Author: Christophe Theron
Date: 21:22:49 04/10/03
Go up one level in this thread
On April 10, 2003 at 23:11:44, Jay Urbanski wrote:
>>I could produce a commercial multiprocessor version of Chess Tiger some day, but
>>I still see no interest to do it at this time.
>
>Dont' say *NO* interest if you mean no commercial interest (as opposed to no
>interest on your part). *I* am interested, and I would buy an SMP version of
>Tiger. :) Given the number of requests on here I would say I'm not alone.
>
>>I don't even see an interest in the foreseeable future. I don't believe in
>>multiprocessing as a major trend.
>
>I have to strongly disagree here. If you look at the *server* market you will
>find that SMP systems dominate. All but the lowest of the low-end servers on
>the market are at least dual capable. Dual-processor systems are the sweet spot
>of the Intel-compatible market.
>
>I could make the case pretty strongly that when you are discussing chess
>engines, you are really talking about *servers* even though most people run them
>on desktop machines. If you accept as your goal having the strongest possible
>chess-playing machine, then you need to go SMP.
My goal is to have the strongest chess program running on single processor
computers. Whatever the size of these computers.
> If you are going to enter a
>chess-playing system into a GM-level tournament then you want to do it using the
>fastest hardware available, and that implies SMP.
That is not what is required to beat grandmasters.
Speed at some point becomes irrelevant when you play against grandmasters. The
speed of the current fastest PCs is more than enough to beat grandmasters. But
current PC programs are not able to achieve this result.
Take a current program and multiply its speed by 4 or 8 and you won't get
significantly better results against grandmasters anyway.
This race for computing power has reached a point of complete futility.
What is needed is more intelligence in the program itself.
The current state of the matches man vs machines is just telling us the vast
superiority of the human intelligence.
What we do with these matches is comparable to organizing a race between a
runner and a Ferrari. We tell them "Go!" and after one kilometer the Ferrari is
ahead. So everybody agrees on the superiority of the car. But if you look
closer, you notice that the car is ahead... by ONE METER.
That's exactly what's happening in man vs machine in chess at this time. We are
using an incredibly sophisticated technology and we achieve a ridiculous result.
Increase the power of the computers by multiplying the number of processors and
all you are doing is to make this even more ridiculous.
My goal is not to hide the stupidity of chess programs by increasing the
computing power until the victory is reached.
My goal is to increase the intelligence of my program which is already running
in my opinion on a sufficiently powerful hardware to achieve the result. And
anyway the average computer gets more powerful every year.
>>I could even add that I don't believe in 64 bits computing either. For me the
>>future of computing in the next years is putting 32 bits single processor
>>computers everywhere (including your watch and your coffee machine).
>
>It all depends on what you are doing. For many applications right now, 32-bit
>is adequate. But the average amount of memory installed on even desktop systems
>is quickly approaching the 4GB limit. Remember that with a 32-bit OS you don't
>even get the full 4GB of address space for applications because that must be
>shared with the kernel. I predict AMD Opteron will make 64-bit desktops common
>by 2004 and that Intel will be forced to reply with Yamhill.
>
>For the server market there are plenty of applications that need 64-bit today
>(and yesterday)
>
>>I agree that a multiprocessor version would be significantly stronger, but on
>>the other hand I have many other high priority tasks that would benefit to a lot
>>more people, and for me it's important.
>>
>>Example of such tasks:
>>* Chess Tiger 16
>>* A Linux version of Chess Tiger
>>* A native ARM version of Chess Tiger for Palm
>>* ...and a few more projects that I prefer to keep secret
>>
>>So I'm very sorry to tell you that you should not hold your breath...
>
>This is of course up to you. Of course whenever these man-vs-machine matches
>are held you can rest assured that Tiger won't be holding up the machine part of
>the battle because it can't run on the fastest hardware out there.
These man-vs-machine are marketing events. The requirement to participate, as we
have seen in the past, is not to have the best program.
I'll probably get my man-vs-machine event some day, with or without SMP.
Actually I don't care if I never get one. It's boring anyway.
Christophe
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.