M ANSARI wrote:To be honest it would not make much of difference for a human if Rybka 3 is on single core i7 or Octal. For engine vs. engine match it would make a huge difference but humans would probably not be able to tell the difference of play between the two platforms. R3 on Athlon 1200 32bit is rated over 3000 ... A single core i7 Nehalem running at the fastest available speed which is 3.2 ghz and using 64 bit orders of magnitude stronger. I think if a match were to be done vs. a human the results of a powerful single core i7 setup and a 40 core cluster would be pretty much the same.
Usually a human loses due to tactics, and his only chance of surviving is to try to close the position as much as possible. If he is able to do that then a 1000 core cluster will make no difference to a single core cluster ... at least not until some form of Monte Carlo module is integrated into the cluster setup.
One more thing I would like to add ... the players that would fare best against computers do not necessarily have to be the top 5 players in the world. I wouldn't be surprised if some lower ranked GM's or even some IM's might have better results due to being able to know how to close positions.
I have to agree with you on almost everything that you stated here, and I also wonder if a chess program after reaching certain speed the benefit of increasing its clock speed really affect the performance against strong human players, unless it is programmed to sacrifice a pawn or two in order to break a closed position. I also noticed that when you provide a strong GM with slower time control instead of faster blitz games, the human GM is able to become more creative and less prone to commit fatal tactical errors.
PS: Usually a human loses due to tactics mostly when playing blitz games. Most games that Anand and Kasparov have lost versus chess programs like Rebel and Genius 3 were at a faster blitz time control, due to time constrain.
M ANSARI wrote:To be honest it would not make much of difference for a human if Rybka 3 is on single core or Octal. For engine vs. engine match it would make a huge difference but humans would probably not be able to tell the difference of play between the two platforms. R3 on Athlon 1200 32bit is rated over 3000 ... A single core Nehalem running at the fastest available speed which is 3.2 ghz and using 64 bit orders of magnitude stronger. I think if a match were to be done vs. a human the results of a powerful single core setup and a 40 core cluster would be pretty much the same.
Usually a human loses due to tactics, and his only chance of surviving is to try to close the position as much as possible. If he is able to do that then a 1000 core cluster will make no difference to a single core cluster ... at least not until some form of Monte Carlo module is integrated into the cluster setup.
One more thing I would like to add ... the players that would fare best against computers do not necessarily have to be the top 5 players in the world. I wouldn't be surprised if some lower ranked GM's or even some IM's might have better results due to being able to know how to close positions.
I disagree that usually a human loses due to tactics.
computers win against humans mainly because of the fact that computers are better positional players than humans and I expect chess programs to beat humans even if you give humans the right to take back mistakes that lose material based on computer analysis(mistakes that do not lose material but lose 0.2 or 0.3 pawns based on computer analysis are positional mistakes and not tactical mistakes).
a strong gm top 2700 play versus a rybka ( small book limit and and small end games limit, only his algoritm... ) have some options today...
maybe rybka 2975 points and some structure have don´t undestand very well ? down to 2600...
if A human have "4 draws" versus rybka, do this mean that the human have = elo rybka ?
of course NOT ¡ and rybka need play versus humans to see his real LEVEL... rybka win a lot of games versus the SAME engines, and it is UP the elo.
if a GM 2700 play 10.000 games versus MI that GM up +3000 elo too ¡ but only rybka have enough time to play 10.000 games versus a "very" down engine level compare... ¡
rybka is not perfect ¡ elo rating ARE over-estimation... because the others engines have a elo inflate to ¡
If a GM work hard, rybka "don´t" win too easy... and that GM have a lot of points of ELo rybka...
It´s necesary do other estimation ratings... elo in vey imprecise, because rybka always wins versus the similar engines...
Uri Blass wrote:Milov could not beat rybka in equal conditions and only got one draw and one loss in normal chess with white.
I think that if GM Milov prepare a little bit longer and get use to play against Rybka he can compete better. I kind of believe in his last statement where he stated that Rybka is NOT unbeatable under normal condition and with plenty of preparation.
Dr.Wael Deeb wrote:I am astonished from the fact that you can't accept a single simple fact:
Humans have lost the war with the computer chess software,try to live with this fact....
Doc,
humans do have lost the war against the computer chess hardware, not against the computer chess software.
Let's be correct
PS:
My opinion is that matches Human vs. Computer have to be held on the average home PC -- that the average customer can see the real strength of software he bought.
Dr.Wael Deeb wrote:I am astonished from the fact that you can't accept a single simple fact:
Humans have lost the war with the computer chess software,try to live with this fact....
Doc,
humans do have lost the war against the computer chess hardware, not against the computer chess software.
Let's be correct
PS: My opinion is that matches Human vs. Computer have to be held on the average home PC -- that the average customer can see the real strength of software he bought.
Correct,I can't agree more
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
The problem is that the average home PC will soon be a computer running a less than $300 Nehalem Core i7 processor ... which is very close to the performance of an 8 core Skulltrail.
M ANSARI wrote:The problem is that the average home PC will soon be a computer running a less than $300 Nehalem Core i7 processor ... which is very close to the performance of an 8 core Skulltrail.
IMHO, you highly overestimate today's average home PC
