Different measured speed...


log in

Advanced search

Message boards : Number crunching : Different measured speed...

Author Message
dwakwiaty
Send message
Joined: 6 Oct 00
Posts: 1
Credit: 7,511
RAC: 0
Poland
Message 1926 - Posted: 25 Jun 2004, 14:39:09 UTC

Hi!

I've installed new S@H on my computer and I'm a little bit confused about measured speed:

When I run setiBoinc on Linux 2.6.6, benchmark gave the following results:
Measured floating point speed 984.01 million ops/sec
Measured integer speed 2251.09 million ops/sec
(http://setiboinc.ssl.berkeley.edu/sah/show_host_detail.php?hostid=6596)

But the same test made by setiBoinc on WinXP gave:
Measured floating point speed 2231.95 million ops/sec
Measured integer speed 3980.85 million ops/sec
(http://setiboinc.ssl.berkeley.edu/sah/show_host_detail.php?hostid=27194)

Why do they differ so much? How can this affect credit granted for each WU?

GoldWolf
Send message
Joined: 24 May 02
Posts: 6
Credit: 68,480
RAC: 87
Belgium
Message 1929 - Posted: 25 Jun 2004, 14:45:58 UTC - in response to Message 1926.

There already is a thread launched on this subject.
see "CPU benchmark difference - Windows and Linux?"

But for the moment there are no real answer to the problem.



Profile Paul D. Buck
Volunteer tester
Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 1990 - Posted: 25 Jun 2004, 17:21:10 UTC - in response to Message 1926.

> Hi!
>
> I've installed new S@H on my computer and I'm a little bit confused about
> measured speed:
>
> When I run setiBoinc on Linux 2.6.6, benchmark gave the following results:
> Measured floating point speed 984.01 million ops/sec
> Measured integer speed 2251.09 million ops/sec
> (http://setiboinc.ssl.berkeley.edu/sah/show_host_detail.php?hostid=6596)
>
> But the same test made by setiBoinc on WinXP gave:
> Measured floating point speed 2231.95 million ops/sec
> Measured integer speed 3980.85 million ops/sec
> (http://setiboinc.ssl.berkeley.edu/sah/show_host_detail.php?hostid=27194)
>
> Why do they differ so much? How can this affect credit granted for each WU?
>

One of the issues that is not discussed much anymore is the actual differences observed when the code is complied using different compilers. Historically, my experience with Microsoft compilers the have never turne dout the fastest code.

The only way to be sure you are geetting identical conversions and there for similar results is to use compilers that can cross-platform compile and make the builds and test them.

It is my understanding that the Visual studio is being used on the windows side but have no idea what is being used on the linux side, but it is either Gnu or possibly an intel compiler (I would guess Gnu cause it is cheaper, which does not imply worse).

<p>


<p>

Message boards : Number crunching : Different measured speed...

Copyright © 2014 University of California