd5[] Ke7 102.Ke4 Bh6 103.Kf3 Bc8 104.Ke4 Ke6 105.Nc4 Bb7+ 106.Kd3 Bf8 107.Nb6 Kxf7[] 108.Nd7 Bg7 109.Kc4 Kg6
  -+  (-1.85)   Depth: 25   00:00:51  7100kN, tb=217577

(, Microsoft 24.05.2010)

9442: Deep Rybka 4 x64 - Naum 4.2_8core, R4rc1_8core_1_1_gauntlet 2010
3N4/3K1P2/8/5kPb/8/b7/8/8 b - - 0 1

8                 
7                 
6                 
5                 
4                 
3                 
2                 
1                 
a  b  c  d  e  f  g  h 
Start<>EndLookup

Analysis by Deep Rybka 4 x64_LP:

94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  +/=  (0.30 !)   Depth: 6   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.15 !)   Depth: 6   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 6   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 7   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 8   00:00:00  0kN
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 9   00:00:00  0kN, tb=1
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 10   00:00:00  1kN, tb=2
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 11   00:00:00  1kN, tb=3
94...Bxf7 95.Nxf7 Bb2 96.g6 Kxg6 97.Ke6 Bc3 98.Nd6 Kg5 99.Kd5
  =  (0.00)   Depth: 12   00:00:00  7kN, tb=20
94...Kg6 95.Ke6 Kxg5 96.Nc6 Bg4+ 97.Ke5 Kg6 98.Kf4 Bh5 99.Ne5+ Kf6
  =  (-0.12 !)   Depth: 12   00:00:00  11kN, tb=41
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Bf3+ 98.Ke6 Kf8 99.g6 Be4 100.Ke5 Bxg6 101.Kf6
  =/+  (-0.27 !)   Depth: 12   00:00:00  13kN, tb=54
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bf5 100.Nf3 Be7 101.Nd4 Bg6 102.Nc6 Bxf7+ 103.Kd4 Bf6+
  =/+  (-0.67 !)   Depth: 12   00:00:00  23kN, tb=151
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg8 97.Ke6 Kg7[] 98.Kf5 Bc5 99.Kf4 Be8 100.Kf5 Ba4 101.Kf4 Bc2 102.Kg4 Bd6 103.Kf4 Bg6
  -/+  (-0.93)   Depth: 12   00:00:00  35kN, tb=237
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg8 97.Ke6 Kg7 98.Kf5 Bc1 99.Nc6 Be8 100.Ne5 Ba4 101.Kg4 Bd1+ 102.Kf5 Bc2+ 103.Kg4 Be4 104.Nd7 Bg6 105.Nc5 Be3 106.Ne6+ Kf7
  -/+  (-1.05 !)   Depth: 13   00:00:00  57kN, tb=479
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg7 97.Ke6 Bd1 98.Kf5 Bc2+ 99.Kg4 Bd6 100.Nc4 Bc7 101.Ne3 Be4 102.Nf5+ Kg6 103.Nh4+ Kf7 104.Nf5
  -/+  (-0.95)   Depth: 13   00:00:00  58kN, tb=480
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg7 97.Ke6 Bd1 98.Kf5 Bc2+ 99.Kg4 Bd6 100.Nc4 Bc7 101.Ne3 Be4 102.Nf5+ Kg6 103.Nh4+ Kf7 104.Nf5 Bd5 105.Ne3
  -/+  (-1.07 !)   Depth: 14   00:00:00  80kN, tb=797
94...Kg6 95.Nc6 Kxf7 96.Ne5+ Kg7 97.Ke6 Bd1 98.Kf5 Bc2+ 99.Kg4 Bd6 100.Nc4 Bc7 101.Ne3 Be4 102.Nf5+ Kg6 103.Nh4+ Kf7 104.Nf5 Bd3 105.Nh4 Be2+
  -/+  (-1.11)   Depth: 14   00:00:00  101kN, tb=1152
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kd5 Bf5 103.Nc6 Bf6 104.Nd4 Bg6
  -/+  (-1.12)   Depth: 15   00:00:00  139kN, tb=1651
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kd5 Bf5 103.Nc6 Bf6 104.Ne5 Ke7 105.Kd4 Be6 106.Ke4 Kf8 107.Kf4 Bg7 108.Ke4 Bh6 109.Kd4 Bc1
  -/+  (-1.12)   Depth: 16   00:00:01  173kN, tb=2116
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Bf6 105.Nd6 Bg4 106.Kc5 Bg7 107.Ne8 Bh6 108.Nd6 Be6
  -/+  (-1.12)   Depth: 17   00:00:01  229kN, tb=2971
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Bf6 105.Nd6 Bg4 106.Kc5 Bg7 107.Ne8 Bh6 108.Nd6 Be6
  -/+  (-1.12)   Depth: 18   00:00:01  294kN, tb=4033
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4
  -/+  (-1.12)   Depth: 19   00:00:02  378kN, tb=5307
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4
  -/+  (-1.12)   Depth: 20   00:00:03  481kN, tb=7051
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4 107.Ne4 Bxf7 108.Nc5
  -/+  (-1.12)   Depth: 21   00:00:04  668kN, tb=10719
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Bf5 103.Kd5 Bg5 104.Nc4 Ke7 105.Nd6 Be6+ 106.Kc6 Bf4 107.Ne4 Bxf7 108.Nc5
  -/+  (-1.12)   Depth: 22   00:00:05  755kN, tb=12417
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Ba3 103.Kd5 Ke7 104.Kc6 Bg2+ 105.Kb6 Bd5
  -/+  (-1.24 !)   Depth: 23   00:00:07  1194kN, tb=21985
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Kf8 101.Kd6 Be7+ 102.Kc6 Ba3 103.Kd5 Ke7 104.Ke4 Be6 105.Nc6+ Kf8 106.Ne5 Bc5 107.Kf4 Bd6 108.Ke4 Ba3 109.Kd4 Bd6
  -/+  (-1.34)   Depth: 23   00:00:10  1677kN, tb=34473
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Be7 101.Kd5 Kf6 102.Kd4 Ke6 103.Ke4 Bg2+ 104.Kd4 Bd6 105.Nc4 Bf8 106.Ne5 Be7 107.Nd3 Kxf7 108.Nf4 Bf6+ 109.Ke3 Bc6
  -+  (-1.46 !)   Depth: 24   00:00:17  2783kN, tb=63766
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5 99.Ne5 Bh3 100.Kc6 Be7 101.Kd5 Kf6 102.Kd4 Ke6 103.Ke4 Bg2+
  -+  (-1.61 !)   Depth: 24   00:00:23  3849kN, tb=96356
94...Kg6 95.Ke6 Bg4+[] 96.Ke5 Kg7 97.Ne6+ Kxf7[] 98.Nf4 Be7[] 99.g6+ Kg7[] 100.Nd5 Ba3 101.Nf4 Bb2+ 102.Ke4 Kf6 103.Ke3 Bc1+ 104.Ke4[] Bf5+ 105.Kf3[] Ba3 106.Kf2 Bc5+ 107.Kg3 Bc8 108.Kf3 Bb7+ 109.Kg4 Bd6
  -+  (-1.83)   Depth: 24   00:00:32  5191kN, tb=141824
94...Kg6 95.Ke6 Bg4+ 96.Ke5 Kg7 97.Kd5 Be7 98.Nc6 Bxg5[] 99.Ne5 Bh3 100.Kd6 Kf8 101.Kd5[] Ke7 102.Ke4 Bh6 103.Kf3 Bc8 104.Ke4 Ke6 105.Nc4 Bb7+ 106.Kd3 Bf8 107.Nb6 Kxf7[] 108.Nd7 Bg7 109.Kc4 Kg6
  -+  (-1.85)   Depth: 25   00:00:45  7100kN, tb=217577
Parent - - By Ingo (***) [nl] Date 2010-05-27 20:31
Hi

I dont want to spoil your fun but just some numbers:

The official estimation for a speed increase for large pages is 15%, the best real number up in that thread was 13% in time to depth increase.

Assuming 60 Elo from R3 to R4 13% is ~8Elo, 15% might be ~9Elo let s say you might be able to win 10Elo.

I fear that it is impossible to see, not to talk about prove, that increase!

Bye
Ingo
Parent - - By Uly (Gold) [mx] Date 2010-05-27 21:00

> 15% might be ~9Elo


Assuming 70 elo per doubling it would actually be (sic) 15 elo.
Parent - By urodets (**) [ua] Date 2010-05-28 08:33
The advantage of huge pages is not only in ~15% speed up, rybka with large hash calculating somehow different compared to ussual, i made dozens of test (on Linux with R3) and it is working better (search, variations, etc)
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-28 08:59
Enabling large pages has the same effect like overclocking your 3 GHz computer to 3.45 GHz, but at no cost.
Parent - - By Dragon Mist (****) [hr] Date 2010-05-28 09:08
Ok, but if I load R4 with large pages, and let it run day in and day out without interruption, and at the same time I use the PC occasionaly for other stuff, mail, internet, word processing etc, will the memory become fragmented once again after a while and does this affects R4?
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-28 09:14
If you don't unload Rybka (in infinite analysis and in normal engines matches you don't), she will keep the same large pages and will work at full speed.
Parent - By Dragon Mist (****) [hr] Date 2010-05-28 09:21
Thanks, that's what I needed to know.
Parent - - By Master Om (*****) [in] Date 2010-05-28 11:36
For optimal performance how to configure page file manually for large pages ? or let windows handle page file ?
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-28 11:47
I prefer to have no page file, but I haven't tried different settings with Rybka + large pages.
Parent - - By Master Om (*****) [in] Date 2010-05-28 13:06
When i set to 512 mb Rybka uses large pages. 1024 mb hash and it does not use. I have 4 GB RAM .
Parent - - By urodets (**) [ua] Date 2010-05-28 13:09
Are you running something else with rybka+GUI? (antivirus, firewalls, other memory hog proccess).
Parent - By Master Om (*****) [in] Date 2010-05-28 13:24
Dont know. But i dont run anything when i run chess. :-)
Parent - - By Ray (****) Date 2010-05-28 11:58

> If you don't unload Rybka (in infinite analysis and in normal engines matches you don't), she will keep the same large pages and will work at full speed.


Yes - but when you go to fire up other applications, you may get "not enough memory" error messages, depending on how much RAM you have. I tried this a few years back, large pages were impossible on a machine with 4GB of RAM.
Parent - By Dragon Mist (****) [hr] Date 2010-05-28 12:07
Well, possible for this particular machine with 4GB of RAM so far.
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-28 12:20
It's always good to have seperated computers. One for internet and office - many for ches ;)
Parent - By Dragon Mist (****) [hr] Date 2010-05-28 12:35
That was my idea, too, but later on, I figured what the hell, all my machinery runs primarily chess, everything else just gets squeezed in between :-)
- - By Uly (Gold) [mx] Date 2010-05-29 06:24
Note for people with low RAM: Have Rybka 4 loaded on Startup, if you fire up Rybka 4 after you've done other things, Large Pages won't be usable.
Parent - - By Bobby C (****) Date 2010-05-29 07:23
Is 4gb considered "low ram"?
Parent - By Uly (Gold) [mx] Date 2010-05-29 08:03
No, I'm using 1GB and it seems low enough to not let me load large pages unless it's the first thing I do at startup.
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-29 11:57
Yes. For a quad I recommend 8 GB, for 8 cores 12 GB, for 12 cores 16 GB
Parent - - By mrolin (*) [fr] Date 2010-05-29 14:07
On my quad I7 8Gb only 1Gb hash with LP is possible.
It's a pity not knowing if LP is Ok or not without looking at log file.
Why not "4 CPUs LP" ?
Parent - - By bnc (***) Date 2010-05-29 15:04

> On my quad I7 8Gb only 1Gb hash with LP is possible.
> It's a pity not knowing if LP is Ok or not without looking at log file.
> Why not "4 CPUs LP" ?


If you use the ChessBase GUI you might want to try and enable the window pane named "Engine Output" . When using GUI 12, it will display a message that indicates whether Large Pages worked or not.
Under GUI11, it seems I had to disable Large Pages and then enable Large Pages before I could see a message in the "Engine Output" window.

On my system with a I7-920 quad and 6gb RAM, I could get Large Pages to work after a boot with a 512gb hash and no 6man EGTBs. Once I tried to increase hash or add EGTBs the Large Page option failed. I am considering adding some RAM  :)
Parent - By mrolin (*) [fr] Date 2010-05-29 15:21
Thanks
I have to wait then end of an engine match i launched (20 games at 25mn) to try it
Parent - - By mrolin (*) [fr] Date 2010-05-30 04:31 Edited 2010-05-30 05:27
yes it's Ok in the main screen with "view, engine text or engine output as you say" (translate from french) but for engines match no way to know but log.
After my engine match i had to reboot for using LP succesfully : 1Gb + EGTBs Gui DF12 Win 7.
LP option is too mutch sensitive to not have a simple way to know if it's OK or not.
With Shedder12 Gui it seems a little better. Creating an engine with LP, and when loading for engine match LP status is displayed.
Parent - By bnc (***) Date 2010-05-30 10:45

> yes it's Ok in the main screen with "view, engine text or engine output as you say" (translate from french) but for engines match no way to know but log.
> After my engine match i had to reboot for using LP succesfully : 1Gb + EGTBs Gui DF12 Win 7.
> LP option is too mutch sensitive to not have a simple way to know if it's OK or not.
> With Shedder12 Gui it seems a little better. Creating an engine with LP, and when loading for engine match LP status is displayed.


Very true, the Engine Output window is not available during an engine match. Another way to verify LP usage is to use the Windows Task Manager.
The memory allocations for the Rybka processes are much smaller if LP is in effect.
Parent - - By Ahmed Mansour (*) [us] Date 2010-05-29 17:07
Hi Kullberg

below is a print from log, can you tell if 'large pages' and other setting s are fine

Much Obliged

           Sat May 29 21:01:35 2010
           "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe"
          
           8 processors, x64 3325 MHz
           Intel(R) Xeon(R) CPU           W5590  @ 3.33GHz, Intel64 Family 6 Model 26 Stepping 5, GenuineIntel, level 6, revision 0x1a05
           98301 mb RAM, 92358 mb available, page size = 4096
                Windows 7 x64
                Process priority: NORMAL_PRIORITY_CLASS
          
                Nalimov path set to C:\autoTB\TB4;C:\autoTB\TB5;C:\autoTB\TB6
                Nalimov cache set to 8
               Large pages not used - user choice.
          
      -->> setoption name xLog value false
      -->> setoption name Max CPUs value 7
                  -- shared synchronization file: RybkaSync12096
                  -- shared exchange file: RybkaEx22096
           Large pages not used - user choice
                  starting child
                  -- hash size: 32
                  -- shared file number: 2
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 6
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 2 7 6 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 2
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 5
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 2 7 5 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 2
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 4
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 2 7 4 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 2
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 3
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 2 7 3 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 2
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 2
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 2 7 2 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 2
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 1
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 2 7 1 2096
                Children restarted.
                Max CPUs set to 7
      -->> setoption name CPU Usage value 100
                CPU usage set to 100
      -->> setoption name NalimovPath value J:\TB
   2            Nalimov path set to J:\TB
      -->> setoption name NalimovCache value 256
                Nalimov cache set to 256
      -->> setoption name NalimovUsage value Rarely
                Nalimov usage set to rarely
      -->> setoption name Main Process Priority value Inherit
                Main process priority set to inherit.
      -->> setoption name Child Process Priority value BelowNormal
                Child process priority set to below-normal.
      -->> setoption name Use Large Pages value true
                Use large pages enabled.
                  -- shared synchronization file: RybkaSync12096
                  -- shared exchange file: RybkaEx32096
           Using large pages, size = 33554432
                Using large pages, size = 33554432
                  starting child
                  -- hash size: 32
                  -- shared file number: 3
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 6
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 3 7 6 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 3
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 5
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 3 7 5 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 3
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 4
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 3 7 4 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 3
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 3
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 3 7 3 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 3
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 2
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 3 7 2 2096
                  starting child
                  -- hash size: 32
                  -- shared file number: 3
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 1
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 32 3 7 1 2096
   3            Children restarted.
      -->> setoption name Always Score Main Move value false
                Always score main move disabled.
      -->> setoption name Display Upperbounds value false
                Display upperbounds disabled.
      -->> setoption name Preserve Analysis value false
                Preserve analysis disabled.
      -->> setoption name Hash File value <empty>
                Hash file set to <empty>
      -->> setoption name Ponder value false
      -->> setoption name TC Buffer value 3
                TC Buffer set to 3 seconds
      -->> setoption name TC Normal Move Time value 8
                TC Normal Move Time set to 8
      -->> setoption name TC Max Move Time value 10
                TC Max Move Time set to 10
      -->> setoption name White Pawn cp value 0
                White pawn adjustment value set to 0 cp
      -->> setoption name Black Pawn cp value 0
                Black pawn adjustment value set to 0 cp
      -->> setoption name White Knight cp value 0
                White knight adjustment value set to 0 cp
      -->> setoption name Black Knight cp value 0
                Black knight adjustment value set to 0 cp
      -->> setoption name White King Bishop cp value 0
                White king bishop adjustment value set to 0 cp
      -->> setoption name Black King Bishop cp value 0
                Black king bishop adjustment value set to 0 cp
      -->> setoption name White Queen Bishop cp value 0
                White queen bishop adjustment value set to 0 cp
      -->> setoption name Black Queen Bishop cp value 0
                Black queen bishop adjustment value set to 0 cp
      -->> setoption name White Rook cp value 0
                White rook adjustment value set to 0 cp
      -->> setoption name Black Rook cp value 0
                Black rook adjustment value set to 0 cp
      -->> setoption name White Queen cp value 0
                White queen adjustment value set to 0 cp
      -->> setoption name Black Queen cp value 0
                Black queen adjustment value set to 0 cp
      -->> setoption name Rook Endgame Scaling value 100
                Rook endgame scaling set to 100 percent.
      -->> setoption name UCI_EngineAbout value www.rybkachess.com
      -->> setoption name Hash value 16000
   5              -- shared synchronization file: RybkaSync12096
                  -- shared exchange file: RybkaEx42096
           Using large pages, size = 8589934592
                Using large pages, size = 8589934592
   7              starting child
                  -- hash size: 8192
                  -- shared file number: 4
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 6
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 4 7 6 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 4
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 5
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 4 7 5 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 4
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 4
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 4 7 4 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 4
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 3
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 4 7 3 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 4
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 2
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 4 7 2 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 4
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 1
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 4 7 1 2096
                Children restarted.
                Hash size set to 8192 mb
      -->> isready
           !!!! readyok
      -->> setoption name Hash value 16000
  10              -- shared synchronization file: RybkaSync12096
                  -- shared exchange file: RybkaEx52096
  11       Using large pages, size = 8589934592
                Using large pages, size = 8589934592
  13              starting child
                  -- hash size: 8192
                  -- shared file number: 5
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 6
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 5 7 6 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 5
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 5
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 5 7 5 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 5
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 4
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 5 7 4 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 5
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 3
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 5 7 3 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 5
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 2
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 5 7 2 2096
                  starting child
                  -- hash size: 8192
                  -- shared file number: 5
                  -- number of processes: 7
                  -- parent id: 2096
                  -- overridden parent id: 2096
                  -- child number: 1
                  -- full command line: "J:\Aquarium2010\Engines\Rybka\sse42\Deep Rybka 4 SSE42 x64.exe" 2096 7 8192 5 7 1 2096
                Children restarted.
                Hash size set to 8192 mb
      -->> isready
           !!!! readyok
      -->> setoption name UCI_AnalyseMode value true
      -->> position startpos

                             (
rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -
)

            r n b q k b n r
            p p p p p p p p
            . . . . . . . .
            . . . . . . . .
            . . . . . . . .
            . . . . . . . .
            D D D D D D D D
            R N B Q K B N R

      -->> go infinite
      -->> stop
             .. stop requested (is searching: true)
      <<-- info depth -2 score cp 48 time 1 nodes 24 nps 24000 pv b1c3
      <<-- bestmove b1c3
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-29 17:24
Large pages are used - hash is 8 GB - everything OK
Parent - - By dragon49 (****) [us] Date 2010-05-30 02:00
I understand the 15% speedup.  Vas, Lucas, please translate this speedup to ELO gain - How many points?  This is all that interests me.

Thanks
Parent - By Dragon Mist (****) [hr] Date 2010-05-30 05:47
It's around 10 ELO.
Parent - - By hal9000 (**) [no] Date 2010-05-30 14:50
Do you know if there's any significant performance hit when using the log during blitz games? The log can be very useful for verifying that large pages is being used, but if it reduces performance I don't want to use it for blitz.
Parent - By dragon49 (****) [us] Date 2010-05-30 15:15
After the weekend, I m going to use the Fritz 12 Gui to have Rybka 4 run a gauntlet against my other (around 50) engines.  It is my understanding that I should use "Large Pages" if Rybka is not unloaded after each game, but not use the option if if it is unloaded after each game.  How does Fritz manage the main engine running the gauntlet?

Thanks
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-30 17:14
I'd switch it off. The task manager can tell you if they are being used.
Parent - By mocha1961 (***) [us] Date 2010-05-30 17:20
what in task manager do i look for? im using vista 64
Parent - - By hal9000 (**) [no] Date 2010-05-30 18:10
I've taken a look in the task manager, but I don't see any significant difference in the memory statistics. Which values are you talking about? I use Windows 7 Pro x64.
Parent - - By Uly (Gold) [mx] Date 2010-05-30 22:42
Rybka should report less RAM that she's really using if LP are working.
Parent - - By cma6 (****) Date 2010-05-30 23:59 Edited 2010-05-31 00:07
I have Win XP Pro x-64. Task Manager shows 2.1 gig (out of 2048 hash size) being used for all 4 cores.
Whether "Large Pages" is checked or not, Task Manager shows the same thing.
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2010-05-31 12:06
This indicates that large pages are not used.
Parent - - By cma6 (****) Date 2010-05-31 14:04
Kullberg, thanks. Then does it seem to you that Windows XP Pro x-64 does not after all allow Large Pages?
Up Topic Rybka Support & Discussion / Rybka Support / Large Pages in Rybka 4
1 2 3 4 5 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill