Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
Hardware

Draw! 243

Posted by michael from the dueling-banjos dept.
An anonymous reader writes "Heise (publisher of the famous german computer magazine c't) started a most unusual CPU benchmark, today. A dual P4 Xeon 2400 and a dual AthlonMP 2000+ have to prove their abilities to ... play chess! The opponents are running two of the best chess AIs (Previews of Deep Fritz 7 and Shredder 6), so there are four different configurations. With each configuration about 55 matches (~24h) are played. As yet AMD/Fritz is leading, but the benchmark has just started. You can follow the duell online [Sorry, site is in german, but the graphics of the java-applet should be multi-lingual]. What's next? Who wouldn't like to see a Linux/Windows mine sweeper death match!"
This discussion has been archived. No new comments can be posted.

Draw! More

Draw!

Comments Filter:

  • Eh? (Score:4, Interesting)

    by SkyLeach ( 188871 ) on Wednesday June 26, 2002 @07:34PM (#3774435) Homepage
    I'm not asking for any other reason than I really want to know:

    What does this prove about performance?

    • Time is limited (Score:4, Informative)

      by nniillss ( 577580 ) on Wednesday June 26, 2002 @07:37PM (#3774468)
      For each game, both opponents have 10 minutes in total plus 2 seconds per move. Everything else being equal (or symmetric, with all combinations of programs being used) it is not too far fetched to assume that the faster machine wins on average
      • The system who is going to have the advantage is the one with the best system optimization. If there are heavy SSE2 optimizations, chances that the Dual Xeon is going to win. If it has heavy pure FPU operations, then the Athlon is going to win. It's that simple. It's no better than any conventional benchmark

        • Re:Time is limited (Score:4, Funny)

          by SirSlud ( 67381 ) on Wednesday June 26, 2002 @07:42PM (#3774523) Homepage
          better shmetter. the coolness factor (which doesnt seem that popular around slashdot in these sombre times) is off the chizz-arts!
        • That's why both machines are using both AI's.

          And if I see 19 of 20 seconds from /. again I'm going to fuckin cry. Seriously slashdot crew, get a fucking clue. Not all posts require a full 20 seconds. Maybe because Taco boy uses a dialing wand and can only type 20 words a day... but for the rest of us... hey this seems like more than 20 seconds.

          Tom

      • Re:Time is limited (Score:3, Insightful)

        by halftrack ( 454203 )
        Machine means all hardware. In a chess game a machine who can't pull as many moves from RAM as it can process would have a disadvantage.
      • Everything else being equal (or symmetric, with all combinations of programs being used) it is not too far fetched to assume that the faster machine wins on average

        Or more accurately, the algorithm with the better implementation.
    • I haven't read the article but one guess would be that if each machine had a specific and equal time limit in which to decide on its next move, the machine with better performance would be able to think more about the next move in the given time period and thus make better choices about which piece to move next.
    • Ability to compute possibilities. Based on both AIs, if one CPU does better than the other, it gives an idea of its ability to run calculations correctly. The ideal situation is that they tie based on the same AI. With different AI's the results should mirror each other, ie AMD with Fritz wins should equal Intel with Fritz wins, etc. That would indicate equal computation capabilities. If one AI dominates, well then that's just the AI.

      • Re:Eh? (Score:5, Insightful)

        by kervel ( 179803 ) on Wednesday June 26, 2002 @07:50PM (#3774627)
        not true. since computer chess is about searching trees, the fastest computer is able to search deeper in the tree and get a more accurate picture of the best move.
        if you want more info about how chess computers
        (and AI in general) work:
        http://www.gamedev.net/reference/programmin g/featu res/chess1/
        http://oellermann.com/cftchess/notes/ shortcuts.htm l
    • Well, with that deep blue computer beating a human, what does that prove?
      faster computing power I guess?

      It's all relitive to italian susage.

    • Re:Eh? (Score:2, Insightful)

      by Anonymous Coward
      I seriously cant understand how you people can sit there and say "how is this a benchmark??"

      Are you kidding me???

      If you cant see how this is a reasonable benchmark at all go back to grade school.

      The simple fact is that the more grunt a computer has, the further a computer can see into the "future" in order to pick the best move. Personally, i think its a brilliant benchmark, and I would also suggest that it can be taken for granted that he'll be running all combinations of chess engine and cpu otherwise, it would all be pointless.
  • How about Core Wars [mcraeclan.com]?
  • Who wouldn't like to see a Linux/Windows mine sweeper death match!

    Wow. It's such a simple question, but your answer immediately determines if you're a lifeless, antisocial dork or not!
  • Is overclocking considered cheating? I hope the AMD doesn't get hot "under the collar"!

  • explanation? (Score:3, Informative)

    by gripdamage ( 529664 ) on Wednesday June 26, 2002 @07:43PM (#3774544)
    Hey everybody complaining about this not being a benchmark... presumably they are going to compare performance such as "AMD Fritz vs. Intel Shredder" against "Intel Fritz vs. AMD Shredder" when the games are all over. Unless the first move for instance contains some random factor the games should be exactly the same, no matter which processor is running the program. So they'll add up how long it takes for each processor to decide to make the same moves. Maybe?
    • Okay so the games are different. Maybe they will seed the random generator on each processor with the same starting value.
    • thats and they are going to compare all four instances

      AMD/Fritz vs Intel/Shredder
      AMD/Fritz vs INtel/Fritz
      AMD/Shredder vs Intel/Shredder
      AMD/Shredder vd Intel/Fritz

      If one CPU hapeens to be better in each of the four instances, you could conlude that is it a better CPU. I do not think time is relevenat, because the processing time will depend on the board staus which will be different in every game. Some moves are hard to compute.
      • Yes I'm totally wrong. I thought they would compare each processors speed at playing the same game. This could be achieved by making sure the random number generator always returns the same values (either manipulating or replacing the random number generator so it isn't random). I'm wrong.

    • Re:explanation? (Score:2, Insightful)

      by tomstdenis ( 446163 )
      Wrong. Get a clue. Each cpu gets 10 mins per GAME and 2 seconds per move.

      The idea is that the faster cpu will beable to try out more moves before actually comitting to a move. More studying presumably means better chess.

      Tom
      • Yes I was wrong, but I think my idea would be a better measure of performance: comparing how long it takes for each computer to calculate the same moves. Their way tests the correctness of the algorithms as well: if the programs work correctly more computations will result in better chess. On the other hand it seems possible that like a person, the programs last guess may not always be it's best guess.

    • Re:explanation? (Score:3, Insightful)

      by tshak ( 173364 )
      There is a reason that IBM's Deep Blue was not a P3 1Ghz - it needed tons of CPU power to compete with Kasparov. In theory, with the same chess engines, the faster computer will win the game each time. That is the point of these tests.

      Also, considering the Athlon is rated at 2000, the P4 should win each time. It would be also be more beneficial to ignore said "ratings", and look at the budget. If they spent $5 building the P4, and $3K building the Athlon, then it's an unfair comparison because they weren't equal budgets. Unless, however practically useless, you want to see who has the fastest CPU regardless of price.

      • Re:explanation? (Score:5, Interesting)

        by DavidRavenMoon ( 515513 ) on Wednesday June 26, 2002 @09:44PM (#3775591) Homepage
        There is a reason that IBM's Deep Blue was not a P3 1Ghz - it needed tons of CPU power to compete with Kasparov.

        I believe the original Deep Blue used PowerPC 604e CPUs. The newer version is a 32-node RS/6000.

        But there is more to Deep Blue than CPU power.

        How Deep Blue Works [ibm.com]

        "Deep Blue is not only the finest chess-playing computer in the world, it is also the fastest. This makes perfect sense, because history has proven that the fastest computers conduct the most extensive searches into possible positions. More searches gives the computer a wider array of moves to choose from and therefore a greater chance of choosing the optimum move."

        "Deep Blue uses 'live' software that can actually generate up to 200,000,000 positions per second when searching for the optimum move. The software begins this process by taking a strategic look at the board. It then computes everything it knows about the current position, integrates the chess information pre-programmed by the development team, and then generates a multitude of new possible arrangements. From these, it then chooses its best possible next move."

        "The software inside of Deep Blue is one all-inclusive program written in C, running under the AIX operating system. Deep Blue utilizes the IBM SP Parallel System called MPI. 'It's a message-passing system,' says Hoane. 'So the search is just all control logic. You're passing control messages back and forth that say, well, what am I doing? Did you finish this? OK, here's your next job. That kind of thing at the SP level.'"

        "The latest iteration of the Deep Blue computer is a 32-node IBM RS/6000 SP high-performance computer, which utilizes the new Power Two Super Chip processors (P2SC). Each node of the SP employs a single microchannel card containing 8 dedicated VLSI chess processors, for a total of 256 processors working in tandem. The net result is a scalable, highly parallel system capable of calculating 60 billion moves within three minutes, which is the time allotted to each player's move in classical chess."

    • Four rounds of play means they are also doing the "mirror match" contests of AMD Fritz vs Intel Fritz and AMD Shredder vs Intel Shredder. Those two will be purer benchmarks, since it eliminates the program differences.

  • Hey (Score:5, Funny)

    by sinserve ( 455889 ) on Wednesday June 26, 2002 @07:44PM (#3774551)
    My MS-DOS 5/286 Commander Keen is challenging Taco's lesbian Sims [slashdot.org] running on a Linux
    ThinkPad, to a side scrolling mud fight.

  • What exactly does speed/power have to do with this?
    • basically, it just shows which computer is better at playing chess, using different chess programs.

      kinda like having me play britney spears at chess: a smart dood versus a dumb bimbo. we may both know the rules, and one might try to infer that i'm smarter, but basically all i'd be doing is showing that i play chess better.

      but you know what they say:
      bewbies are the ultimate equalizer of intellect.

      in which case i'd lose cause i can't concentrate, or lose cause i want her to win. screw having a good score in chess when you can get to first base!
  • all this processing power, and i'm watching two computers too stupid to draw.

    move 125! move 126! move 127!

    ...

    move 32489! move 32490!
    [insert blue screen of death on both computers here]

    • Chess games are finite (Score:4, Informative)

      by yerricde ( 125198 ) on Wednesday June 26, 2002 @08:17PM (#3774902) Homepage Journal

      all this processing power, and i'm watching two computers too stupid to draw.

      Chess has several ko rules that will end the game after no progress. For instance, if twenty-five rounds have passed without a capture or pawn move, or if the same board position has appeared three times, the game is a draw.

      • or if the same board position has appeared three times, the game is a draw.

        Uhm, isn't it in three consecutive moves from both white and black?
        • No. This is a common mistake. The rule states that if an identical position, with the same player to move, has occured three times in the game, then the game is a draw by "repetition of position". It isn't a draw by virtue of repetition of moves, but by repetition of position.
  • Is there an over-riding force to prevent stalemate moves from continuing forever?
  • I'm not sure raw performance can be gauged by 'victories' in chess, seeing as how each situation will present new, and different calculations to be decided.

    It would be much more interesting to see them each perform calculations based on say ten thousand different chess scenarios, and show that side by side in the java applet. This way they are both presented with the same problem solving task.

  • Google Translation [google.com]

    A condition for complete regarding of this side is a Java suited Browser with switched on Java. Applet used from us to the representation of the animated chessboard with numerous Web Clients under different operating systems one tested. Incompatibilities with certain Browserkonfigurationen are not to be excluded nevertheless.

  • It would be more useful (and more entertaining) if each box were given a battery pack and a set of wheels, and AI to duke it out in the ring.

    Somehow, this reminds me of the Rocky movies...ah yes, Intel's P4 clad in patriotic colors as Uncle Sam....
  • this is much better than what's currently on the tv. been watching this 1 game for the past 10 minutes...they have distinct personalities.
    at one point they played a game of 'follow-me-around;' quite comical.
    too bad I cant read German...
  • AMD is up 11-7, with 8 stalemates. AMD has 15 points (57.7%) while Intel has 11 (42.3%). I assume the points are some sort of chess match scoring mechanism. I've only had 2 years of rudementary German, so I can't translate the article or anything close.

  • Update: submission is from yesterday (Score:5, Informative)

    by nniillss ( 577580 ) on Wednesday June 26, 2002 @07:48PM (#3774603)
    The first match ended with a slight advantage for Shredder/Intel versus Fritz/AMD: 20 wins, 22 losses, 29 draws. The present match is for the flipped combinations: now Shredder/AMD leads with 11 wins, 7 losses, and 8 draws against Fritz/Intel. These results indicate a superiority of the AMD architecture (2x Athlon MP 2000+) over the Intel architecture (2x Intel Xeon 2.4 GHz) (see here [heise.de]) which would be in line with the chess SPECINT result.

    Note that both programs learn from game to game within each match, but are reset after the match. In the first match, Shredder started very weak and had a steeper learning curve against Fritz. Since in the second match Shredder/AMD already started strong, a landslide victory for this combination appears likely.

  • Can't wait for Big Blue to rock up and kick both their asses :)
  • My guess on the "how is this a benchmark" question is that they're actually playing speed chess.

    Each side is evaluating moves with a maximum time limit per move. If they make a decision sooner, that's fine. But if the time expires, they just take the best move they've found so far.

    In theory, whichever side can evaluate more moves to evaluate the alpha-beta minimax tree to a deeper level within the time allowed _should_ win.

    So, in effect, they are evaluating which side can look at more nodes/sec, but taking an average over the entire game, and reducing the answer to a one-or-the-other result instead of a numerical comparison.

    Like most benchmarks, all it tells you is how well the benchmark ran. But, I have to admit, this is a pretty novel approach.

  • I hope neither of these machines is also hosting the web server, because it looks like it just got slashdotted.

    :(

    • Still responding fine here for the last hours. Maybe it's just your network connection.

      It's somewhat tough to bring heise.de down. :-) These guys usually know what they're doing.

  • A better idea (Score:5, Funny)

    by cybian ( 27589 ) on Wednesday June 26, 2002 @07:50PM (#3774628)
    I have a better idea. Put up two identical websites; one on AMD, one on Intel. Post the links on Slashdot and see which one stays up the longest.

  • The contest is already running for more than a day, it was not started today:

    According to the contest's page [heise.de] the contest was started on June 25th, 5pm local time.

    More or less just nitpicking. :-) But the programs have already played several dozend matches. (and scored nearly exactly 1:1 the last time I had a look - 29 won by Athlon, 29 won by Xeon, for example.)

    The reason that the counters are at "0" at the moment is that they are being reset at about this time of day everyday.

  • I may not be the best at understanding german, but I can certainly translate this:

    Anmerkung zur Runde 1: Fritz/AMD gewinnt geget Shredder/Intel 20 Partien, remiseirte 29 Partien und verliert 22 Partien

    into

    Noted for round 1: Fritz/AMD won against Shredder/Intel in 20 games, got a draw in 29 games and lost 22 games.

    Translating that into points, AMD has 34½ points (20 wins * 1 point/win + 29 draws * ½ point/draw) and Intel has 36½ points (22 * 1 point/win + 29 draws * ½ point/draw).

    How is AMD winnning?

    And at the moment, that score means nothing, as Shredder could be the better program, but being held back by the Intel CPU, just as Fritz could be the better program being held back by the AMD CPU - we won't know until the bitter end.

    Talk about premature conclusions ... this is even more premature than CNN declaring Gore, then Bush, then Gore, then Bush the winner of Florida ...

    Next time - learn the language of the article before drawing conclusions.
  • A rock/paper/scissors benchmark.

  • Queen's Rook to Queen's Rook 3.999998456

  • Off topic, yes - but can someone point me in the direction for the Java Plugin for Mozilla. I just hate to use IE these days.
  • I found a page that has jpgs of the processor and the motherboard, but I couldn't find any mention of the OS or how much RAM they plugged into these guys?

    anyone else find ?

  • What this proves is simply that we love good old fashioned duels, and chess is one of the oldest and most universal. This is not about which chip is smarter, but about rooting for a sentimental favorite and cheering until a winner is crowned.

    GO AMD!

    • "This is not about which chip is smarter, but about rooting for a sentimental favorite and cheering until a winner is crowned."

      That's the wrong game. You're thinking of checkers.
  • ...that chess is one of the most computationally intense applications known to man. Assuming you want to plot and evaluate full strategies, that is.

    Mind you, I couldn't even beat the chess program on my Sinclair Spectrum with 48Kbytes of RAM, let alone a modern machine ;-)

    -psyconaut
  • According to information in the corresponding Heise forum, doubling the computer power brings about 100-150 ELO points (all combinations should be in the 3000 ELO range). A citation is Ernst A. Heinz: Scalable Search in Computer Chess, Braunschweig/Wiesbaden 2000, p. 126. Since AMD has about 20% advantage in the 186.crafty SPECINT benchmark (see here [specbench.org] and here [heise.de]), one should expect a 20-30 ELO advantage for AMD. Weighted over all combinations of programs, the AMD architecture should score about 53-54% if the cited benchmark was significant.

    In any case, the about 400 games to be played should suffice to find the true chess champion among both dual-processor setups.

  • article:

    ---"Who wouldn't like to see a Linux/Windows mine sweeper death match!"

    Does a crash count as a withdraw?

  • tomshardware (Score:5, Funny)

    by brer_rabbit ( 195413 ) on Wednesday June 26, 2002 @08:20PM (#3774922) Journal
    Too bad this wasn't on Tom's Hardware instead. Regardless of the winner of the chess match, Tom's would determine the winner by highest frame rate in Battle Chess.

    "Despite losing to the AMD, the Intel with the GeForce8 XP 512GB AGPxxx had a frame rate of 1882 FPS. Any chess player would appreciate the 4X anti-aliased graphics of the rook rock-monster pounding the opponent's pawns to pieces."

  • Actual contest is... (Score:3, Informative)

    by Krieger ( 7750 ) on Wednesday June 26, 2002 @08:30PM (#3774993) Homepage
    4 seperate rounds of roughly 55 games day. We've caught it at the beginning of the second day. AMD lost the first day by a couple points 20 wins, 29 ties, and 22 losses. AMD is currently winning the second round as of 8:28PM EST 12-7-9. Everyday they reset the computers (since the programs learn over the course of the day), and switch the programs. Obviously they're going to attempt to fill the product matrix and see which processor performs better with the different programs.

    I'm curious to see how the different platforms perform...
  • How many people ask why this shows which processor is faster.
  • Hey, people quit complaining about how this is pointless or what-not.

    It's just damn cool. I'm interested in the results because, well it's just cool.

    Go ahead mod me down. This is "key", people who think this isn't "key"... well they aren't "key". Scornful moderators aren't key.
  • Here are my benchmarks, if the lameness filter lets them by.

    Lower scores are better.

    CPUGigstone:

    bash$ time for (( x=1 ; x1000000 ; x++ )); do :; done

    NetworkGigstone:

    time for (( x=1 ; x100 ; x++ )); do wget www.slashdot.org; done

    DeveloperGigstone:

    time for (( x=1 ; x10 ; x++ )); do cd /usr/src/linux; make ; done

    HardDiskGigstone:

    time for (( x=1 ; x10 ; x++ )) do; dd if=/dev/urandom of=/tmp/testfile bs=1G count=1; rm /tmp/testfile; done

    All the benckmarks you will ever need.

  • FP? (Score:1)

    by SAN1701 ( 537455 )
    I don't know much about chess algorithms, but aren't they exaustive calculations with integers only?
    I mean, if we are not using (or using too litle) floating point units, the results of the test can be very biased. And, for financial, scientific, gaming, DTP, etc, FP is very used.
    Anyway, this is one of the most fun bench tests I've seen.

  • um (Score:2)

    by Ravagin ( 100668 )

    Who wouldn't like to see a Linux/Windows mine sweeper death match!

    Sorry? They're going to play each other in minesweeper?

  • Linux/Windows Texas Cage Match (Score:5, Funny)

    by rocjoe71 ( 545053 ) on Wednesday June 26, 2002 @08:50PM (#3775149) Homepage
    Who wouldn't like to see a Linux/Windows mine sweeper death match!

    They already tried to pit Windows versus Linux in a chess match:

    1) the Windows machine refused to make its first move-- Microsoft executives explained later that they shouldn't have to make the first move as this could lead to a compromise of it's security system, thereby leaving its horsey vulnerable to worm attacks.

    2) Microsoft later on changed the rules of the chess game citing their freedom to 'innovate' chess, creating new game pieces like 'bazooka' and 'platypus'. Unfortunately, they wouldn't tell anybody else how to use the new pieces or even document that there _were_ new pieces. Once it became apparent that the new pieces were there the Linux camp asked to have them removed but Microsoft refused on the basis that removing the pieces would irretrievably 'break' the game of chess.

  • So... does this mean that Deep Blue has a better benchmark score than Gary Kasparov?

    Fascinating...
  • It seems to me that the end result of this would be that it is a stalemate (Draw).

    I mean, both AMD and Intel processors can crunch the same amount of data, just not in the same amount of time...

    and the other guy's turn kind of depends on how long one person/machine takes...
  • "What's next? Who wouldn't like to see a Linux/Windows mine sweeper death match!"
    Just have one Lindows box, sitting the the corner, playing with itself.

  • Weird opening moves ... (Score:3, Interesting)

    by Hektor_Troy ( 262592 ) on Wednesday June 26, 2002 @11:53PM (#3776425)
    I've been watching this match on and off for the last couple of hours, and it seems to me, that they have some pretty weird opening moves (as in "non classical") ...

    1. d2-d4 Ng8-f6
    2. c2-c4 g7-g6
    3. g2-g3 Bf8-g7
    4. Ng1-f3 0-0

    Not having played a lot of chess on time, I'm not entirely sure, if these are regular moves to upset the board quickly - any players willing to venture some guesses?

  • Oh, come on! That wouldn't even be a challenge. Windows wins hands down... I mean, you stick with what you're good at, right?

  • Personally, I'm looking forward the Intel Shredder vs. AMD Leonardo, Donatello, Raphael and Michaelangelo.

    Will the rooks look like Technodromes? Will Krang (?) play the king? April could definitely be the queen, although it's a bit of a wierd combination with Splinter, who'd have to be the other king..

    Game on!

Slashdot Top Deals

One man's constant is another man's variable. -- A.J. Perlis

Close