Benchmark

[PY 222]

Thanks to graeme for providing the benchmark utility for us to run on.

System Specs: Single Opteron 142, 1603.641GHz, 1GB RAM.
OS: Debian 3.0

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+13
Finding SP Mins
Minima, saddle: 0.132682 <- 0.178873 -> -3.02704e-06

================
BENCHMARK FINISH
================

Run time (s): 45

[PY 222]

System Specs: P4 2.8GHz HT, 2813.541GHz, 1GB RAM. (running 1 client)
OS: Debian 3.0

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+13
Finding SP Mins
Minima, saddle: 0.13268 <- 0.178873 -> -4.36321e-06

================
BENCHMARK FINISH
================

Run time (s): 49

[Bok]

I'll try some tomorrow. I think we need IB to try it on his FX55@2.8Ghz :)

Bok

[PY 222]

System Specs: Athlon MP 2800+, 2133.436GHz, 1GB RAM. (running 1 client)
OS: Debian 3.0

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+13
Finding SP Mins
Minima, saddle: 0.132682 <- 0.178873 -> -3.02704e-06

================
BENCHMARK FINISH
================

Run time (s): 40


Hmm... looks like the MP 2800+ is faster than my Opteron. :crazy:

[Mustard]

This is good as it will allow us to work out the best package for running eon!

[PY 222]

System Specs: P4 2.8GHz, 2813.593GHz, 1GB RAM.
OS: Debian 3.0

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+13
Finding SP Mins
Minima, saddle: 0.13268 <- 0.178873 -> -4.36321e-06

================
BENCHMARK FINISH
================

Run time (s): 50

[Mustard]

AMD64-3200, 512 mem, OS- Slack 9.1 32bit.
===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+13
Finding SP Mins
Minima, saddle: 0.132682 <- 0.178873 -> -3.02704e-06

================
BENCHMARK FINISH
================

Run time (s): 33



So this puppy rocks on running a 32bit OS on eon! :)

[PY 222]

Wow... 33s. Now that is rocking! :cheers:

[jasong]

Has anybody thought about putting in slower-faster RAM to see if that has any effect on benchmarks? Also, if there are any chips where the main difference is cache size, that would be an illuminating test(to see the effect of cache size)

[jasong]

Would it be possible for someone, sometime in the next 3-4 days(preferably someone who has Administrative privileges so they can edit months down the road), to make a new topic that consolidates the benchmark info. At the very least, they could copy and paste all the messages together and delete duplicate information like the information printed during the run.

Maybe I'm wrong, but I think these runs are exactly the same, meaning the relevant information is system specs and how long the run took.

[IronBits]

AMD Athlon 64 FX-55 - 2804 MHz
1 GB Corsair DDR-SDRAM PC3200 2-2-2-5
A8N-SLI DELUXE, 1.XX
nVidia nForce4 rev. A3
Windows XP SP2 (32bit)

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 14

[Mustard]

Now that is rocking right along IB!!!!!

[Thor]

What am I doing wrong?

I downloaded the package, started the benchmark exe, program runs but closes on finish so I can't see the time...

This happens on Win2000 SP4..


Thor

...definitly faster than 30 seconds

[Mustard]

run it from the command line, and re-direct the output to a file, such as:

benchmark_cmd > output.txt

and the file output.txt will show you the results.

[black_civic55]

how do you do that, cause i have the same problem as thor

[Mustard]

benchmark.exe > output.txt

from the commad line

[Thor]

ok , got it ! this is the result!

XP3200 @ 2420mhz 1GB DDR400 @220Mhz FSB
Win2k SP4
===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 19


That Rocks!

Thor

[Nitrousine]

XP2500+ @ 1.83Ghz 256x2 Generic DDR333 Dual Channel
XP Home
===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 24

[graeme]

More benchmark data:

36s: 2.0 GHz Opteron 246 using the x86_64 client
35s: 2.0 GHz Opteron 246 using the x86 client (so much for the 64 bit client!)
26s: 2.5 GHz G5 OS X Mac
107s: 1.5 GHz G4 Powerbook

This CuGB client is written in C++ and built with the gcc compiler,
(msvc on windows). At some point I'll post another benchmark for
a potential in which the core is in fortran. This could make a
significant difference for in the timings for different systems.
For now, though, these current timings are relevant.

[jasong]

CPU-Z version 1.29

Number of CPUs 1
Name AMD Athlon MP
Code name Thoroughbred
Specification AMD Athlon(tm) XP 2200+(might be wrong, might be 2500+)

Package Socket A
Instructions Sets MMX, Extended MMX, 3DNow!, Extended 3DNow!, SSE
Clock Speed 1747.8 MHz
Clock multiplier x10.5
Front Side Bus Frequency 166.5 MHz
Bus Speed 332.9 MHz
Stock frequency 1400 MHz
P-Rating 2200+
L1 Data Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L1 Instruction Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L2 Cache 256 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 1747.8 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 64 bits


Memory Type DDR
Memory Size 512 MBytes
Memory Frequency 166.5 MHz (1:1)
DRAM Interleave 2-way
CAS# 2.5
RAS# to CAS# 3
RAS# Precharge 2
Cycle Time (tRAS) 7


Memory SPD
------------------------------------------------------------------------------

DIMM #2

General
Memory type DDR-SDRAM
Manufacturer (ID) Nanya Technology (7F7F7F0B00000000)
Size 512 MBytes
Max bandwidth PC2700 (166 MHz)
Part number M2U51264DS8HC1G-6K
Manufacturing date Week 30/Year 04

Attributes
Number of banks 2
Data width 64 bits
Correction None
Registered no
Buffered no

Timings table
Frequency (MHz) 133 166
CAS# 2.0 2.5
RAS# to CAS# delay 3 3
RAS# Precharge 3 3
TRAS# 6 7



Run time (s): 30

(stupid question: I have two 512MB RAM modules with slightly different ratings, correct or not?)

[jasong]

observation:apparently bang for buck starts with a good Sempron, although I'm still not sure how RAM plays a part, or if it's important at all.

If someone has a lot of different kinds of computers, it would be very kind of them to experiment to see how RAM affects things.

[jasong]

CPU-Z version 1.29
------------------------------------------------------------------------------

Number of CPUs 1
Name AMD Athlon XP
Code name Palomino
Specification AMD Athlon(tm)
Package Socket A
Instructions Sets MMX, Extended MMX, 3DNow!, Extended 3DNow!, SSE
Clock Speed 1249.1 MHz
Clock multiplier x12.5
Front Side Bus Frequency 99.9 MHz
Bus Speed 199.9 MHz
Stock frequency 1666 MHz
L1 Data Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L1 Instruction Cache 64 KBytes, 2-way set associative, 64 Bytes line size
L2 Cache 256 KBytes, 16-way set associative, 64 Bytes line size
L2 Speed 1249.1 MHz (Full)
L2 Location On Chip
L2 Data Prefetch Logic yes
L2 Bus Width 64 bits

Memory SPD
------------------------------------------------------------------------------

DIMM #1

General
Memory type SDRAM
Size 128 MBytes
Max bandwidth PC125 (125 MHz)
Part number
Serial number FFFFFFFF
Manufacturing date Week 255/Year 255

Attributes
Number of banks 1
Data width 64 bits
Correction None
Registered no
Buffered no

Timings table
Frequency (MHz) 125
CAS# 3.0
RAS# to CAS# delay 3
RAS# Precharge 3
TRAS# 6


DIMM #2

General
Memory type SDRAM
Size 128 MBytes
Max bandwidth PC133 (133 MHz)
Part number
Serial number 300E1200
Manufacturing date Week 32/Year 99

Attributes
Number of banks 1
Data width 64 bits
Correction None
Registered no
Buffered no

Timings table
Frequency (MHz) 100 133
CAS# 2.0 3.0
RAS# to CAS# delay 2 3
RAS# Precharge 2 3
TRAS# 5 6


DIMM #3

General
Memory type SDRAM
Manufacturer (ID) Infineon (C1494E46494E454F)
Size 128 MBytes
Max bandwidth PC133 (133 MHz)
Part number HYS64V16300GU-7.5
Serial number 011C007F
Manufacturing date Week 48/Year 01

Attributes
Number of banks 1
Data width 64 bits
Correction None
Registered no
Buffered no

Timings table
Frequency (MHz) 100 133
CAS# 2.0 3.0
RAS# to CAS# delay 2 3
RAS# Precharge 2 3
TRAS# 5 6

# of seconds: 42

[graeme]

I've put a summary of your timing data at:
http://eon.cm.utexas.edu/faqs.php

It looks like windows is wining out for this client.
It also seems that almost everyone is running AMD.

Not to get off topic, but does anyone else wish that
Apple would partner with AMD instead of Intel?

[PS2pcGAMER]

1.7GHz (735) Dothan, 400Mhz FSB, 1GB PC2700RAM, Windows XP


===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132682 <- 0.178873 -> -3.02416e-006

================
BENCHMARK FINISH
================

Run time (s): 38

[Thor]

Did another run with a freshly booted system.

Specs are again a XP3200@ 2420mhz(220MhzFSBx11) with 1GB DDR400RAM@220Mhz, W2k SP4

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 18


found another second;)

Greetz Thor

will do another run at stock 2200Mhz and post it here


Edit:

Now this is kind of interesting:
same specs as above but at stock 2200Mhz (200Mhz FSBx11) RAM at 200Mhz:

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 19

ran it three times, seems to be correct..

Will fire up an XP 2400 and an Athlon 800 and a P4 2.66Ghz later as references



Next EDIT:
P4 2.66Ghz, 512MB RIMM at 533Mhz, W2k SP4

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132682 <- 0.178873 -> -3.02416e-006

================
BENCHMARK FINISH
================

Run time (s): 36

That is quite slow I think...especially if you see the next:


XP2400 @stock 200mhz(256kb L2 Cache), W2k SP4

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 26

AMD seems to beat Intel big time...

[PCZ]

One thing thats obvious from the results posted so far is that Windows is MUCH faster than nix's running the benchmark.

Is this also the case when running the client ?
If so this is a complete reversal of the Status Quo as i understood it.

My XP 2400's are taking around 40 seconds running linux and low 20's running windows :confused:

My A64 3200 venice @ 2600mhz takes 14 seconds running XP.

BTW
PY222 WTH are you running on EON.

[Mustard]

Yep, looks like the benchmark is much faster on windoze than linux. As for running the client I honestly don't know. I end up with lower "times" in the time column under linux running the "real" client, than when I grind the same stuff under a win system of the same cpu/mb/mem. So who knows? The benchmark also shows the variation between m/b's and memory using the same cpu. So I guess it's good for "tuning" a system.

But reality for me is that I can't afford to go purchase 20 copies of windows, so I run linux. And if it runs slower, then that's the way the ball bounces I guess.

[PCZ]

But reality for me is that I can't afford to go purchase 20 copies of windows, so I run linux. And if it runs slower, then that's the way the ball bounces I guess.

Yep totally agree.

I have my doubts that the benchmark is reflecting the performance of the client .

One thing I have seen is that runing on windows a lot of time is wasted with the ZZZ
nonsence.
My linux nodes connect much more reliably.

[TeeJay]

Originally posted by PCZ
One thing thats obvious from the results posted so far is that Windows is MUCH faster than nix's running the benchmark.

Is this also the case when running the client ?
If so this is a complete reversal of the Status Quo as i understood it.


I also was under the impression that nix was 3 times faster than Win...
I don't get it... same client graeme ?

[graeme]

The benchmark represents the time to complete a typical work unit
for the current server. This is a fairly new C++ based potential, so
the difference between windows and linux is the difference between
gcc and msvc.

In the past, we were using potentials written in Fortran. Here,
linux seemed to win out, mainly because we were using an old MS
fortran compiler. About six months aog we switched to a better
(Lahey) fotran compiler and saw a fairly similar performance to
Linux. It will be nice to see a benchmark for that code as well.

Another factor in the past was a time consuming windows screen
saver. I know these are not popular for people on this forum, but
for the general statisics, about 30% of windows cpu time for a
screen saver launched client was lost to the opengl graphics.

I think that Linux would do better, for the current client if we built
it with a commercial C compiler such as the Intel compiler. This
typically produces faster code than gcc.

[jasong]

Originally posted by graeme
I think that Linux would do better, for the current client if we built
it with a commercial C compiler such as the Intel compiler. This
typically produces faster code than gcc. Am I mistaken in assuming the key word here is "commercial?" As in,"We're on a fixed budget, so the project has to suffer."?

[PCZ]

graeme

So can we conclude then that the client we are currently using is faster in windows than linux ?
The windows compiler is better, so produces tighter code ?

I can remeber a time when the windows client was very slow compared to the nix client.
If memory serves the windows client was running in debug mode accounting for it's lack lustre performance.
Thats been fixed now and combined with the better compiler the windows client is now faster than the linux one.

So what to do about the nix client ?
Could you use gcc 4.0 it is supposed to produce code as fast as ICC.

[graeme]

The slow windows client in the past was due to a slow windows (MS) fortran compiler.
When we moved to a faster (Lahey) fortran compiler, and for the current client which is
entirely in c/c++, the windows client speed has improved dramatically.

By commercial, I mean that it's written by a company who's bottom line is to make
a compiler which produces the fastest code possible. The Intel compiler is free for
academic use, so cost is not a factor here, as compared to gcc.

When I first saw the timing numbers, I was also concerned about gcc version. The linux
client is built on an old Redhat machine with gcc 2.96. I had a lot of difficulty getting a
linux client working on all linux systems, and using old static libraries was the only solution
I could find. But then I compared the client as compared to one built with gcc 3.2, and the
benchmark time is the same. It is possible that gcc 4.x might help, but on OS X I have
done this test and again found no improvement.

I think it's also remarkable that 64 bit code on the Opteron does not tend to improve
speed over 32 bit code. I've seen this for a variety of codes in both fortran and C. I wonder
if this has to do with the fact that 32 instructions are already well optimized and the 64 bit
are not. It would be interesting to know if this is a hardware or software issue.

This benchmark has been very informative. I'm going to build a fortran based benchmark,
right now, to see if that makes a significant difference. It will also give a more well rounded
view of eon performance.

[Mustard]

PCZ -- I've noticed the same on the linux nodes connecting much more reliably than windows also. :)


Graeme -- It's good that you built the benchmark. As we are seeing, it has raised some interesting points, and hopefully will result in improvements to the eon clients over a time span.

It is also very good of you to spend your time in this forum discussing your project and client/server issues. As you have seen over some time here, there are some very talented individuals that become involved in this effort with you. Bottom line is that everyone's concern is the success of your project. Far too many projects out there do not have the communication level between the crunching participants and the project staff that is shown here. That is a big feather in your hat!!! And it is really nice that in spite of some growing pain issues in getting things worked out, that there isn't any finger pointing or name calling going on here. I'm pretty amazed at how far you have come since the UW days! It is a much much better project now.

So in spite of the on and off zzz issues, just wanted to give you a warm and fuzzy for your (and your team at the school too) efforts at being so responsive to the crunching members here! :)

Bruce

[graeme]

I've posted a fortran based benchmark named AlPyr_bench on the eon download page.

This could have significicantly different timing values as compared to the CuGB C++
based potential.

[Mustard]

Yep, it was significant.... took 60s longer on linux for my amd64-3200.

[graeme]

The times from this AlPyr benchmark can not be compared to the CuGB benchmark. They are completely different systems. They can only be compared with other machines using the same benchmark. If you post some numbers, I'll make an additional table.

[PCZ]

The situation reverses and nix is quicker on this benchmark

AMD xp 2400 linux
{took 40 secs on c++ benchmark}

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
pyr.con
Done Init.
Initial Energy: -7724.12
Full Relax Energy: -7726.73
Energy: -7726.73
Dimer: Start
Image Done: U: 0.822366 FC: 586 TS: 141
Dimer: Done
HSize: 111
Good Pref: 5.23604e+13
Finding SP Mins
Minima, saddle: -0.00018356 <- 0.822366 -> 0.808473

================
BENCHMARK FINISH
================

Run time (s): 134



AMD xp 2400 Windows XP
{took 28 secs on c++ benchmark}

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
pyr.con
Done Init.
Initial Energy: -7724.12
Full Relax Energy: -7726.74
Energy: -7726.74
Dimer: Start
Image Done: U: 0.70176 FC: 840 TS: 204
Dimer: Done
HSize: 144
Good Pref: 7.58982e+013
Finding SP Mins
Minima, saddle: -9.89229e-005 <- 0.70176 -> 0.422072

================
BENCHMARK FINISH
================

Run time (s): 201

[PCZ]

AMD 3200+ at 2600mhz windows XP
{took 14 secs on c++ bench}


===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
pyr.con
Done Init.
Initial Energy: -7724.12
Full Relax Energy: -7726.74
Energy: -7726.74
Dimer: Start
Image Done: U: 0.70176 FC: 840 TS: 204
Dimer: Done
HSize: 144
Good Pref: 7.58982e+013
Finding SP Mins
Minima, saddle: -9.89229e-005 <- 0.70176 -> 0.422072

================
BENCHMARK FINISH
================

Run time (s): 75

An XP 2400 took 201 seconds to run this benchmark.
Even allowing for the 30% clock speed advantage the A64 is considerably faster.
Around twice as fast.

[graeme]

I see a problem with the AlPyr benchmark. For some reason, the Linux and Windows benchmarks are doing different calculations. The WinXP/Linux pair that PCZ posted, for example, are different. This must have something to do with the fortran compilers.

I'll need to look into this further, and for now, we'll have to call this test off.

[jasong]

Edit: Info superceded by previous post.

[graeme]

Thanks for the nice comments, Lexx. I do appreciate all the time everyone is putting into this project, and the civil nature of the discussion. You guys are great. We'll try to keep up the communication. I know we need to do better showing the results. Alok should have a visualization page ready by the end of the month to view a movie of the current simulation. I also need to write a little decription of the different simulation projects to explain what we're trying to learn.

[PS2pcGAMER]

I just want to verify this benchmark is NOT multithreaded (i.e. it can only take advantage of one CPU). Correct?

I don't know much about eon and I only have single processor machines...so I can't really be sure.

[rjkc4]

I second that post Bruce. Great wording.

Great job Graeme.

Excellent project!!!

Kevin

[PS2pcGAMER]

I started to put together some stuff in PHP to organize this. Would an eon benchmark database be useful to anyone?

Basically you would input info about your computer and the results and then it would go into a database. Then there would be a listing of all the times and specs of the computer and it would be searchable. Good idea or not?

You can see what I have started here (http://jeff.canyonlake.net/eon/add.php). If people are interested, I'll finish it.

[black_civic55]

i would deff post mine on there if i could still figure out how to keep the damn benchmark window open when it finishes.

p.s. how would i enter my benchmark for a pentium D? dont see it as a choice

[PS2pcGAMER]

It is listed there as "Intel PD Smithfield (Dual Core)". I should probably name it better.

I put together a simple batch file to output the results. I'm assuming you are on windows.

http://jeff.canyonlake.net/eon/RUNME.bat

Put it in the same directory as benchmark.exe and just run it. It will output the results into a file called "output.txt" in the same folder.

[graeme]

I think I've got the AlPyr benchmark working. It should give the following output
on all systems (the floating point numbers should be very close).

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data/init.dat
pyr.con
Done Init.
Initial Energy: -7724.12
Full Relax Energy: -7726.73
Energy: -7726.73
Dimer: Start
Image Done: U: 0.900297 FC: 492 TS: 106
Dimer: Done
HSize: 177
Good Pref: 1.18226e+14
Finding SP Mins
Minima, saddle: 0.525074 <- 0.900297 -> -0.000187188

================
BENCHMARK FINISH
================

Here are a few timings:

110s linux32 Redhat E3 2.0GHz Opteron 246
507s win32 WinXP 0.7GHz PIII
78s macosx OSX 10.3 2.5GHz G5 PowerMac
399s macosx OS X 10.4 1.5GHz G4 PowerBook

Cool idea with the database.

[PS2pcGAMER]

My 1.7ghz Dothan CPU gave 150 seconds for this 2nd benchmark.


===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
pyr.con
Done Init.
Initial Energy: -7724.12
Full Relax Energy: -7726.74
Energy: -7726.74
Dimer: Start
Image Done: U: 0.900361 FC: 492 TS: 106
Dimer: Done
HSize: 177
Good Pref: 1.18229e+014
Finding SP Mins
Minima, saddle: 0.525138 <- 0.900361 -> -0.000124041

================
BENCHMARK FINISH
================

Run time (s): 150




BTW, I have put up the database. The URL might change (but I'll put in redirects). Right now it works, but I have a lot to add. You can view the current inputted benchmark results here (http://jeff.canyonlake.net/eon/output.php) and you can input your own results here (http://jeff.canyonlake.net/eon/add.php).

[black_civic55]

ok thanks got it to work, my bad on the database/ accidently entered 3.0 instead of 3000 for speed

p.s. how do i know if my p4 is a willamette, northwood or prescott??

its a presario 2100 laptop if that helps.... is it like right in front of me and i should enter prescott??

o jesus i think i messed up everything haha. just erase my entry and ill do it again later tonite!!! :machgun:

[PS2pcGAMER]

Early P4's were Wilamette. The newest P4's are Prescott. Northwoods were released in between. [url=http://www.cpuid.com]CPU-Z will tell you. It is ok if people guess on there if they don't know.

[black_civic55]

thanks, just wondering if you plan to have it where we can edit our own entries for mistakes or new info. if thats possible...

[PS2pcGAMER]

Originally posted by black_civic55
thanks, just wondering if you plan to have it where we can edit our own entries for mistakes or new info. if thats possible...

I'd have to create a login script of some sort, which is kinda beyond my ability right now. Maybe sometime down the line.

[Thor]

I did a little test to see how the L2 cache affects the client when doing the CuGB_bench.
Unfortunatly I can't fully recreate a XP2400 with my XP3200. I can't change multipliers so I have to go with 181Mhz FSB for the CPU and RAM.

A "real" XP2400 does 2000Mhz with 133Mhz FSB and in my case with 166Mhz DDR RAM.
Except for that, the only difference shoud be the L2 cache, which is 256kb for the XP2400 and 512kb for the XP 3200.

Here again the result for the xp 2400 posted earlier:

XP2400 @stock 2000mhz(256kb L2 Cache), W2k SP4

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 26


Here is the result for my underclocked XP3200

XP3200@ 1991mhz, w2k SP4 1Gig DDR RAM @ 181 Mhz

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 21

It is 5 seconds faster but I'm not shure if it is due to the higher FSB or the bigger L2 cache.
I would think the latter because between my overclocked XP3200 and stock speed is only a 1second difference...

Maybe someone else can replicate hat a little closer, but I thought it was wrth a shot:D

Greets Thor

[jay.pound]

===============
BENCHMARK START
===============

LOADING INITIALIZATION FILE

data\init.dat
first.con
Done Init.
Initial Energy: 100.444
Full Relax Energy: 100.444
Energy: 100.444
Dimer: Start
Image Done: U: 0.178873 FC: 378 TS: 77
Dimer: Done
HSize: 219
Good Pref: 2.39828e+013
Finding SP Mins
Minima, saddle: 0.132679 <- 0.178873 -> -5.71678e-007

================
BENCHMARK FINISH
================

Run time (s): 17


this is only using 1 processor to run, the machine has 4, 4gb ecc ddr 3200 memory cas 2.5. 1.26TB array, pci express x16 geforce 5900 ultra, running windows 2003 server 64bit enterprise edition. didnt shut down any processes or do any tweaking to get a better score. each cpu has 1mb L2 cache.
I'm currently running 4 clients on this machine along with my search engine. so there is always around 3gb of memory used.