Path: news.mathworks.com!not-for-mail
From: "Yi Cao" <y.cao@cranfield.ac.uk>
Newsgroups: comp.soft-sys.matlab
Subject: Re: 64-bit : Vista vs Linux
Date: Wed, 23 Apr 2008 19:58:02 +0000 (UTC)
Organization: Cranfield University
Lines: 94
Message-ID: <fuo4c9$ltm$1@fred.mathworks.com>
References: <fugcs9$3ni$1@fred.mathworks.com> <muylk34n32f.fsf@G99-Boettcher.llan.ll.mit.edu> <fuo28q$9in$1@fred.mathworks.com>
Reply-To: "Yi Cao" <y.cao@cranfield.ac.uk>
NNTP-Posting-Host: webapp-05-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1208980682 22454 172.30.248.35 (23 Apr 2008 19:58:02 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Wed, 23 Apr 2008 19:58:02 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 69713
Xref: news.mathworks.com comp.soft-sys.matlab:464852



"Suresh " <sureshn@aps.anl.gov> wrote in message 
<fuo28q$9in$1@fred.mathworks.com>...
> Hi Peter,
> 
> I wanted to understand this better. I am thinking of 
rebuilding the vista 
> hardware with Linux and do a direct comparison that way. 
I just want the best 
> performance and matlab runs equally well on Linux too (we 
are not doing any 
> simulink). I will post my test results later for people 
who are interested.
> 
> Thanks.
> 
> Suresh
> 
> 
> Peter Boettcher <boettcher@ll.mit.edu> wrote in message 
> <muylk34n32f.fsf@G99-Boettcher.llan.ll.mit.edu>...
> > 
> > [Please post your responses at the end, after trimming 
from the quoted
> > material]
> > 
> > "Suresh " <sureshn@aps.anl.gov> writes:
> > 
> > > "Yi Cao" <y.cao@cranfield.ac.uk> wrote in message
> > > <fulpra$k3f$1@fred.mathworks.com>...
> > >
> > >> "Suresh " <sureshn@aps.anl.gov> wrote in message 
<fulolp$982
> > >> $1@fred.mathworks.com>...
> > >> > 
> > >> > You are right about the memory being faster and 
that is a direct
> > >> > factor in what I am doing. But still I cannot 
explain a factor of 2
> > >> > in speed other than the fact the Linux OS handles 
resources much
> > >> > better than Windows.
> > >> > 
> > >> > 
> > >> > Vista Hardware:
> > >> > Dual-Core Intel? Xeon? Processor 5160 3.00 GHz 
1333 MHz 
> > >> > FSB 4 MB L2 
> > >> > cache EX615AV 
> > >> > Dual-Core Intel? Xeon? Processor 5160 3.00 GHz 
1333 MHz 
> > >> > FSB 4 MB L2 
> > >> > cache (2nd) EX616AV 
> > >> > HP 8GB (4x2GB) DDR2-667 ECC FBD RK355AV 
> > >> > HP 250GB SATA 3Gb/s NCQ 7200 (1st) EM060AV 
> > >> > 
> > >> > Linux Hardware:
> > >> > 2.33 GHz Intel Core Duo
> > >> > 8 GB DDR2-8 MHz RAM
> > >> > Hard disk same speed as well as capacity.
> > 
> > >> How about MATLAB bench results? Do you still get a 
factor 
> > >> of 2 performance difference?
> > 
> > > Actually the benchmark shows that the 2 machines 
should be
> > > very similar.
> > 
> > That indicates that your code is bottlenecked somewhere 
other than at
> > the CPU.  You'll have to keep searching if you want to 
understand this.
> > Try some disk benchmark software (iozone, fio, etc).  
Even if the drives
> > are the same, there could still be a difference in 
filesystems, location
> > of the data on disk, caching/readahead algorithms at 
the OS level, etc.
> > 
> > -Peter
> 
> 

I cannot recall exactly, but as I am aware that the file 
system of Linux (ext2?) as well as other unix systems is 
much more efficient than the one of windows (NTFS, or 
FAT32). I guess this is the main reason for you to see the 
2x difference since your computation is memory/disk 
operation intensive. To comfirm this, you may perform some 
tests on large file operations as as load, save, fread, 
fwrite, etc on both systems to see the difference.

hth
Yi Cao