[Beowulf] commercial clusters

[Vincent Diepeveen]

----- Original Message ----- 
From: "Geoff Jacobs" <gdjacobs at gmail.com>
To: "Vincent Diepeveen" <diep at xs4all.nl>
Cc: "Robert G. Brown" <rgb at phy.duke.edu>; "Jim Lux" 
<James.P.Lux at jpl.nasa.gov>; <beowulf at beowulf.org>; "Angel Dimitrov" 
<stormlaboratory at yahoo.com>
Sent: Friday, September 29, 2006 9:19 PM
Subject: Re: [Beowulf] commercial clusters


> Vincent Diepeveen wrote:
>> The basic flaw i see in some assumptions taken here is that you as a
>> human want
>> to do a running contest with a car.
>>
>> In reality that isn't the case.
>>
>> People just want to see whose car is better, not whether they can outrun
>> the car.
> Your analogy is poor. Any computer playing chess must conform to the
> same rules as a human player. A better analogy would be a foot-race
> between a human and a bipedal robot. Do you think 100k Ordinary
> People[tm] would be willing to lay down $10k for such a curiosity?

Well the hard facts is that hundreds of thousands of $3000-$20k 
chesscomputers (dedicated with a 10Mhz chip) have been sold until the pc was 
faster than the dedicated chesscomputers. A vaste multiple indeed of that of 
$50-$1000 computers has been sold at the time.

Around 1983 or so when the dutch programmers started producing dedicated 
computers they handsdown kicked 99.9% of the population already, so that's 
just not the question here,
to counter RGB.

By now it's not 99.9% but more like that it is a handful who make a chance 
in a match and just a couple of thousands on the planet who now and then can 
win a game from the machine.

The first parallel pc chess software was crafty in 1998, followed by me in 
1999.
First commercial sales were done by chessbase around 1999 of deepfritz6.

Amazingly world title 1999 went still to a single cpu program (Shredder), 
despite MiT joining for example with Cilkchess at 512 cpu's origin3800 
(other participants like p.conners at 200+ node dual P2 machine, zugzwang at 
a 512 cpu T3E etc).

Nowadays all strong pc programs are also SMP. Not all of them run that well 
on quad machines though and majority of them has a very poor speedup out of 
4 cores already (most sit around 2.0 - 2.4).

So the move to low latency clusters now is logical.

Vincent

>> Additional they want to analyze their own running (most important
>> feature) in the most objective manner possible, or they want to compete
>> with it in a 300 KM/h contest.
>>
>> To look like a winner, people are prepared to pay any price.
> If this were the case, we'd all drive Bugatti Veyrons and Porsche GT1s.
>
>> Vincent
>>
>>
>
>
> -- 
> Geoffrey D. Jacobs
>
> Go to the Chinese Restaurant,
> Order the Special
>