Crafty on 767Mhz Alpha at Paris WMCCC?
Richard A. Fowell
As far as I know, I'm starting an unfounded, inflammatory rumour here,
but it's something for Robert Hyatt to consider.
What inspires this is a look at the Dark Thought information
posted at the WMCCC site. Under "Program Descriptions" at
http://www.dcs.qmw.ac.uk/~icca/WMCCC15/descript.htm
it says:
" Thanks to the courtesy of both Digital Equipment Corp. and Kryotech
Inc., DarkThought will run on a superfast 767MHz DEC Alpha-21164a
during the 15th WMCCC in Paris."
The last data I saw, Crafty ran faster on an Alpha than on a Pentium Pro
at the same clock speed. At 767 MHz ... and I'm sure DEC wouldn't mind
buying two chances for some publicity ... heck, they paid for a 2-page
center spread ad for the Alpha in the November 1997 MacWorld (pp. 112-113).
If they are willing to try to sell Alphas to diehard Mac users ...
Are you tempted, Robert?
fow...@netcom.com (Richard A. Fowell)
P.S.: As I recall, the WMCCC is restricted to commercially available
microprocessors - is DEC shipping this 767MHz machine?
(not that a 500MHz alpha would be bad, either)
Dann Corbit
Robert Hyatt wrote in article <60k79b$s...@juniper.cis.uab.edu>...
[snip]
>That is a sticky issue in my book. Years ago anything went. Richard used
>to show up with 680x0 processors running at much higher clock rates than
>anyone else. So it *was* legal. Now, I don't know. But I would venture
>that a 767 alpha would not meet any real "commercially available"
>criteria. But I don't even know that that is required for the WMCCC...
From the rules page: http://www.dcs.qmw.ac.uk/~icca/WMCCC15/rules.htm
>>
"The chess-playing code of a participating program must execute on a single,
generally available microprocessor. The Tournament Director has the right to
demand that a computer be opened for inspection in order to verify that this
is the case."
<<
The clock speed is not specified, but "generally available" might rule out
extra high speed chips. I think you would have to have a clarification to
be sure. I wonder why a single CPU is a requirement. Why not allow a
connection machine? If it is hardware expense that is the sticky point,
they could have a dollar figure that cannot be breached. It might also be
interesting to see which program can achieve the highest rating points per
dollar. Alternatively, they could set up identical machines and have all
the competitors use those, if pure strength of algorithms is the desired
search result.
--
C-FAQ ftp sites: ftp://ftp.eskimo.com ftp://rtfm.mit.edu
Hypertext C-FAQ: http://www.eskimo.com/~scs/C-faq/top.html
C-FAQ Book: ISBN 0-201-84519-9.
Want Software? Algorithms? Pubs? http://www.infoseek.com
brucemo
Robert Hyatt wrote:
> That is a sticky issue in my book. Years ago anything went. Richard used
> to show up with 680x0 processors running at much higher clock rates than
> anyone else. So it *was* legal. Now, I don't know. But I would venture
> that a 767 alpha would not meet any real "commercially available"
> criteria. But I don't even know that that is required for the WMCCC...
I don't know what the legality issues are involving over-clocking, whether
you do it yourself, have the chip manufacturer set it up for you, or have a
third party do it for you.
I would think that an over-clocked processor is still "generally available".
bruce
Ethan Michael O'Connor
Robert Hyatt wrote:
> That is a sticky issue in my book. Years ago anything went. Richard used
> to show up with 680x0 processors running at much higher clock rates than
> anyone else. So it *was* legal. Now, I don't know. But I would venture
> that a 767 alpha would not meet any real "commercially available"
> criteria. But I don't even know that that is required for the WMCCC...
I think that the Kryotech workstation would meet "commercially available"
rules if neccesary. The whole workstation as set up is not yet comm. available,
but it is built around what starts as a normal 600MHz. Alpha. They just
cool it to -40C to get it to overclock that much safely. Enhancing the cooling
system around a widely available chips seems like it would have to fall into
the acceptable category... otherwise you might have to consider ruling out
any chip which required a heatsink to run at its rated speed!
-Ethan O'Connor
zud...@mit.edu
Robert Hyatt
Richard A. Fowell (fow...@netcom.com) wrote:
: As far as I know, I'm starting an unfounded, inflammatory rumour here,
: but it's something for Robert Hyatt to consider.
: What inspires this is a look at the Dark Thought information
: posted at the WMCCC site. Under "Program Descriptions" at
: http://www.dcs.qmw.ac.uk/~icca/WMCCC15/descript.htm
: it says:
: " Thanks to the courtesy of both Digital Equipment Corp. and Kryotech
: Inc., DarkThought will run on a superfast 767MHz DEC Alpha-21164a
: during the 15th WMCCC in Paris."
: The last data I saw, Crafty ran faster on an Alpha than on a Pentium Pro
: at the same clock speed. At 767 MHz ... and I'm sure DEC wouldn't mind
: buying two chances for some publicity ... heck, they paid for a 2-page
: center spread ad for the Alpha in the November 1997 MacWorld (pp. 112-113).
: If they are willing to try to sell Alphas to diehard Mac users ...
: Are you tempted, Robert?
Been tempted. :) Jason Deines is actually going to Paris to operate
Crafty. He has been pursuing DEC to see if he could work out a loaner
deal with them. So far, no real progress I don't think. I'd love to
run on such a machine. But I also am quite sympathetic to running on
the "stock" machine they are providing. IE, a Cray T90 processor might
qualify nowadays... but hardly seems fair.
But we are looking for an alpha. If you have any ideas, feel free to
point us in the right direction. Cray always considered the positive
P/R and press results from Cray Blitz as ample compensation for the machine
time they provided to us. I'd think DEC could get some good milage from
an "alpha crafty" or "crafty alpha" entry...
: fow...@netcom.com (Richard A. Fowell)
: P.S.: As I recall, the WMCCC is restricted to commercially available
: microprocessors - is DEC shipping this 767MHz machine?
: (not that a 500MHz alpha would be bad, either)
That is a sticky issue in my book. Years ago anything went. Richard used
Keith Ian Price
On 28 Sep 1997 03:49:36 GMT, zud...@athena.mit.edu (Ethan Michael
O'Connor) wrote:
>
>Robert Hyatt wrote:
>
>> That is a sticky issue in my book. Years ago anything went. Richard used
>> to show up with 680x0 processors running at much higher clock rates than
>> anyone else. So it *was* legal. Now, I don't know. But I would venture
>> that a 767 alpha would not meet any real "commercially available"
>> criteria. But I don't even know that that is required for the WMCCC...
>
>I think that the Kryotech workstation would meet "commercially available"
>rules if neccesary. The whole workstation as set up is not yet comm. available,
>but it is built around what starts as a normal 600MHz. Alpha. They just
>cool it to -40C to get it to overclock that much safely. Enhancing the cooling
>system around a widely available chips seems like it would have to fall into
>the acceptable category... otherwise you might have to consider ruling out
>any chip which required a heatsink to run at its rated speed!
>
>-Ethan O'Connor
>zud...@mit.edu
What if they cooled it to -40 degrees Fahrenheit? Is the 600Mhz Alpha
available yet? I still only see ads for the 533 and lower.
kp
mclane
kpr...@teleport.com (Keith Ian Price) wrote:
>What if they cooled it to -40 degrees Fahrenheit? Is the 600Mhz Alpha
>available yet? I still only see ads for the 533 and lower.
>kp
I thought the championship is about SOFTWARE and who has the best
software.
It looks some people try to change this again.
If one team would use a Pentium Pro 2000 Mhz, how would the other
teams look ??
No - sorry. I thought this is a championship trying to find out about
the strongest program. It looks it develops again into something
different. Who has the fastest Mercedes for Paris !
Richard A. Fowell
>kpr...@teleport.com (Keith Ian Price) wrote:
>>What if they cooled it to -40 degrees Fahrenheit? Is the 600Mhz Alpha
>>available yet? I still only see ads for the 533 and lower.
>
>>kp
>
>I thought the championship is about SOFTWARE and who has the best
>software.
>
I don't think so ...but I think I explain this best in the last three
paragraphs. Much of the magic of this event is that people think it
is about who has the best software, though.
I'd say it has more to do with advancing computer chess,
by creating interest in computer chess, and computer chess
isn't just software, but algorithms, programmers, compilers and hardware.
You have a point, the hardware issue was one reason
that the annual microcomputer championships were separated from
the rarer "Absolute Championships", one presumes.
My take on this:
The hardware resources for Belle, Cray Blitz, Deep Thought, and the like,
were available to few, and thus discouraged programmer participation.
The WMCCC rule of "a single commercially available microprocessor"
was introduced to give all programmers a shot at the championship.
Note that special purpose hardware is still a valid way to extend
computer chess, hence the continuance of the occasional "Absolute"
championship. Mind you, this de-emphasis of special purpose hardware
may have put a severe crimp in that branch of computer chess.
However, the rule that programs can compete on
"any single commercially available microprocessor"
has been around for a long time - so it's a level field in
that sense. And I believe entries on Sparc, PowerPC, etc.,
have occurred in that past.
So, Alphas seem fair game to me.
If other entries wanted to get DEC to sponsor them on an Alpha,
they could. Kryotech workstations may be a little short, but the
Alpha web pages are advertising the 600MHz Alpha (which Kryotech
apparently "pushes"), so other teams ought to be able to
get a 600MHz Alpha, I would think. And I don't think non-Intel
entries are anything new.
The hardware/compiler element of computer chess is an element,
too - otherwise we'd all still be running on 68020's.
I'd definitely oppose making this a "Uniform Platform" event -
the world is already too "Intel-compatible" for my tastes.
Especially when the Intel chips aren't that great for chess
software in the first place (as opposed to, say, Alphas).
The Alpha did slightly better, even when normalized for clock rate,
than the Pentium Pro on the last Crafty benchmarks I saw.
The "single commercially available microprocessor" rule
accomplishes the original goal - making this an event that any
programmer can win. I don't see any point in further restrictions.
Programmers whose software can be moved to whatever the current
hottest machine (like Dark Thought, Crafty, Ferret, HIARCS ...)
get an advantage from that, but it seems to me that this is a better
thing for computer chess than otherwise.
From the hardware manufacturer's point of view, of course, this
is an opportunity to show that "our microprocessor is best for chess" -
which is one reason that programmers are able to get manufacturers
to loan them machines for these events - which is where I expect
the AMD K6s came from. Why did IBM fund Deep Blue? Why did Intel
fund the PCA? The "a single commercially available microprocessor"
encourages hardware manufacturer participation/funding of computer
chess, and that is good for computer chess.
We want hardware manufacturers had some incentive to provide hardware
support for computer chess, rather than entirely focusing on
multi-media extensions. I think that making the speed of running
"Crafty" one element of the industry-standard benchmark will at least
help the bitboard types get better compiler support, and perhaps
hardware support for FirstOne, LastOne and the like.
I notice on the "Dark Thought" page that:
"as supplied by Digital Equipment Corp. under external technology
research contract DE-0032-97.", so it seems that the Dark Thought
team has managed to tap the hardware manufacturer's interest into
furthering computer chess - seems like a good thing to me.
>It looks some people try to change this again.
>
>If one team would use a Pentium Pro 2000 Mhz, how would the other
>teams look ??
I imagine that the PP2000 vendor would be willing to sponsor more
than one team - after all, that gives them more chances to advertise
that the world champion ran on their machine.
Many of the entries can run just fine on an Alpha:
Dark Thought, Crafty, Ferret, certainly.
HIARCS and AnMon, probably.
"Arthur" usually runs on a PowerPC, as I recall.
>
>No - sorry. I thought this is a championship trying to find out about
>the strongest program. It looks it develops again into something
>different. Who has the fastest Mercedes for Paris !
The question of which is the strongest program is better answered
by the SSDF approach. They use level platforms, and they don't publish
results until they have something halfway statistically significant, and
they are careful about pointing out the statistical limitations of the
results (which many of us, myself included, frequently, gleefully, ignore).
Alas, they don't support hardware other than Pentiums).
This may be part of the answer of why Ed doesn't participate.
He's doing well on the SSDF - why gamble his program's reputation in
a short event where some weaker program can get lucky?
The WMCCC is more about arousing interest in programmers,
manufacturers, and the public. Also a good way for good chess programmers
to meet. Winning the WMCCC is unlikely to be statistically significant,
yes? But we humans love these events and the human contact -
it's all about motivation. There is thrill in competition, particularly
in relatively short events. A fifty-game per round, round robin done
over the Internet takes too long to hold interest, and lacks the human
contact.
fow...@netcom.com (Richard A. Fowell)
brucemo
Dan Thies wrote:
> Well, a Connection Machine is not a microcomputer. If this were the
> "world supercomputer chess championship" or the "world experimental-
> processor-that-only-one-team-can-get-their-hands-on chess
> championship" then I'd see no problem with a Connection Machine or
> a non-production Alpha chip.
I don't know it is production or not. It may be a stock 533 or 600 mhz
chip, over-clocked. If it is not production, that could be a problem.
Last year, teams showed up in Jakarta with over-clocked chips. It never
even occurred to me to challenge this practice. Apparently people have
been over-clocking their machines for years.
I think a challenge to this would be weak, since the processor is
production, but the surrounding stuff has been modified. People used to
build their own computers to take to these things. I can't see how you
could declare that this was illegal, since you didn't use a standard
motherboard or whatever.
> >Alternatively, they could set up identical machines and have all
> >the competitors use those, if pure strength of algorithms is the desired
> >search result.
>
> Okay, whose processor? Motorola, Intel, Cyrix, Digital, AMD?
> MacChess can't be rewritten for a Pentium overnight.
Exactly.
bruce
brucemo
Dann Corbit wrote:
> The clock speed is not specified, but "generally available" might rule out
> extra high speed chips. I think you would have to have a clarification to
> be sure. I wonder why a single CPU is a requirement. Why not allow a
> connection machine? If it is hardware expense that is the sticky point,
> they could have a dollar figure that cannot be breached. It might also be
> interesting to see which program can achieve the highest rating points per
> dollar. Alternatively, they could set up identical machines and have all
> the competitors use those, if pure strength of algorithms is the desired
> search result.
This is going to involve gray areas, since at some point between a washing
machine controller and a Cray you move from micro to mini to mainframe to
supercomputer.
A very imprecise definition is that a microcomputer is something that has one
processor and can sit next to your office trash can without looking out of
proportion.
If someone brings something that I can throw a table cloth over and eat
dinner on, I will argue that it is a mini.
The dollar figure might be an interesting way to go though.
A tournament using identical machines would be nice, but there are major
reasons for not doing it. For one thing, if you standardize on a machine,
you will probably standardize on the Intel architecture, and although I am a
shareholder in all of the obvious companies I cannot argue that this is a
fair thing to do. Why make Intel the only fair computer? What is wrong with
the Power Mac or the Alpha or the K6, other than that they may be better?
Today's extra high speed chip is tomorrow's Intel chip. 787 mhz may sound
like a lot of speed, but in a few years we will be handing these machines
down to our kids, who won't want them, because they are slow. Two and a half
years Frans Morsch brought a 120 mhz machine to Hong Kong, and we were all
impressed. I now use something faster than that to read email, it is too
slow for chess.
There shouldn't be a rule that you can't go faster than the fastest Intel
processor. That's gross.
bruce
brucemo
mclane wrote:
> I thought the championship is about SOFTWARE and who has the best
> software.
No, it has never been completely about software.
The sponsor for the 1995 Paderborn tournament was Peacock. They provided
us all with 120 mhz machines. Except for one team, Genius, which they
provided with a 133 mhz machine, "by prior arrangement."
The speed difference is not that big a deal, but it was a big enough deal
that a special arrangement was made.
And some of the other teams ran on Sparc's, probably by choice, since
these were faster than those Pentiums. Being a Windows app hurt me
seriously there.
At Jakarta, they had 133 mhz Pentiums. I wrote something to the ICCA,
before the tournament started, suggesting that a P6/200 was two point
something times faster than this machine, and that was a big enough
difference that those who brought their own machines would have a
fantastic edge, and could they please try to get faster machines. They
couldn't, and so you saw people bringing in P6's from all over the world.
Including me, damned right.
I simply decided that I would start buying my new "work" machine each
year in August or September.
> No - sorry. I thought this is a championship trying to find out about
> the strongest program. It looks it develops again into something
> different. Who has the fastest Mercedes for Paris !
It was this way years ago, as well, I have been told. There has always
been a sort of arms race.
bruce
Dan Thies
>From the rules page: http://www.dcs.qmw.ac.uk/~icca/WMCCC15/rules.htm
>>>
>"The chess-playing code of a participating program must execute on a single,
>generally available microprocessor. The Tournament Director has the right to
>demand that a computer be opened for inspection in order to verify that this
>is the case."
><<
>
>The clock speed is not specified, but "generally available" might rule out
>extra high speed chips. I think you would have to have a clarification to
>be sure. I wonder why a single CPU is a requirement. Why not allow a
>connection machine?
Well, a Connection Machine is not a microcomputer. If this were the
"world supercomputer chess championship" or the "world experimental-
processor-that-only-one-team-can-get-their-hands-on chess
championship" then I'd see no problem with a Connection Machine or
a non-production Alpha chip.
It seems like the WMCCC's definition of a microcomputer is adequate,
but if the 767-Mhz Alpha is not generally available, and they allow
its use, then their implementation of the definition would seem
inadequate.
>Alternatively, they could set up identical machines and have all
>the competitors use those, if pure strength of algorithms is the desired
>search result.
Okay, whose processor? Motorola, Intel, Cyrix, Digital, AMD?
MacChess can't be rewritten for a Pentium overnight.
Dan (Knowchess)
da...@taic.net
: Why did IBM fund Deep Blue? Why did Intel
: fund the PCA?
Intel has dropped support of the PCA, has it not? Does anyone know
exactly when this happened, and what the current state of the PCA is?
Is the PCA dead, dead and buried, or actually decomposing?
Steve Mayer
da...@taic.net wrote:
: : Why did IBM fund Deep Blue? Why did Intel
: : fund the PCA?
: Intel has dropped support of the PCA, has it not?
: Does anyone know exactly when this happened
At the time (or the _announcement_?!) of the first Kasparov-Deep Blue
match, yes?
: , and what the current state of the PCA is?
: Is the PCA dead, dead and buried, or actually decomposing?
I think it's in the same condition as the inner city schools program
Intel had until the first (or the announcement of?!) the first
Kasparov-DB match, yes?
-Steve
Keith Ian Price
On Sun, 28 Sep 1997 21:16:15 GMT, mcl...@prima.ruhr.de (mclane) wrote:
>kpr...@teleport.com (Keith Ian Price) wrote:
>>What if they cooled it to -40 degrees Fahrenheit? Is the 600Mhz Alpha
>>available yet? I still only see ads for the 533 and lower.
>
>>kp
>
>I thought the championship is about SOFTWARE and who has the best
>software.
>
>It looks some people try to change this again.
>
>If one team would use a Pentium Pro 2000 Mhz, how would the other
>teams look ??
>
>No - sorry. I thought this is a championship trying to find out about
>the strongest program. It looks it develops again into something
>different. Who has the fastest Mercedes for Paris !
>
>
>
I agree. "Generally available" is true for a cooling fan for a
Pentium. A mechanism to cool a processor to -40 degrees is not. I'm
not even sure the 600 Mhz Alpha is generally available yet. Maybe it's
a "Beta Alpha" ;>). I haven't seen any systems advertised for 600 Mhz
yet. But if it were running on a stock Alpha, I think that's fair. It
isn't the WPCCCC.
kp
Komputer Korner
This is the very reason that I for years have been saying that the World
micro computer championships are a fraud and always have been. The world
computer championships every 3 years is an open contest where anything
goes. As soon as you put restrictions on that you run into problems unless
you run it like the Uniform platform championships.
--
- -
Komputer Korner
The inkompetent komputer
If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Robert Hyatt <hy...@crafty.cis.uab.edu> wrote in article
<60k79b$s...@juniper.cis.uab.edu>...
>
Ernst A. Heinz
Dan Thies wrote:
>>From the rules page: http://www.dcs.qmw.ac.uk/~icca/WMCCC15/rules.htm
>>>
>>"The chess-playing code of a participating program must execute on a single,
>>generally available microprocessor. The Tournament Director has the right to
>>demand that a computer be opened for inspection in order to verify that this
>>is the case."
>><<
>>
>>The clock speed is not specified, but "generally available" might rule out
>>extra high speed chips. I think you would have to have a clarification to
>>be sure. I wonder why a single CPU is a requirement. Why not allow a
>>connection machine?
>
> Well, a Connection Machine is not a microcomputer. If this were the
> "world supercomputer chess championship" or the "world experimental-
> processor-that-only-one-team-can-get-their-hands-on chess
> championship" then I'd see no problem with a Connection Machine or
> a non-production Alpha chip.
Right.
> It seems like the WMCCC's definition of a microcomputer is adequate,
^^^^^^^^^^^^^
> but if the 767-Mhz Alpha is not generally available, and they allow
> its use, then their implementation of the definition would seem
> inadequate.
Dan,
you misread the rules which speak of a generally available microprocessor
as opposed to microcomputer. The CPU of the 767MHz DEC/Kryotech machine
is a standard 600MHz DEC EV56 Alpha-21164a microprocessor albeit
overclocked.
But as Bruce has already pointed out, overclocked machines were always
present at ICCA WMCCCs (e.g. Nimzo in Jakarta) and they perfectly adher
to the the current ICCA rules.
=Ernst=
Vincent Diepeveen
>kpr...@teleport.com (Keith Ian Price) wrote:
>>What if they cooled it to -40 degrees Fahrenheit? Is the 600Mhz Alpha
>>available yet? I still only see ads for the 533 and lower.
>
>>kp
>
>I thought the championship is about SOFTWARE and who has the best
>software.
Correct i share that vision.
>It looks some people try to change this again.
>
>If one team would use a Pentium Pro 2000 Mhz, how would the other
>teams look ??
I'm more concerned about good book of Hydra (is it true that this
becomes nimzo 4?) at the standard K6-233
(nice machines by the way, that SDRAM will give Diep another few %),
and Fritz on a PII-300, than i am about the possibility/fact of Crafty/
DarkThought/Guru running on an Alpha.
>No - sorry. I thought this is a championship trying to find out about
>the strongest program. It looks it develops again into something
>different. Who has the fastest Mercedes for Paris !
You can always try to overclock your Pentium Pro to 2000 Mhz, if
you are out of options.
I tried to get an alpha, but it seems that only Digital sometimes
borrows them out.
Vincent.
--
+----------------------------------------------------+
| Vincent Diepeveen email: vdie...@cs.ruu.nl |
| http://www.students.cs.ruu.nl/~vdiepeve/ |
+----------------------------------------------------+
Robert Hyatt
brucemo (bru...@seanet.com) wrote:
: Dan Thies wrote:
: > Well, a Connection Machine is not a microcomputer. If this were the
: > "world supercomputer chess championship" or the "world experimental-
: > processor-that-only-one-team-can-get-their-hands-on chess
: > championship" then I'd see no problem with a Connection Machine or
: > a non-production Alpha chip.
: I don't know it is production or not. It may be a stock 533 or 600 mhz
: chip, over-clocked. If it is not production, that could be a problem.
: Last year, teams showed up in Jakarta with over-clocked chips. It never
: even occurred to me to challenge this practice. Apparently people have
: been over-clocking their machines for years.
: I think a challenge to this would be weak, since the processor is
: production, but the surrounding stuff has been modified. People used to
: build their own computers to take to these things. I can't see how you
: could declare that this was illegal, since you didn't use a standard
: motherboard or whatever.
My major problem with this issue, is that this is *exactly* why I started over
and wrote Crafty. I wanted something that would run and play decent chess on
a commodity processor. I had gotten very tired of the continual negotiations
necessary to schedule a $60 million computer for a chess tournament. That is
why I haven't been very active in trying to line up an alpha for Paris. I
simply don't want to get involved in that *again*. It takes a lot of time,
plus requires a lot of porting/tuning efforts to extract max performance. It
sounds like Cray Blitz all over again.
I'm going to be quite satisfied, even with a 2x performance penalty, to run on
the AMD. At least I know we have a machine to run on. In many years past, CB
was admitted on a "contingency" because we *never* knew for certain that we would
be able to get machine time. It depended on production schedules and test schedules
for new machines. Including, once, where the end-customer agreed to let Cray
delay shipment for a week so we could use their machine at the manufacturing
facility. That was always a nail-biting wait-and-see situation that doubled
the stress-level. No more for me...
Robert Hyatt
Vincent Diepeveen (vdie...@cs.ruu.nl) wrote:
: In <60k79b$s...@juniper.cis.uab.edu> hy...@crafty.cis.uab.edu (Robert Hyatt) writes:
: >Richard A. Fowell (fow...@netcom.com) wrote:
: >: As far as I know, I'm starting an unfounded, inflammatory rumour here,
: >: but it's something for Robert Hyatt to consider.
: >
: >: What inspires this is a look at the Dark Thought information
: >: posted at the WMCCC site. Under "Program Descriptions" at
: >: http://www.dcs.qmw.ac.uk/~icca/WMCCC15/descript.htm
: >
: >: it says:
: >
: >: " Thanks to the courtesy of both Digital Equipment Corp. and Kryotech
: >: Inc., DarkThought will run on a superfast 767MHz DEC Alpha-21164a
: >: during the 15th WMCCC in Paris."
: >
: >: The last data I saw, Crafty ran faster on an Alpha than on a Pentium Pro
: >: at the same clock speed. At 767 MHz ... and I'm sure DEC wouldn't mind
: >: buying two chances for some publicity ... heck, they paid for a 2-page
: >: center spread ad for the Alpha in the November 1997 MacWorld (pp. 112-113).
: >: If they are willing to try to sell Alphas to diehard Mac users ...
: >
: >: Are you tempted, Robert?
: >
: >Been tempted. :) Jason Deines is actually going to Paris to operate
: >Crafty. He has been pursuing DEC to see if he could work out a loaner
: >deal with them. So far, no real progress I don't think. I'd love to
: >run on such a machine. But I also am quite sympathetic to running on
: >the "stock" machine they are providing. IE, a Cray T90 processor might
: >qualify nowadays... but hardly seems fair.
: >
: >But we are looking for an alpha. If you have any ideas, feel free to
: >point us in the right direction. Cray always considered the positive
: >P/R and press results from Cray Blitz as ample compensation for the machine
: >time they provided to us. I'd think DEC could get some good milage from
: >an "alpha crafty" or "crafty alpha" entry...
: What about Diep Alpha?
: Diep can use that level 1 and 2 cache of the alpha!
: Don't need to tell what happens if Diep searches on almost the same
: depth as the opponent. Against Rebel8 searching 2 ply less means 50% score.
: Against Fritz Diep may search 3 ply less sometimes...
: suppose it searches 1 ply less!
: I'm gonna autoplay Diep against Crafty using winboard.
: Do you have a good book for crafty to use?
large works fine, just remember to also download "learn.dat" and after
you build the large book, type "import learn.dat"...
: Just turning off the huge book and the tournament book of Diep
: results lot of times in same line.
: At ICS i saw crafty using a different book against computers than
: it uses against humans, like i mailed you. How to get it?
Crafty doesn't do this. It does modify its book selection algorithm a bit if
it notices it is playing a computer, but it doesn't use two books on ICC. In
fact, "crafty" has been using the same opening book for all games since about
18 months ago or so... with no changes other than the evolution caused by
learning...
Vincent Diepeveen
Just turning off the huge book and the tournament book of Diep
results lot of times in same line.
At ICS i saw crafty using a different book against computers than
it uses against humans, like i mailed you. How to get it?
Vincent
>: fow...@netcom.com (Richard A. Fowell)
>
>: P.S.: As I recall, the WMCCC is restricted to commercially available
>: microprocessors - is DEC shipping this 767MHz machine?
>: (not that a 500MHz alpha would be bad, either)
>
>That is a sticky issue in my book. Years ago anything went. Richard used
>to show up with 680x0 processors running at much higher clock rates than
>anyone else. So it *was* legal. Now, I don't know. But I would venture
>that a 767 alpha would not meet any real "commercially available"
>criteria. But I don't even know that that is required for the WMCCC...
>
--
Robert Hyatt
Keith Ian Price (kpr...@teleport.com) wrote:
: On Sun, 28 Sep 1997 21:16:15 GMT, mcl...@prima.ruhr.de (mclane) wrote:
: >kpr...@teleport.com (Keith Ian Price) wrote:
: >>What if they cooled it to -40 degrees Fahrenheit? Is the 600Mhz Alpha
: >>available yet? I still only see ads for the 533 and lower.
: >
: >>kp
: >
: >I thought the championship is about SOFTWARE and who has the best
: >software.
: >
: >It looks some people try to change this again.
: >
: >If one team would use a Pentium Pro 2000 Mhz, how would the other
: >teams look ??
: >
: >No - sorry. I thought this is a championship trying to find out about
: >the strongest program. It looks it develops again into something
: >different. Who has the fastest Mercedes for Paris !
: >
: >
: >
: I agree. "Generally available" is true for a cooling fan for a
: Pentium. A mechanism to cool a processor to -40 degrees is not. I'm
: not even sure the 600 Mhz Alpha is generally available yet. Maybe it's
: a "Beta Alpha" ;>). I haven't seen any systems advertised for 600 Mhz
Make that an "alpha (beta)"... :)
: yet. But if it were running on a stock Alpha, I think that's fair. It
Ren Wu
On Sat, 27 Sep 1997 19:48:32 -0700, brucemo <bru...@seanet.com>
wrote:
>Robert Hyatt wrote:
>
>> That is a sticky issue in my book. Years ago anything went. Richard used
>> to show up with 680x0 processors running at much higher clock rates than
>> anyone else. So it *was* legal. Now, I don't know. But I would venture
>> that a 767 alpha would not meet any real "commercially available"
>> criteria. But I don't even know that that is required for the WMCCC...
>
>I don't know what the legality issues are involving over-clocking, whether
>you do it yourself, have the chip manufacturer set it up for you, or have a
>third party do it for you.
>I would think that an over-clocked processor is still "generally available".
>bruce
Sorry I wouldn't be able to attent paris event, but for those do go,
and willing to get a few extra points by tring overclocking, please
see Tom's Hardware Guide page at
And you will know everything about overclocking.
Personlly, I don't think that WMCCC can ever become 'fair', Different
platform, or even same platfom with differnt components will endup
with different implement speed.
It would be nice to have one PC program only tourament and eveyone
runs on exactly same hardware.
I like Don' uniform tournment idea very much, and I really hope that
he can resume this event soon.
Ren.
- remove one loop if you reply by email
Rolf Tueschen
mcl...@prima.ruhr.de (mclane) wrote:
>brucemo <bru...@seanet.com> wrote:
>>It was this way years ago, as well, I have been told. There has always
>>been a sort of arms race.
>>bruce
>Brilliant. So why is this event held ? To measure something ? No.
>To help the programmers finding out something ? No.
>To give the programmers a target ?! Maybe.
>It reminds me on beauty-queens-elections.
>The girls have to walk half-naked over a stage, and in the dark some
>old guys are sitting with big eyes --- they claim they want to find
>out who is the most beautiful girl - but in fact they want to make sex
>with one girl from the stage ...
No, not very probable. I would at least choose 4 of them and do it right
on stage. Above all it's clear that you did never f* a girl because it's
not the eyes that become big. Hahaha. Czub?
Rolf Tueschen
mcl...@prima.ruhr.de (mclane) wrote:
>Would thy allow Schumacher to use a car that has a 30000 Ltr. fuel ?
You mean a flyer or a real space rocket?
>Would they allow 2 Karate guys fighting, and the one uses a gun ?
You would surely need one against my Karate... Czub?
mclane
mclane
fow...@netcom.com (Richard A. Fowell) wrote:
>The WMCCC is more about arousing interest in programmers,
>manufacturers, and the public. Also a good way for good chess programmers
>to meet. Winning the WMCCC is unlikely to be statistically significant,
>yes? But we humans love these events and the human contact -
>it's all about motivation. There is thrill in competition, particularly
>in relatively short events. A fifty-game per round, round robin done
>over the Internet takes too long to hold interest, and lacks the human
>contact.
>fow...@netcom.com (Richard A. Fowell)
I just wonder because in other sports the FAIRNESS part is more
stressed than in computerchess.
Would thy allow Schumacher to use a car that has a 30000 Ltr. fuel ?
Would they allow Boris Becker a rack that is 2 meters in size ?
Would they allow 2 Karate guys fighting, and the one uses a gun ?
But you are right with the thrill.
Maybe one reason Chris and I participate although his program makes
only 4000 NPS.
Also I am awaiting massive free beers from Bruce.
One beer alone is not enough...
Ernst A. Heinz
Richard,
Thank you very much for writing a marvelously insightful post about what
you think the WMCCC events are all about. Your post saved me a good deal
of time because I intended to write something similar.
Let me quote some essentials that I wholeheartedly agree with:
> I'd say the WMCCC has more to do with advancing computer chess,
> by creating interest in computer chess, and computer chess
> isn't just software, but algorithms, programmers, compilers and hardware.
>
> [...]
>
> The "single commercially available microprocessor" rule
> accomplishes the original goal - making this an event that any
> programmer can win. I don't see any point in further restrictions.
>
> [...]
>
> We want hardware manufacturers have some incentive to provide hardware
> support for computer chess, rather than entirely focusing on
> multi-media extensions.
>
> [...]
>
> The WMCCC is more about arousing interest in programmers,
> manufacturers, and the public. Also a good way for good chess programmers
> to meet. Winning the WMCCC is unlikely to be statistically significant,
> yes? But we humans love these events and the human contact -
> it's all about motivation. There is thrill in competition, particularly
> in relatively short events.
The publicity aspect is especially interesting for the ICCA and university
teams like us ("PR money makes the world go round").
Moreover, I believe that having the world's fastest workstation at the WMCCC
in Paris will greatly increase the "newsworthiness" of the championship.
Hopefully, it helps the local organizers secure better media coverage
(e.g. TV) which in turn will boost the visibility of the ICCA and computer
chess in general. The more publicity we can get the better -- especially
w.r.t. funding future computer chess events!
=Ernst=
brucemo
Komputer Korner wrote:
>
> This is the very reason that I for years have been saying that the World
> micro computer championships are a fraud and always have been. The world
> computer championships every 3 years is an open contest where anything
> goes. As soon as you put restrictions on that you run into problems unless
> you run it like the Uniform platform championships.
Well, for years you've been using the wrong word then.
The WCCC (the one Fritz won, in Hong Kong) allows for unlimited hardware.
You have supercomputers competing with custom hardware and microcomputers.
Absolutely no complaint is made about this, you can come with whatever
you've got. There is no effort made to handicap things, but you wouldn't
refer to this as a "fraud", would you?
The WMCCC is a limited hardware event. The limits aren't very well defined,
but you can't bring a supercomputer and you can't multi-process. There may
be some limits on the size of your case, as well. Other than that, anything
goes. This event is, likewise, not a fraud.
Do you want to argue that the Alpha I am using is a mini? If so, perhaps
you would also argue that the P6/200 I used last year is a mini, since the
disparity between it and the loan machines (P5/133's) was greater. Is
Fritz's PII/300 also a mini, or are only non-Intel machines eligible for
mini status?
If Don Beal gets his uniform platform tournament going again, I will compete
in it, assuming he provides machines that can run my program (anything that
runs Windows NT or Windows '95, so this shouldn't be a problem). I competed
in the last one, in 1994.
But his tournament is not the world microcomputer championship, which is not
a uniform platform event.
bruce
Dann Corbit
Seems to me that the three contests mentioned answer three different
questions.
WCCC
Using the best available hardware and software, what is the most potent
chess playing engine that can be currently produced? The platforms may
cost millions of dollars. Contests like this give us marvels like "Deep
Blue". Only a billionaire could afford to buy the winning entry to one of
these contests [unless something revolutionary happens in programming].
WMCCC
Using hardware obtainable at reasonable cost and possible to place in your
bedroom, what is the most potent chess playing engine that can be currently
produced? Most of the solutions at a contest like this would at least be
'obtainable' by a real chess nut.
uniform platform tournament
On a completely level playing field, what is the most potent chess playing
engine that can be currently produced? This shows programming skill, but
penalizes those who have designed for a different platform. It might also
be considered as a test of portability. The contest winners are generally
programs that anyone could afford [though some may not be available {
:-()}].
All three are exciting contests in their own right. I enjoy watching
machine against machine or man against machine as much as man against man,
woman against woman, or man against woman.
--
C-FAQ ftp sites: ftp://ftp.eskimo.com ftp://rtfm.mit.edu
Hypertext C-FAQ: http://www.eskimo.com/~scs/C-faq/top.html
C-FAQ Book: ISBN 0-201-84519-9.
Want Software? Algorithms? Pubs? http://www.infoseek.com
brucemo
Dann Corbit wrote:
> WMCCC
> Using hardware obtainable at reasonable cost and possible to place in your
> bedroom, what is the most potent chess playing engine that can be currently
> produced? Most of the solutions at a contest like this would at least be
> 'obtainable' by a real chess nut.
Right, although some micros cost fifty grand.
People have suggested putting a price ceiling on the machines, but there are some
problems with this, since some machines may have been tweaked or home-built, and
some machines may not be general-purpose machines. A "chess machine" would sell for
more than a general-purpose computer, since you are paying for hardware and software
at the same time.
I would rather there was a ceiling on cost, since my credit card doesn't have a huge
limit, but others might enjoy making deals with manufacturers. Perhaps I will try
to do this next year.
I will always have good hardware at these events.
> uniform platform tournament
> On a completely level playing field, what is the most potent chess playing
> engine that can be currently produced? This shows programming skill, but
> penalizes those who have designed for a different platform. It might also
> be considered as a test of portability. The contest winners are generally
> programs that anyone could afford [though some may not be available {
All of the tournaments are lotteries, so it's not like you should talk down to your
peers for a year if you win one.
I would like to win the WMCCC (or the WCCC, ha), since there is a recognized title
that goes along with that, but I think I would more "proud" if my program won a UPC
that was well-attended.
bruce
Komputer Korner
--
- -
Komputer Korner
The inkompetent komputer
If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
brucemo <bru...@seanet.com> wrote in article <34312E...@seanet.com>...
> Komputer Korner wrote:
> >
> > This is the very reason that I for years have been saying that the
World
> > micro computer championships are a fraud and always have been. The
world
> > computer championships every 3 years is an open contest where anything
> > goes. As soon as you put restrictions on that you run into problems
unless
> > you run it like the Uniform platform championships.
>
> Well, for years you've been using the wrong word then.
>
> The WCCC (the one Fritz won, in Hong Kong) allows for unlimited hardware.
> You have supercomputers competing with custom hardware and
microcomputers.
> Absolutely no complaint is made about this, you can come with whatever
> you've got. There is no effort made to handicap things, but you wouldn't
> refer to this as a "fraud", would you?
Of course not, but see below.
>
> The WMCCC is a limited hardware event. The limits aren't very well
defined,
> but you can't bring a supercomputer and you can't multi-process. There
may
> be some limits on the size of your case, as well. Other than that,
anything
> goes. This event is, likewise, not a fraud.
Why is the world micro not a fraud? The rules are confusing and are unfair
if you have a program that doesn't run on the fastest micros, since you
then are at a disadvantage. The ICCA has always recognized the terms
amateur and professional meaning that they recognize the PC micro market.
Alphas that run UNIX are a clear hardware advantage over the commercial PC
market. Can anybody define a supercomputer? Are bit slicers allowed? What
does the size of a case got to do with anything? What is the exact
definition of multiprocessing? No one can provide a fool proof definition
of this since who is going to be able to tell what is on a motherboard and
there may be specialized processors on it which are dedicated to graphics,
endgame databases, and who knows what else?
Richard A. Fowell
In article <01bcce17$6a171ac0$645bb5cf@ALAN> "Komputer Korner" <kor...@netcom.ca> writes:
<snip >
>Why is the world micro not a fraud? The rules are confusing and are unfair
>if you have a program that doesn't run on the fastest micros, since you
>then are at a disadvantage.
>The ICCA has always recognized the terms
>amateur and professional meaning that they recognize the PC micro market.
>Alphas that run UNIX are a clear hardware advantage over the commercial PC
Or even Macintoshes, in the case of HIARCS:
On my "33 problems, 4.5 minutes each" HIARCS 6.0 benchmark:
HIARCS 6.0 Mac on a Mac 9600/350 - 39489 nps
HIARCS 6.0 PC on a AMD K6/200 - 21396 nps
Does that mean Mac entries should be banned, too?
It's hard enough to get people to write chess software
for the Mac as it is - have some pity!
(Also, I suspect you can get 500Mhz Alpha systems running Windows NT
pretty darned cheap - probably cheaper than the Mac 9600/350 <sigh>.
But I have my eye on the 266 MHz G3 Mac upgrade card that started
shipping last Saturday - anyone in LA have one I can benchmark HIARCS on?)
fow...@netcom.com (Richard A. Fowell)
Bill Pemberton
In article <01bcce17$6a171ac0$645bb5cf@ALAN>,
Komputer Korner <kor...@netcom.ca> wrote:
>
>Why is the world micro not a fraud? The rules are confusing and are unfair
>if you have a program that doesn't run on the fastest micros, since you
>then are at a disadvantage. The ICCA has always recognized the terms
>amateur and professional meaning that they recognize the PC micro market.
>Alphas that run UNIX are a clear hardware advantage over the commercial PC
>
But it's still a micro.... It's not the World WinTel Chess
Championship.
Unix is also no test, the Intel based machine can run Unix and the
Alpha can run Windows NT. Again, it's not World
Must-run-on-Microsoft-OS Chess Championship either. The Pentium is a
microprocessor and the Alpha is a microprocessor, I don't see a
problem.
--
Bill
Robert Hyatt
Bill Pemberton (wf...@virginia.edu) wrote:
: In article <01bcce17$6a171ac0$645bb5cf@ALAN>,
: --
: Bill
I don't really see a big problem. You can nearly buy a 533mhz alpha
machine for the same price you'd pay for a good normal PC... certainly
within a factor of 2x at most. But I wouldn't like to see this go to
any extremes. IE, the base processor on a Cray T90 could qualify as
a microprocessor now. Running at 500mhz, super-scalar, vector processing,
with a memory bandwidth at least 1000x that of the best PC's around.
Only problem with this is that it might cost you 5 million bucks to show
up with one. But you'd be so much faster than the 533 mhz alpha that it
wouldn't be funny at all.
I think the point many are trying to make is that this is (should be)
an event for everyone. The WCCC is really not, because when I show up
with my 60 million dollar computer, I have a *huge* advantage right up
front. And when you show up with your PC, you get to serve as
cannon-fodder for the Cray, DB, the Futjitsu, the Hitachi, and so
forth. If the ICCA isn't careful, the WMCCC can also become an event
that gives a significant advantage to the player with the deepest
pockets. It was that way in the commercial ranks years ago, with
different people showing up with processors running at non-standard
clock rates, hand-picked by the manufacturer to run at that clock
rate. It's become less of a problem of late, with most running on
PC's, albiet the fastest PC's they can find. But fortunately, that
type of machine doesn't cost a farm to buy.
Wonder how an alpha cooled to -180 degrees C, and overclocked to
1 gigahertz would work? The alpha would be cheap.. but the cooling
system might not be. :) Clearly there has to be a limit. Otherwise
I'm talking to Cray next year. :)
Komputer Korner
The problem is that you can't set limits that will cover all possibilities
and have it fair. The computer world moves too fast for even the small ICCA
bureaucracy. No one has come up with rules that that would outlaw the
single Cray processor that Bob is talking about. Only uniform platform or
the wide open non-rules of the Open computer WCCCC are fair. Anything else
and the rules committee are always playing catchup to the technology.
- -
Komputer Korner
The inkompetent komputer
If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Robert Hyatt <hy...@crafty.cis.uab.edu> wrote in article
<60tkr0$l...@juniper.cis.uab.edu>...
brucemo
Komputer Korner wrote:
>
> The problem is that you can't set limits that will cover all possibilities
> and have it fair. The computer world moves too fast for even the small ICCA
> bureaucracy. No one has come up with rules that that would outlaw the
> single Cray processor that Bob is talking about. Only uniform platform or
> the wide open non-rules of the Open computer WCCCC are fair. Anything else
> and the rules committee are always playing catchup to the technology.
Oh gad, here you are again with your pronouncements.
It's a limited hardware event. There are tons of examples of competitions
where you could use better hardware, but are prohibited by the rules
(motorcycles, baseball, bicycling, etc), but the competition is not
uniform-hardware.
There are often attempts to run right up to the edge of legality, but it is
still possible to have such an event without labelling it something nasty.
bruce
Martin Zentner
Ernst A. Heinz wrote:
> Moreover, I believe that having the world's fastest workstation at the
> WMCCC in Paris will greatly increase the "newsworthiness" of the
> championship.
>
> Hopefully, it helps the local organizers secure better media coverage
> (e.g. TV) which in turn will boost the visibility of the ICCA and
> computer chess in general. The more publicity we can get the better --
> especially w.r.t. funding future computer chess events!
>
> =Ernst=
Here is my opinion on this:
Everyone who is bringing something, that meets the criteria of being a single
cpu should be allowed to do this, no matter at which clock speed it's running.
The only problem I see is: What's going to happen in the future, when new processor's
(in other words a single cpu) consist of an array of cpu's. It's just ONE chip !
Who cares, how many processes it can run at "load 1". So far noone is concerned,
if a multi-threaded program is using a single cpu, but in the future ?!
Some of us have better "connections" to hardware companies than others. As long
as it would be theoretically possible to get into the same position to get one
of those fast machines, why not take the advantage ? Everyone agrees, that in
Formula1 racing it's not just the driver winning the championship, but the engine
and car design as well. If it says: "Hardware has to be generally available" I read
it that way: If it might be possible to bring this machine (in general) just do it.
To Thorsten Czub, who is overdoing things again: Noone is breaking any rules,
they are just "stretched". I know, there is a strict rule about the amount of fuel
in a formula1 car (actually they specified the density of the stuff, to make sure,
that noone can use "cooled" fuel to reduce volume: this has "legally" been done before
as a "streching" of the rules.) And I "guess" there is a rule, that prohibits guns in
karate contests. :-) [I don't know exactly about the rule for tennis rackets, but I
remember, that it might not be longer than the height of the net.] Be sure, that in
every competition everyone *legally* tries to do the best he/she can. I have no problem
with "unfair" contests as long as they are according to the rules. If you don't like
the rules, noone forces you to compete... Then again: Noone expects to win a grandprix
in a "slow" standard sportscar, but if you can't win when driving the best car, who cares
anyway ?
In fact it's more challenging to compete against faster machines, because you're
somehow in an outsider position. Noone expects you to win, and it is more fun if
you do. In formula1 you have to show, that you're good enough to drive one of the
top cars, but still I would say, they are generally available: Just go and get one,
if not this year, then next year.
One last point:
A commercial program has 2 choices to go for the title:
a) bring it's own fast hardware (at least the speed of other commercial entries)
b) accept the commercial tournament victory on slow hardware, but not the overall title.
(as Virual Chess did in Jakarta)
For amateur entries, it is fun to run on the fastest possible hardware, but if
you win the tournament and try to sell your program afterwards: Try to tell your
future customers, that he/she needs an "overclocked, water-cooled, hand-selected,
whatever ... " workstation to beat something that's on the market for years and
plays strong chess on a "cheap" home-computer.
Hope to see you all at Paris
-Martin (author of XXXXII)
--
Martin Zentner, University of Passau [http://www.uni-passau.de/]
E-Mail: mailto:zen...@phil.uni-passau.de
WWW: http://www.phil.uni-passau.de/linguistik/staff/zentner/
Komputer Korner
I wouldn't label it so harshly if there weren't commercial considerations
that depend on the outcome. The world micro championship has always been
used as hype for the winner's commercial aspirations, even to the point of
putting it on the packaging. People that buy these programs are not
knowledgeable like the readers of the r.g.c.c. When they see the packaging
that says world micro champion they are very impressed and they buy the
program. Of course that program is usually strong but it may have got sales
over another program only because of winning on faster hardware. Uniform
platform competitions are the answer to this problem.
--
- -
Komputer Korner
The inkompetent komputer
If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
brucemo <bru...@seanet.com> wrote in article <343316...@seanet.com>...
brucemo
Graham, Douglass wrote:
> This is a good point. In sports that have faced the problem for a long time,
> what we see is that VERY tight specifications are written for the equipment that
> is used.
Yes. The WMCCC doesn't have tight specifications, your application is more a
proposal, and if they don't think that you qualify they would probably turn you
down or talk it over with you.
If I wanted to do something strange, like bring a super-cooled machine, I would
probably talk it over with them first.
> In 1997, it would be reasonable to specify a single processor, 266 Mhz clock and
> 64 Mb RAM. They're close to the top of the normal PC range, and everyone could
> easily buy (or borrow) one.
A "megaherz" isn't a measure of computing power. With some processors you get
more bang for each cycle. This is like saying, "you may put 50 units of fuel in
your car". This would very much discriminate against the makers of tanks whose
volume is described in liters, as opposed to those whose volume is described in
gallons. A 533 mhz alpha, for instance, is not 533/200 times faster than a 200
mhz Pentium Pro, for my purposes.
Using an Alpha is perfectly fair. The only arguments against it are that:
1) It won't run Windows '95.
2) The processor is not made by Intel.
3) Programs written in x86 assembler will have a hard time porting.
I think these are all bogus arguments.
My machine is the same size and weight as a normal PC (actually, it is smaller
than my Pentium Pro). It has one processor in it. It uses a normal PC bus, as
far as my limited hardware knowledge can understand. It runs Windows NT, so its
screen is visually identical to that of my Pentium Pro. It cost six grand, which
sounds like a lot, but I had a lot of memory and a big SCSI disk put in, as well
as a network card, a sound card, and a nice graphics card. You can probably get a
serviceable >= 500mhz machine for under four grand.
bruce
Chris Carson
Bruce,
Have you run the LCTII test suite on your hardware that you
are using for Paris? If so, could you let me know the MHZ,
Rating, avg. NPS, avg. ply, and Hash size.
THANKS! :)
If anyone has this info for DarkThought, Hiarcs (AMD 233),
or Shreder (AMD 233), Please post. THX! :)
--
Best Regards,
Chris Carson email: chris-...@ti.com
Drew Sarkisian
On 2 Oct 1997 01:38:41 -0700, Graham Douglass wrote:
>In article , brucemo says...
>
>>Oh gad, here you are again with your pronouncements.
>>
>>It's a limited hardware event. There are tons of examples of competitions
>>where you could use better hardware, but are prohibited by the rules
>>(motorcycles, baseball, bicycling, etc), but the competition is not
>>uniform-hardware.
>>
>>There are often attempts to run right up to the edge of legality, but it is
>>still possible to have such an event without labelling it something nasty.
>>
>>bruce
>
>This is a good point. In sports that have faced the problem for a long time,
>what we see is that VERY tight specifications are written for the equipment that
>is used.
>
>In view of this experience from other sports, wouldn't it be a jolly good idea
>to apply the same principles to this sort of event?
>
>In 1997, it would be reasonable to specify a single processor, 266 Mhz clock and
>64 Mb RAM. They're close to the top of the normal PC range, and everyone could
>easily buy (or borrow) one.
Way too naiive. Clock speeds across different architectures do not
scale appropriately...even variations of the same "core" architecture
can have vastly different performance characteristics (watch a PowerPC
750 266Mhz outperform a PowerPC 604e 300Mhz (integer calculations) for
a quick example).
--Drew
brucemo
Graham, Douglass wrote:
> These people have long experience of making a hardware/driverware sport work
> successfully. They should be copied in micro computer chess, say I.
How about a specific implementation this time?
bruce
Ernst A. Heinz
Bob Hyatt wrote:
> I don't really see a big problem. You can nearly buy a 533mhz alpha
> machine for the same price you'd pay for a good normal PC... certainly
> within a factor of 2x at most. But I wouldn't like to see this go to
> any extremes. IE, the base processor on a Cray T90 could qualify as
> a microprocessor now. Running at 500mhz, super-scalar, vector processing,
> with a memory bandwidth at least 1000x that of the best PC's around.
I don't think the CPU of a Cray T90 could qualify as a *microprocessor* because
it consists of several chips just like the IBM POWER2.
> Wonder how an alpha cooled to -180 degrees C, and overclocked to
> 1 gigahertz would work? The alpha would be cheap.. but the cooling
> system might not be. :)
A 1GHz Alpha would be an extremely nice animal ... :-)
As for costs and availability of the Kryotech cooling system, their WWW page
at URL = <http://www.kryotech.com> mentions commercially available Intel PPro
overclocked to at least 266MHz at as little as 800-1000 US$ additional cost.
Hence, I do not think that this technology violates the "general availability"
spirit of the WMCCC events.
> Clearly there has to be a limit.
Yes, I fully agree. This should be discussed at the players' meeting in Paris.
> Otherwise I'm talking to Cray next year. :)
If Cray has a real microprocessor available then, you are welcome to do so.
=Ernst=
brucemo
Ernst A. Heinz wrote:
> As for costs and availability of the Kryotech cooling system, their WWW page
> at URL = <http://www.kryotech.com> mentions commercially available Intel PPro
> overclocked to at least 266MHz at as little as 800-1000 US$ additional cost.
>
> Hence, I do not think that this technology violates the "general availability"
> spirit of the WMCCC events.
I cannot buy one of these Alphas. They are not for sale. So the availability
of these cooling units for a 266 Mhz P6 is moot.
I doubt this matters, since the processor *is* generally available.
bruce
Robert Hyatt
Ernst A. Heinz (hei...@ira.uka.de) wrote:
: Bob Hyatt wrote:
: > I don't really see a big problem. You can nearly buy a 533mhz alpha
: > machine for the same price you'd pay for a good normal PC... certainly
: > within a factor of 2x at most. But I wouldn't like to see this go to
: > any extremes. IE, the base processor on a Cray T90 could qualify as
: > a microprocessor now. Running at 500mhz, super-scalar, vector processing,
: > with a memory bandwidth at least 1000x that of the best PC's around.
: I don't think the CPU of a Cray T90 could qualify as a *microprocessor* because
: it consists of several chips just like the IBM POWER2.
: > Wonder how an alpha cooled to -180 degrees C, and overclocked to
: > 1 gigahertz would work? The alpha would be cheap.. but the cooling
: > system might not be. :)
: A 1GHz Alpha would be an extremely nice animal ... :-)
: As for costs and availability of the Kryotech cooling system, their WWW page
: at URL = <http://www.kryotech.com> mentions commercially available Intel PPro
: overclocked to at least 266MHz at as little as 800-1000 US$ additional cost.
: Hence, I do not think that this technology violates the "general availability"
: spirit of the WMCCC events.
: > Clearly there has to be a limit.
: Yes, I fully agree. This should be discussed at the players' meeting in Paris.
: > Otherwise I'm talking to Cray next year. :)
: If Cray has a real microprocessor available then, you are welcome to do so.
: =Ernst=
I suspect they will. But it doesn't change things, as it will become a battle
of "pocket depth" like the WCCC has become. I don't mind that battle, but it
would be much more fun to ignore the hardware procurement problems and simply
work on the chess engine. Now we are faced (again) with spending valuable
"chess time" chasing hardware leads to get faster machines. I'm not so
interested in what I'd call "clock tricks" (overclocking) because that kind
of performance is generally not a linear improvement since they can't freeze
the bus and memory as well to ramp it up proportionally. But wouldn't it be
ugly to show up with a 3-5 million dollar microcomputer and whip everyone,
not with the best program, but with a 10x hardware advantage? This can end
up at that level.
The flip side would be that if a single machine were chosen, it would cause
grief to the programs that used a different processor. So there's no easy
answer. But an "arms race" is also not the greatest plan in my opinion...
Dan Thies
On 29 Sep 1997 08:46:52 GMT, hei...@ira.uka.de (Ernst A. Heinz) wrote:
>Dan Thies wrote:
>> It seems like the WMCCC's definition of a microcomputer is adequate,
> ^^^^^^^^^^^^^
>> but if the 767-Mhz Alpha is not generally available, and they allow
>> its use, then their implementation of the definition would seem
>> inadequate.
>
>Dan,
>
>you misread the rules which speak of a generally available microprocessor
>as opposed to microcomputer. The CPU of the 767MHz DEC/Kryotech machine
>is a standard 600MHz DEC EV56 Alpha-21164a microprocessor albeit
>overclocked.
I wasn't trying to comment on whether or not the 767MHz Alpha would be
acceptable or not. If the CPS is generally available, I'd say they
can use whatever they want to - including overclocking. As I said
before, I don't know (do any of us know?) how they got to 767MHz with
the Alpha - if it's just overclocked, it seems to fit right in the
definition of "generally available."
>But as Bruce has already pointed out, overclocked machines were always
>present at ICCA WMCCCs (e.g. Nimzo in Jakarta) and they perfectly adher
>to the the current ICCA rules.
Yes. Actually, I don't know why there aren't more programs running on
Alphas, the 500MHz chips are not terrbily expensive.
Dan