Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Ply Depth in relation to Elo again...

Author: Robert Hyatt

Date: 11:37:07 05/23/00

Go up one level in this thread


On May 23, 2000 at 12:41:24, Joshua Lee wrote:

>
>
>
>>>Deeper Blue  1997 (15at a minimum)17-18 Ply 2817(USCF)200 MNPS
>2767 or 2717FIDE     I just want to know if these numbers were acurate and which
>ones were not fix them if you knew the correct ratings. This way we can get an
>idea of where PC's Lie compared to Deep Thought ...
>
>Deeper Blue is twice as fast as Deep Blue obviously but what would the
>equivalent processor speed have to be to match it? 2Thz like i said or no?
>Deep Blue   1996 (14Ply Min 16-17 Ply  2642(FIDE?)100 MNps

That is hard to say.  I would compare DB's speed (200M nodes per second),
and then find the very slowest commercial program you can find, maybe
Hiarcs, and use that to estimate the speed of DB.  DB's eval was very
sophisticated (even if not tuned particularly well in some cases of course).
If DB did 200M, and hiarcs can do 200K on a 1ghz machine, then 2thz is about
right.

Roughly guessing of course...

And remember, 200M was the "nominal" speed.  It was capable of hitting 1B
nodes per second when all the hardware was busy, but this took deep depths
to begin to touch that number.  I think I read that they could actually reach
70% of max in a real game, so 700M would be roughly the max speed.



>
>looking at it now twice as fast netted Deeper Blue 1 ply
>512???Ghz would be 50MNps  and (13Ply Minimum)15 Ply so
> 2642-50= 2592or -60 =2582 or is it -90 =2552???
>Here is where it get's tricky because we know Deeper Blue's rating , we Know
>Deep Blue's rating so when we go to the Processing power that is 50% lower
>going from 1024 to 512  we though that the rating is 50 or 60 or 90 points
>I proved beyond a shadow of a doubt that this cannot be true in every case
>because the ratings get junky down the list.
>
>
>>This is a problem.  I personally believe that DB is on the order of 300 points
>>better than the next best computer.  But this is in head-to-head play with
>>computers.  Whether it is that much stronger if both programs were playing only
>>humans is another issue.
>>
>>Computer vs computer exaggerates search speed and missing knowledge.  Against
>>humans this isn't as important, overall...
>
>Deep Blue is 2642 and looks 16-17 Ply (14 Ply minimum) 2642(FIDE
>128Ghz  12.5Mnps   (11PlyMin)   13Ply   2492     2462             2372
>Deep Thought II        1991                  6-7 Million Nps 64Ghz?
>64Ghz    6.25Mnps      (10Ply)     12Ply    2442     2402            2282
>
>Can't be because it would make the PC's even lower besides gary Kasparov was the
>one who said that Deep Thought in 1989 was 2480-2500FIDE
>300 points would make the rest of the list even lower if i knew the hardware
>then that would make a difference. also I played over the 89 game and at 12Ply
>in analysis it prefered 39.Kg1 to g3 -0.95 and gave Deep Thoughts move g3 as
>-1.16.  ofcourse this took maybe 3 minutes on their hardware and took atleast an
>hour on mine it says something about how strong the eval is in PC programs.
>
>You said : doubling the cpu speed is generally said to produce 50-60 rating
>points.Since the typical effective branching factor is around 3.0, every time
>the speed is tripled, we get another ply, and using the 60 point figure above,
>a ply would be about 90 rating points, roughly.  But there is nothing that
>says that as we go deeper this doesn't taper off.  Nor is there anything that
>says that as we go deeper, the gain doesn't actually get larger...
>
>Keeping your  last two sentences in mind...
>
>
>                                         -50    -60          -90
>2.048Thz 200 Mnps 15Ply Min  17Ply     2692     2702         2732 <---??
>
>Deeper Blue                1997   (15at a minimum)17-18 Ply  2817 (USCF)
> 200 MNPS 2.048Thz  2817USCF -50 or -100   2767 or 2717FIDE
>Deep Blue                    1996             16-17 Ply (*)14 Ply 2642(FIDE
>(*2535 Positional Play Strength?) 100 MNps 1.024Thz?? 32node *6 =192 processors
>1.024Thz 100 Mnps 14PlyMin   16Ply     2642     2642            2642
>512Ghz  50Mnps   13Ply Min       15Ply   2592       2582           2552
>256Ghz  25Mnps    (12Ply Min)    14Ply  2542      2522            2462
>128Ghz  12.5Mnps   (11PlyMin)   13Ply   2492     2462             2372
>Deep Thought II        1991                  6-7 Million Nps 64Ghz?
>
>64Ghz    6.25Mnps      (10Ply)     12Ply    2442     2402            2282
>32Ghz   3.125Mnps    (9Ply)        11Ply     2392    2342             2192
>Deep Thought            1989                2480-2500 (FIDE) 2 Mnps 20Ghz?*
>16Ghz  1.562.500nps    (8)           10Ply   2342     2282             2102
>8Ghz     781,250nps       (7)         9Ply     2292      2222             2012
>Deep Thought 0.02    1988     2551(USCF) 2501 or  2451(FIDE?)750KNps* 7.453Ghz
>In 1988 DEEP THOUGHT and Grandmaster Tony Miles shared first place in the U.S.
>Open championship. DEEP THOUGHT had a 2745 performance rating. 2745-180 = 2565?
>don't you subtract 180 to get a true rating from a performance one?
>4Ghz     390,625nps      (6)          8Ply      2242    2162              1922
>Hitech       1995              200KNps
>2Ghz    195,312nps         (5)           7Ply      2192    2102
>1832
>Hitech                          1988 2413(USCF) 2363 or 2313FIDE?   175 KNps
>In 1985 HITECH achieved a performace rating of 2530.  2530-180 = 2350
>Belle                             1983 2208         160KNps
>1Ghz    97,656nps           (4)            6Ply     2142    2042 1742
>Hiarcs 7.32                  2000           98-134KNps  1Ghz
>Hiarcs 7.32                  2000 2660(LCTII)2460(FIDE est)87-123KNps 880AMD
>Belle                             1983 2208         160KNps
>
> Hiarcs looks at a Minimum of 7 usually 8 but on average 9 or 10 sometimes 11Ply
>in the middlegame on my computer which is 880Athlon
>
>I swear i read that Deep Thought searched 12Ply in 1989 against Kasparov
>I don't know if this is so and my other thought is how much of a difference does
>the extensions Hiarcs gives in a position i.e if it say's that it is looking
>9/30ply
>how does that compare with Deep Blue saying it is looking 16Ply?
>anotherwords in the same position is 16 Hiarcs ply the same as 16 Deep Blue ply
>maybe that is the key to the differences in the list above.
>thankyou


I can guarantee you that when DB says 16 plies, they are searching many branches
to at least double that, because of singular extensions, and all the other
search extensions they use.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.