Xeon 3.6 not performing....any ideas

Perle
Perle
Joined: 22 Jan 05
Posts: 47
Credit: 1,118,780,288
RAC: 513,232
Topic 192660

It really seems to not be performing as strong as it should.
This system is on a intel se7520af2 motherboard, 1 gig of ram, 36 gig raptor drive.
It was built to be cruncher.
Motherboard has the latest chipset drivers and bios.
Intel(R) Xeon(TM) CPU 3.60GHz
Number of CPUs 4
Microsoft Windows Server 2003
Measured floating point speed 1592.57 million ops/sec
Measured integer speed 1896.29 million ops/sec

The latest Einstein "pulsar" wu's are taking at least 30 hours.
As a comparision my amd 1800x2 is doing these in 26 hours.
My C2D at 2.66 does these in about 10hrs.
Also...on spinhenge....3.6 xeon is completing wu's in 1hr18min.
The AMD 1800x2 is at 53 minutes.
The C2D is at 38 minutes.

any ideas or suggestions are welcomed.

Michael Karlinsky
Michael Karlinsky
Joined: 22 Jan 05
Posts: 888
Credit: 22,245,334
RAC: 0

Xeon 3.6 not performing....any ideas

Quote:

any ideas or suggestions are welcomed.

This one?

4 CPU's, means 2 real cores with hyperthreading enabled (four WUs running in parallel)? If so, one WU takes longer on your XEON, but it is doing twice as much WUs as the mentioned AMD X2. (which has two cores, no HT, so max. two WUs are running in parallel).

HTH

Michael

Udo
Udo
Joined: 19 May 05
Posts: 203
Credit: 8,945,570
RAC: 0

RE: any ideas or

Message 63035 in response to message 63034

Quote:

any ideas or suggestions are welcomed.

there are also different WUs available...
Some grant 160 credits others grant over 400 credits!

Udo

Perle
Perle
Joined: 22 Jan 05
Posts: 47
Credit: 1,118,780,288
RAC: 513,232

Yes, these are 3.6 HT xeons,

Yes, these are 3.6 HT xeons, so the OS see's 4 cores.
Xeon 3.6mhz 800mhz FSB 2mb L2 cache

My point is that this systems production does not seem what it should be....or rather what it used to be.
I have not changed anything as far as the OS, or drivers.
I have checked all the usual items, background process's, virii, etc.

It doesnt get used for anything....its just a cruncher.

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9,352,143
RAC: 0

One other thing to keep in

One other thing to keep in mind is we are essentially Beta testing the new app, backend components, raw data structure, and IIRC a 'two pass' strategy for narrowing the hunt for the most likely candidates of interest and speeding things up overall which will be used for the next full length run. So just about everything is in flux right at the moment including the rates, which appear to be highly dependant on CPU type and family. Also, if you look at the stderr file for the result your running you'll see they have pretty much full debugging options enabled, which will tend to slow the app down as well.

All of this including the bug's which are causing a variety of app crashes and other malfunctions will get worked out over the next few weeks, as well as optimizing the app to improve overall runtime performance.

Alinator

Perle
Perle
Joined: 22 Jan 05
Posts: 47
Credit: 1,118,780,288
RAC: 513,232

*nods head in

*nods head in agreement*

...one would think tho....that a 3.6 xeon 'should' do better.

as a unfair comparison my xeon 5355 system does the 'pulsars' in aboot 15 hours

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9,352,143
RAC: 0

No doubt that it will

No doubt that it will eventually. Part of it could be with essentially no optimiztions at this time you may be seeing the effect of the differences in the architectures between the different generations of the Intel processors, which are 'masked' when the apps running on it are highly optimized.

If that's the case, it's pretty clear why Intel abandoned NetBurst for Core. ;-)

Alinator

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 482
Credit: 125,223,309
RAC: 165,520

RE: *nods head in

Message 63040 in response to message 63038

Quote:

*nods head in agreement*

...one would think tho....that a 3.6 xeon 'should' do better.

as a unfair comparison my xeon 5355 system does the 'pulsars' in aboot 15 hours


The advantages of 8MB of L2 cache, and 4 true cores.

Pav Lucistnik
Pav Lucistnik
Joined: 7 Mar 06
Posts: 136
Credit: 853,388
RAC: 0

You really can't expect old

You really can't expect old "pentium4" architecture Xeons to outperform modern "Core" chips... Any Athlon64 at half the clock will smoke them.

If this is a new box you just got for big bucks then someone really got you :)

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9,352,143
RAC: 0

RE: RE: *nods head in

Message 63042 in response to message 63040

Quote:
Quote:

*nods head in agreement*

...one would think tho....that a 3.6 xeon 'should' do better.

as a unfair comparison my xeon 5355 system does the 'pulsars' in aboot 15 hours


The advantages of 8MB of L2 cache, and 4 true cores.

...and a pipline that doesn't extend from here to the moon. ;-)

Perle
Perle
Joined: 22 Jan 05
Posts: 47
Credit: 1,118,780,288
RAC: 513,232

nawl. This system was put

nawl.

This system was put together about 8 months ago.
The 3.6 xeon system is in a Chenbro 3u case with a 650watt server power supply.....new ones are $540...mine was $20.oo off craigslist.

The motherboard and procs and 1 gig of ram (4x256 ddr2-400 ecc)
I traded for building a small shed for a customer to park his Harley in....about $250.oo in lumber and 4 hours of my time to build it.
Then it was great reason for a BBQ so it was all the bratwurst and beer I could consume, it was a great party.

The C2Q x5355 clovertown system cost me $1500.oo for the procs, mobo and a gig of ram(FB-dimms).
It is in a Koolance case on liquid cooling, procs are usually around 95 degrees F.
The Koolance 601 case I have had for years....with many different systems in it....I got it when AMD 2500 barton core procs were the fastest cpu, it was $200.oo then.....how ever long ago that was.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.