It's time for me to upgrade to a new system. I was hoping to get some
advice from fellow chess enthusiasts.
My main question is whether I should go with one of the new dual-core
processors, or would I be better to stay with the faster (clock speed)
single core processors?
Do the dual-core processor 'behave' like two separate CPUs? I do not
currently own (or plan on purchasing) any of the Chessbase 'Deep' engines,
so would there be any advantage using a dual-core processor system with for
example, Fritz 9? I have noticed that with hyperthreading enabled on my
laptop that the benchmark runs slightly better (987 - 971), but that the
processor usage is only 51% in comparison to 98% with hyperthreading
disabled. Am I running Fritz at 'full performance' even though my CPU is
only running at 51%? I assume that the dual-core processor (or
hyperthreading) would have little/no effect on the playing of other games,
e.g. car-racing.
Before I forget, please direct me to a previous post/posts if these
questions have already been answered.
My other question is whether to purchase an AMD or Intel CPU. I certainly
do not wish to start a 'CPU war of words', but I think I read here
previously that AMD had better integer processing, or something, so it was
better for chess. Would I notice any difference between the two companies? I
must say I have always bought Intel up until now, and if I purchase a
dual-core processor I will almost certainly purchase Intel.
Those are all the questions I can think of at the moment. If anyone has
any advice on specific specs to enhance my chess experience, it would be
much appreciated. )
TIA.

--

RocketMan

On 2006-01-24, RocketMan wrote:

Hi!

It's time for me to upgrade to a new system.

Really?

My other question is whether to purchase an AMD or Intel CPU.

I can tell you that we buy only AMD64 or Opterons for the
institute right now, no intel. This is for the compute
servers as well as the users workstations. It's a matter of
how much you get for your money and we came to the
conclusion that we are better off with AMD. Though I admit,
that this cluster here is not running chess but physics

I certainly do not wish to start a 'CPU war of words',
but I think I read here previously that AMD had better
integer processing, or something, so it was better for
chess. Would I notice any difference between the two
companies?

What I notice is that Intels Pentium M's are actually pretty
nice cpu's and faster than similar desktop processors. If I
compare my notebook to my desktop e.g. they are not that far
apart as the clockspeed would suggest they should be. Anyway
some numbers (can't include my notebook as I do not have it
with me today):

model name : AMD Athlon(TM) XP 3200+
cpu MHz : 2201.129
cache size : 512 KB
bogomips : 4358.14

model name : AMD Athlon(tm) 64 Processor 3500+
cpu MHz : 2200.114
cache size : 512 KB
bogomips : 4358.14

model name : Intel(R) Pentium(R) 4 CPU 2.80GHz
cpu MHz : 2794.279
cache size : 1024 KB
bogomips : 5570.56

model name : Intel(R) Pentium(R) 4 CPU 3.00GHz
stepping : 1
cpu MHz : 2994.279
cache size : 1024 KB
bogomips : 5980.16

Those are all the questions I can think of at the moment.

IMHO you missed a crucial one. Most people miss it. Do you
really need so much computing power?

I mean these boxes consume a considerable ammount of energy
and most of it is dissipated as heat, so you have all the
fans arround and stuff (even if the "green aspect" is no
argument to you . My office e.g. has a noise level which
is really ugly sake of that and coming to the 4 boxes
located here it's unbearable hot especially in summer. So we
try to relocate the computing power into a climatised and
well shut off room.

So, if I would need a new box, well I'd go with a _mobile_
processor. Actually not that much looking on the GHz and
such stuff. This is a reason why I would very closely
examine the new iMacs as they use these processors. (They
are dual core as well.) Chess Software there is probably an
issue one would have to check out, but I think almost
everything from the Unix world will do. But well for the
next years I'll be pretty happy with my ThinkPad running an
Intel Pentium M (with Linux of course

--

Kind regards,
Alexander Wagner

Well, I am working in a computer science lab and we have a very large
numbers of PCs with AMD (single and dual), Intel PIV (single and dual),
Xeon and laptops with pentium M.
It is true that the Pentium M is very fast compared to its clock
frequency, however BogoMips are not a relevant way to test computer
speed for chess.

I would recommand instead to try the crafty "perf" command; I did that a
few months ago and I'll try to dig out the results. Using the LCT2 test
would even be better.

However, one thing to consider is whether you are ready to go for XP64.
There is a great difference in speed on 64 bits processors for all chess
engines using bitmaps moves generators, but XP64 is still a bit buggy
(on my machine, the nvidia WDM drivers are not available, the sound
card software is buggy, and the ethernet card driver exhibits some
unstability, all these bugs disappear when the machine boots under XP32).

Last, a simple advice: you can use Shredder 7.04 (buy 7.0 and download
the upgrade to 7.04 from chessbase) on dual core machines. It uses both
processors, and its results on the LCT2 test are excellent on my P-IV D
3.2Ghz. It's a fairly low cost solution for extremely good results.


Alexander Wagner wrote:
On 2006-01-24, RocketMan wrote:

Hi!


It's time for me to upgrade to a new system.


Really?


My other question is whether to purchase an AMD or Intel CPU.


I can tell you that we buy only AMD64 or Opterons for the
institute right now, no intel. This is for the compute
servers as well as the users workstations. It's a matter of
how much you get for your money and we came to the
conclusion that we are better off with AMD. Though I admit,
that this cluster here is not running chess but physics


I certainly do not wish to start a 'CPU war of words',
but I think I read here previously that AMD had better
integer processing, or something, so it was better for
chess. Would I notice any difference between the two
companies?


What I notice is that Intels Pentium M's are actually pretty
nice cpu's and faster than similar desktop processors. If I
compare my notebook to my desktop e.g. they are not that far
apart as the clockspeed would suggest they should be. Anyway
some numbers (can't include my notebook as I do not have it
with me today):

model name : AMD Athlon(TM) XP 3200+
cpu MHz : 2201.129
cache size : 512 KB
bogomips : 4358.14

model name : AMD Athlon(tm) 64 Processor 3500+
cpu MHz : 2200.114
cache size : 512 KB
bogomips : 4358.14

model name : Intel(R) Pentium(R) 4 CPU 2.80GHz
cpu MHz : 2794.279
cache size : 1024 KB
bogomips : 5570.56

model name : Intel(R) Pentium(R) 4 CPU 3.00GHz
stepping : 1
cpu MHz : 2994.279
cache size : 1024 KB
bogomips : 5980.16


Those are all the questions I can think of at the moment.


IMHO you missed a crucial one. Most people miss it. Do you
really need so much computing power?

I mean these boxes consume a considerable ammount of energy
and most of it is dissipated as heat, so you have all the
fans arround and stuff (even if the "green aspect" is no
argument to you . My office e.g. has a noise level which
is really ugly sake of that and coming to the 4 boxes
located here it's unbearable hot especially in summer. So we
try to relocate the computing power into a climatised and
well shut off room.

So, if I would need a new box, well I'd go with a _mobile_
processor. Actually not that much looking on the GHz and
such stuff. This is a reason why I would very closely
examine the new iMacs as they use these processors. (They
are dual core as well.) Chess Software there is probably an
issue one would have to check out, but I think almost
everything from the Unix world will do. But well for the
next years I'll be pretty happy with my ThinkPad running an
Intel Pentium M (with Linux of course

RocketMan wrote:

My main question is whether I should go with one of the new dual-core
processors, or would I be better to stay with the faster (clock speed)
single core processors?

You need to factor in the price tag, and whether you use dual-capable
programs, or multiple (demanding) programs at the same time.

Do the dual-core processor 'behave' like two separate CPUs?

Yes.

I do not
currently own (or plan on purchasing) any of the Chessbase 'Deep' engines,
so would there be any advantage using a dual-core processor system with for
example, Fritz 9?

There is no advantage (for non-Deep engines), unless you, for example,
want to have it analyzing and do some other demanding task at the same time.

I have noticed that with hyperthreading enabled on my
laptop that the benchmark runs slightly better (987 - 971),

The Fritz benchmark "cheats" in the sense that it will use the second
processor, even though it can't actually use it for playing games or
analyzing.

I assume that the dual-core processor (or
hyperthreading) would have little/no effect on the playing of other games,
e.g. car-racing.

A dual-core CPU is something entirely different than hyperthreading.
With hyperthreading, you only have one core that is pretending to be
two, and there is almost never a speedup. A dual core, well, that's
another matter.

More and more programs will take advantage of a second CPU. I believe
video drivers from one company (dont remember if it was ATI or Nvidia,
or both) are already designed to take advantage of them, so it would
help all games.

My other question is whether to purchase an AMD or Intel CPU. I certainly
do not wish to start a 'CPU war of words', but I think I read here
previously that AMD had better integer processing, or something, so it was
better for chess. Would I notice any difference between the two companies?

You would pay more and get less performance when you buy an Intel. This
may or may not change in the future.

I must say I have always bought Intel up until now, and if I purchase a
dual-core processor I will almost certainly purchase Intel.

Very bad idea right now. Just look at some tests of the CPUs in your
price range.

--
GCP

RocketMan wrote:

Do the dual-core processor 'behave' like two separate CPUs? I do not
currently own (or plan on purchasing) any of the Chessbase 'Deep' engines,
so would there be any advantage using a dual-core processor system with for
example, Fritz 9? I have noticed that with hyperthreading enabled on my
laptop that the benchmark runs slightly better (987 - 971), but that the
processor usage is only 51% in comparison to 98% with hyperthreading
disabled. Am I running Fritz at 'full performance' even though my CPU is
only running at 51%? I assume that the dual-core processor (or
hyperthreading) would have little/no effect on the playing of other games,
e.g. car-racing.

Dual Cores are almost exactly like two separate CPU's. There are two
Physical Processors. For chess, this means that one core can run
chess, the other the OS or other programs. With hyperthreading, you
are taking one processor, and moving multiple tasks throught the same
core at the same time. Chess does not work well for this, as it wants
to greedily hog the processor to itself. And yes you are getting full
performance.

But you should see other advantages, the newer systems work harder at
slower speeds, and are actually "faster" for the chess programs. At the
Last human/computer championships the Fritz team got excellent results
with a Pentium M laptop.

Ultimately, Yes the Core Duo will be and is an excellent chip for
running chess engines. Your system will be snappier when running chess.
Chess engine matches may very well be able to run on both cores.

Johnny T wrote:

Dual Cores are almost exactly like two separate CPU's. There are two
Physical Processors. For chess, this means that one core can run
chess, the other the OS or other programs. With hyperthreading, you
are taking one processor, and moving multiple tasks throught the same
core at the same time. Chess does not work well for this, as it wants
to greedily hog the processor to itself. And yes you are getting full
performance.

But you should see other advantages, the newer systems work harder at
slower speeds, and are actually "faster" for the chess programs.

This seems to me to be a very interesting point. Unfortunately,
I am not a programmer and, hence, I am not sure I understand.
Would you be kind enough to explain?

At the
Last human/computer championships the Fritz team got excellent results
with a Pentium M laptop.

Ultimately, Yes the Core Duo will be and is an excellent chip for
running chess engines. Your system will be snappier when running chess.
Chess engine matches may very well be able to run on both cores.

Has anyone succeeded in doing so yet?

Regards,

Major Cat

Major Cat wrote:

This seems to me to be a very interesting point. Unfortunately,
I am not a programmer and, hence, I am not sure I understand.
Would you be kind enough to explain?

It is why AMD has "Performance Ratings". Intel found they could gain
Higher Clock speeds with longer pipelines. AMD found that they could
get more work done with shorter pipelines at slower clockspeeds. Chess
programs tend to have very tight loops that do things like move
generation and position evaluation that are better on smaller pipeline
machines, like AMD processors, and Pentium M and the newer Core
processors. The Core processors are also not as register "starved" as
the older pentium designs and have relatively high amounts of level 2
cache which is well suited for chess engine work. (Though I believe
that if someone wanted to, the transmeta processor could probably be
made into the fastest chess engine processor, but I am pretty sure that
nobody wants to do the work.)

Many thanks! 8)

Johnny T wrote:
It is why AMD has "Performance Ratings". Intel found they could gain
Higher Clock speeds with longer pipelines. AMD found that they could
get more work done with shorter pipelines at slower clockspeeds. Chess
programs tend to have very tight loops that do things like move
generation and position evaluation that are better on smaller pipeline
machines, like AMD processors, and Pentium M and the newer Core
processors. The Core processors are also not as register "starved" as
the older pentium designs and have relatively high amounts of level 2
cache which is well suited for chess engine work. (Though I believe
that if someone wanted to, the transmeta processor could probably be
made into the fastest chess engine processor, but I am pretty sure that
nobody wants to do the work.)

On 2006-01-25, James wrote:

Hi!

It is true that the Pentium M is very fast compared to its
clock frequency, however BogoMips are not a relevant way
to test computer speed for chess.

Sure. It was the easiest even level to come up with for
comprison. I even wasn't able to include a M machine. Sorry
for that.

I would recommand instead to try the crafty "perf"
command;

This is surely a bit better, you'll admit though that it
takes some time and during the week I'll not get real
results I guess as all machines have some load.

However, one thing to consider is whether you are ready to
go for XP64.

Hm, well beeing a bit conservative I admit I would not have
bought 64bit this year at all at least not for all machines
below 3GB of RAM and as long as I can get 32bit. Even Linux
seems to have some issues and there are some apps not
compiling on 64bit yet. (Though for our setup the difference
between 64bit sarge and 32bit sarge is only minor. Still the
feeling on 64bit is not that ready as on 32. IMHO!!!)

In any case I still stick to my main point wether one really
needs so much power (at least a normal home user should IMHO
really think about it) and wether it is not better to make
the decision not based on GHz but on things like power
consumption (ie. heat production and stuff).

--

Kind regards,
Alexander Wagner

I made some tests with crafty only with different kind of machines, OS
and compilers.

I used the following FEN position (exract from the lct2 test suite):
rq2rbk1/6p1/p2p2Pp/1p1Rn3/4PB2/6Q1/PPP1B3/2K3R1 w - - 0 1
The move to find is Bxh6

The time recorded is the one displayed by crafty in its analysis tree
when showing Bxh6 as the new best move.
Hashsize was set to 24M, hashpawn to 6M
Under Linux, the compiler used by default was gcc 4.0.3
Under windows, I used the executables available on the net, but I
strongly suspect (see timings below) that they were compiled using the
intel compiler.
WIN32 is Windows XP Pro
WIN64 is windows XP 64
LIN32 is linux 32 bits
LIN64 is linux 64 bits.

These are the raw results:

PIV D 3.2Ghz 1M cache
WIN32 1CPU : 70
WIN32 2CPU : 56
WIN64 1CPU : 51
WIN64 2CPU : 31
LIN32 1CPU : 80
LIN32 2CPU : 45
Intel compiler LIN32 1CPU : 70
Intel compiler LIN32 2CPU : 45

Opteron 146 2Ghz 1M cache
LIN64 1CPU : 49

Pentium M 1.6Ghz 1M cache
WIN32 1CPU : 70
LIN32 1CPU : 80
Intel compiler LIN32 1 CPU : 70

Dual Xeon 2.8Ghz 512K cache
LIN32 1CPU : 90
LIN32 1CPU : 52

PIV 3.0Ghz 1M cache
LIN32 1CPU : 90

ATI shuttle PIV 3.2 GHz 1M cache
LIN32 1CPU : 90

Some comments:
- Results with 2 CPUs are quite unstable. I suppose that it depends on
the way crafty handles parallelism: small variations on the load of the
machine might (I guess) change the way the tree is cut. It looks also
like the gcc executable gets "better" when used with multiple CPUs.
Results with 1CPU are stable
- The intel compiler is better than gcc for this application on single
CPU machine. It is the default compiler in Robert Hyatt's makefile
- The Pentium M is extremely fast, as fast as a PIV with twice its
frequency.
- Two PIV on two different machines can give different results (the ATI
shuttle is a terrible machine for a lot of different reasons...)
- The 64 bits version of crafty is much faster than the 32 bits one,
which is quite normal as crafty is a bitboard chess program.

It looks like a pentium M 2.2Ghz could be indeed a very good choice, at
least for crafty with WIN32 (perhaps a bit expensive, but the 2.0Ghz is
affordable). And I completely agree with Alexander Wagner: the M
processor dissipates much less power, and you can then use fans making
much less noise (even very good fans have problems dissipating the 130W
of the PIV-D 3.2).
The only faster alternative seems to be a 64bits system.

Hope it helps

James


Alexander Wagner a écrit :
On 2006-01-25, James wrote:

Hi!

It is true that the Pentium M is very fast compared to its
clock frequency, however BogoMips are not a relevant way
to test computer speed for chess.

Sure. It was the easiest even level to come up with for
comprison. I even wasn't able to include a M machine. Sorry
for that.

I would recommand instead to try the crafty "perf"
command;

This is surely a bit better, you'll admit though that it
takes some time and during the week I'll not get real
results I guess as all machines have some load.

However, one thing to consider is whether you are ready to
go for XP64.

Hm, well beeing a bit conservative I admit I would not have
bought 64bit this year at all at least not for all machines
below 3GB of RAM and as long as I can get 32bit. Even Linux
seems to have some issues and there are some apps not
compiling on 64bit yet. (Though for our setup the difference
between 64bit sarge and 32bit sarge is only minor. Still the
feeling on 64bit is not that ready as on 32. IMHO!!!)

In any case I still stick to my main point wether one really
needs so much power (at least a normal home user should IMHO
really think about it) and wether it is not better to make
the decision not based on GHz but on things like power
consumption (ie. heat production and stuff).