Which CPUs was used against Kramnik in 2002 ?

[Vinvin]

Is there any official information about that on the Internet ?

I found "Fritz ran on four Intel Pentium 4 Xeon CPUs at 2.8 GHz." from https://en.wikipedia.org/wiki/X3D_Fritz
and "8 processors at 2.4 Ghz + 256 GB memory" in French https://fr.wikipedia.org/wiki/Fritz_(pr ... 3%A9checs)

What's the exact model of this Intel processor ?

[Joost Buijs]

I think it was an 8 processor system based on large cache Pentium III Xeons running at 900 MHz. Nothing like 2.4 GHz.
It costed an arm and a leg, but the Sheik could probably afford it.

[Milos]

I think it was an 8 processor system based on large cache Pentium III Xeons running at 900 MHz. Nothing like 2.4 GHz.
It costed an arm and a leg, but the Sheik could probably afford it.

Since Prestonia Xeon 2.0A (based on Nortwood architecture) was released in early 2002 and ran on 2GHz, I seriously doubt they would had played on such an old and outdated hardware as P3 Xeons. Moreover, its retail price was only 417 USD so hardly it costed "an arm and a leg".

[Rom77]

"Computerschach und Spiele" #5, 2002, page 10:

Die Hardware, auf der Deep Fritz läuft, ist ein 8 x 900 MHz schneller Compaq mit vier GB Speicher. Nicht das allerschnellste, was es heute gibt, aber sicher wesentlich leistungsfähiger als die gängigen Dualsysteme. Der Rechner, der etwas dicker als ein Standard-PC ist, aber um ein Vielfaches schwerer, wurde Mitte September in das Trainingscamp von Vladimir Kramnik geliefert. Mathias Feist installierte dort die letzte Version von Deep Fritz (er war über die beobachteten 3,5 Millionen Stellungen in der Sekunde nur mäßig begeistert) und Kramnik konnte mit seinen Sekundanten an die Arbeit gehen.

The hardware on which Deep Fritz runs is an 8 x 900 MHz Compaq with four GB of memory. Not the fastest thing there is today, but certainly much more powerful than the common dual systems. The computer, which is slightly thicker than a standard PC, but much heavier, was delivered to Vladimir Kramnik's training camp in mid-September. Mathias Feist installed the last version of Deep Fritz (he was only moderately enthusiastic about the observed 3.5 million positions per second) and Kramnik was able to work with his seconds.

"Computerschach und Spiele" #6, 2002, page 10:

Der 8 x 900 MHz Rechner von Compaq

The 8 x 900 MHz computer from Compaq

"Computerschach und Spiele" #6, 2002, page 19:

In Bahrain erzielte das Programm auf dem verwendeten achtmal 900-MHz-Rechner Suchtiefen von 40 Halbzügen (=20 Zügen) in einigen besonders genau betrachteten Abspielen.

In Bahrain, the program achieved search depths of 40 half-moves (= 20 moves) on the eight times 900 MHz computer used in some particularly carefully considered plays.

[Ajedrecista]

Hello Vincent:

I have found the following information: firstly under CPW, there is a link to Selective Search computer magazine (page 9 of that issue):

http://www.chesscomputeruk.com/SS_103.pdf

The greatly anticipated million dollar 'Brains in Bahrain' match between 2807 rated Vladimir Kramnik and an 8 processor (Xeon 933 Mhz's each I believe) Deep Fritz 7 finally took place in October.

Please note the 'I believe', so it is not totally sure.

Secondly, under the official web of the event via The Week in Chess 339 (7th May 2001):

https://web.archive.org/web/20011223221 ... ents&p=mvm

Code: Select all

KRAMNIK VS DEEP FRITZ: HEAD TO HEAD

     Vladimir Kramnik                                Deep Fritz
            26                     AGE                   10
          RUSSIAN              NATIONALITY             GERMAN
           6'4"                   HEIGHT                12CM
          95 KG                   WEIGHT                50 G
      1 MOVE/SECOND               SPEED        6,000,000 MOVES/SECOND
BRAIN - 100 BILLION NEURONS    PROCESSORS            8 PENTIUMS

Finally, there is some info in the same site concerning the engine qualification between Deep Fritz and Deep Junior, where both played on identical powerful Pentium III dual processing 933 MHz computers:

https://web.archive.org/web/20020109144 ... p=mvm_deep

To win its position to challenge Kramnik this October, Deep Fritz beat rival software Deep Junior by contesting 25 games on identical powerful Pentium III dual processing 933 MHz computers. IBM was invited to pitch its formidable Deep Blue against Deep Fritz for this contest, but declined to do so.

------------

More references to 6,000,000 positions per second:

https://web.archive.org/web/20021204234 ... com/about/

[...]

Deep Fritz 7 unveiled to play Kramnik in Bahrain
6,000,000 positions per second, running on a XXXMHz Pentium Xx

Almost!

https://web.archive.org/web/20021205053 ... stats.html

2001: October - Deep Fritz 7 to be launched for the Man v Machine challenge 6,000,000 moves per second - the most powerful chess computer on earth.

Regards from Spain.

Ajedrecista.

[Milos]

Finally, there is some info in the same site concerning the engine qualification between Deep Fritz and Deep Junior, where both played on identical powerful Pentium III dual processing 933 MHz computers:

https://web.archive.org/web/20020109144 ... p=mvm_deep

To win its position to challenge Kramnik this October, Deep Fritz beat rival software Deep Junior by contesting 25 games on identical powerful Pentium III dual processing 933 MHz computers. IBM was invited to pitch its formidable Deep Blue against Deep Fritz for this contest, but declined to do so.

------------

More references to 6,000,000 positions per second:

https://web.archive.org/web/20021204234 ... com/about/

What does it mean dual processing??? Petium III Xeons are all single core, single thread machines.
Also having 8 CPUs running 6Mnps would mean that each core is running at least 800knps. That seems extremely high for such a crappy 933MHz core. Only possible in late end-game.
Someone was seriously lying when advertising those events.

[Ajedrecista]

Hello Milos:

What does it mean dual processing??? Petium III Xeons are all single core, single thread machines.
Also having 8 CPUs running 6Mnps would mean that each core is running at least 800knps. That seems extremely high for such a crappy 933MHz core. Only possible in late end-game.
Someone was seriously lying when advertising those events.

I know very little about hardware, so I guess it was some piece of marketing. Since we do not have the logs of Deep Fritz of that match, it is only speculation.

Regards from Spain.

Ajedrecista.

[Dr.Wael Deeb]

Finally, there is some info in the same site concerning the engine qualification between Deep Fritz and Deep Junior, where both played on identical powerful Pentium III dual processing 933 MHz computers:

https://web.archive.org/web/20020109144 ... p=mvm_deep

To win its position to challenge Kramnik this October, Deep Fritz beat rival software Deep Junior by contesting 25 games on identical powerful Pentium III dual processing 933 MHz computers. IBM was invited to pitch its formidable Deep Blue against Deep Fritz for this contest, but declined to do so.

------------

More references to 6,000,000 positions per second:

https://web.archive.org/web/20021204234 ... com/about/

What does it mean dual processing??? Petium III Xeons are all single core, single thread machines.
Also having 8 CPUs running 6Mnps would mean that each core is running at least 800knps. That seems extremely high for such a crappy 933MHz core. Only possible in late end-game.
Someone was seriously lying when advertising those events.

True ....

People would take the published data for granted ....

All was done in the sake of advertising ....

Cheers,
Dr.D

[syzygy]

It seems Fritz 9 reached about 480knps on a 1Ghz Pentium III:
http://www.jens-hartmann.at/Fritzmarks/

I have no idea when Fritz 9 was released.

This article mentions 4Mnps:
http://news.bbc.co.uk/2/hi/europe/1466810.stm

[Milos]

It seems Fritz 9 reached about 480knps on a 1Ghz Pentium III:
http://www.jens-hartmann.at/Fritzmarks/

I have no idea when Fritz 9 was released.

This article mentions 4Mnps:
http://news.bbc.co.uk/2/hi/europe/1466810.stm

That Pentium III 1GHz number seems to be inflated since Pentium 4 and 2GHz shows 688knps (and all other Pentium 4 benchmarks agree). So at best Pentium III 1GHz should get around 340knps, and Pentium III Xeon 933MHz around 300knps. Which leads us to assumption that 8 CPUs at best scaling (which was almost certainly not the case in practice) would have around 2.4Mnps. So they lied a lot.