Measured floating point speed & OS

Myu
Myu
Joined: 24 Feb 10
Posts: 15
Credit: 911136
RAC: 0
Topic 195769

Hi all !

I'm wondering why on a Windows platform (Win 7) i got 2.7K MFPS and 1.8K on a GNU/Linux platform (Fedora 14 - Linux 2.6.35.x)

CPU-frequency scaling disabeled on the linux, CPU run @ 2.93 Ghz

Any idea? :)

GNU/Linux <3

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3562358667
RAC: 8

Measured floating point speed & OS

Quote:

Hi all !

I'm wondering why on a Windows platform (Win 7) i got 2.7K MFPS and 1.8K on a GNU/Linux platform (Fedora 14 - Linux 2.6.35.x)

CPU-frequency scaling disabeled on the linux, CPU run @ 2.93 Ghz

Any idea? :)

Because the boinc bogobenchmarks have at best an accidental relationship with reality.

mikey
mikey
Joined: 22 Jan 05
Posts: 12061
Credit: 1834324605
RAC: 21086

RE: Hi all ! I'm wondering

Quote:

Hi all !

I'm wondering why on a Windows platform (Win 7) i got 2.7K MFPS and 1.8K on a GNU/Linux platform (Fedora 14 - Linux 2.6.35.x)

CPU-frequency scaling disabeled on the linux, CPU run @ 2.93 Ghz

Any idea? :)

Because when Windows is up and running it is doing 50 gazillion things in the background and as DanNeely said this all effects the numbers. Over time they will stabilize but comparing a Windows machine to a Linux machine the numbers will never be the same. That is part of why a Linux machine runs faster on the same hardware than a Windows machine.

FrankHagen
FrankHagen
Joined: 13 Feb 08
Posts: 102
Credit: 272200
RAC: 0

RE: Because when Windows is

Quote:
Because when Windows is up and running it is doing 50 gazillion things in the background and as DanNeely said this all effects the numbers. Over time they will stabilize but comparing a Windows machine to a Linux machine the numbers will never be the same. That is part of why a Linux machine runs faster on the same hardware than a Windows machine.

if you compare application performance with the same code built with the same compiler, you'll often see it's less than 1% difference.

PG-LLR's for instance.

the benchmark differences have at least one reason: MS-VC vs. GCC.
and afaik the newer boinc-clients for linux 64 have been compiled with SSE and verctorisation enabled.

and there are some other silly aspects in there..

Comment viewing options

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