Advices on computer hardware

[lechristophe]

Hi all,

I would like to get answers to the following question : what would be
the best choice for a post-aquisition, image processing computer ? It
will run ImageJ and possibly other image processing and visualization
software (depending on the OS). I'm wondering which processor,
archirtecture, OS to choose (let's say I have to keep it under 5000€,
possibly less, so no fancy workstations here):

- what about 64 bits architectures ? Are the OSes ready yet (XP, Vista,
OS X, Linux ?) Can I expect a real difference in speed ? Are the
applications already optimized (I've heard about Volocity 64, I've heard
about the problems for ImageJ) ?

- PC or Mac ? Are the new Mac Pro an option ? What are the options for
high-end PCs ?

- To what features and specs should I particularily pay attention,
keeping in mind the image processing use (RAM, writing speed of hard
disks, others ?)

Best Regards,

--
Christophe Leterrier

Postdoc
INSERM UMR641 Neurobiology of ionic channels
IFR Jean Roche - Mediterranee University
Marseille, France

[Frederick Ross]

On 10/19/06, Christophe Leterrier <[hidden email]> wrote:
> I would like to get answers to the following question : what would be
> the best choice for a post-aquisition, image processing computer ? It
> will run ImageJ and possibly other image processing and visualization
> software (depending on the OS). I'm wondering which processor,
> archirtecture, OS to choose (let's say I have to keep it under 5000€,
> possibly less, so no fancy workstations here):

Shouldn't be hard to keep the price below this.  I do most of my image
processing on an old Powerbook, though that has started to feel kind
of constrained, and I dropped $1500 for a souped up Dell.  Now if
you're doing rotation of 3D surface renderings in real time you need
slightly more, but not hugely so.  Though if you're planning to run
Huygens on massive time lapse videos of 3D stacks, you'll want to drop
in the neighborhood of 5000 for a big ol' quad processor machine.

> - what about 64 bits architectures ? Are the OSes ready yet (XP, Vista,
> OS X, Linux ?) Can I expect a real difference in speed ? Are the
> applications already optimized (I've heard about Volocity 64, I've heard
> about the problems for ImageJ) ?

The OSes have been ready for a long time (well, I don't know about
Windows, but Linux has been running on 64 bit architectures for well
over a decade at this point).  You can't expect a difference in speed
from having 64 bits.  Indeed, it will make very little difference to
your perception of speed.  If you were doing high precision integer
calculations, it would get you out of the arbitrary precision package
more often, and it makes boxing/unboxing issues in pure functional
languages slightly nicer, but for just churning through 16 bit
unsigned integer data, it's completely indifferent.

> - PC or Mac ? Are the new Mac Pro an option ? What are the options for
> high-end PCs ?

The Macs are pretty, but if you intend to write your own code, they're
very hell to program on.  I keep a Linux installation next to OS X on
my Powerbook for this purpose, but I unfortunately use MATLAB quite a
bit at the moment and their Linux version is only for x86
architectures.  You'll get a really pretty computer for the price from
Apple, but you'll get every bit as much computer for half the price
from one of the PC vendors.  Also, OS X doesn't manage memory well, so
you will have to restart from time to time to clear it out if you're
going through large datasets.  Linux is in many ways a better option,
unless you're attached to OsiriX.  2D graphics in Java at this point
is much slower on OS X than on Linux, though I think that's supposed
to be rectified soon.  ImageJ certainly feels faster on Linux.

> - To what features and specs should I particularily pay attention,
> keeping in mind the image processing use (RAM, writing speed of hard
> disks, others ?)

Frontside bus speed, disk speed, and RAM size are the three that
you'll notice.  You also probably want to indulge in a nice video
card.  If you go a route where there's driver support and you're
planning to do 3D, a volume rendering board might be a good option.

--
Frederick Ross
Graduate Fellow, (|Siggia> + |McKinney>)/sqrt(2) Lab
The Rockefeller University
Je ne suis pas Fred Cross!

[Mark J. Chopping]

> The Macs are pretty, but if you intend to write your own code, they're
> very hell to program on.... You'll get a really pretty computer for the
> price from Apple, but you'll get every bit as much computer for half the
> price from one of the PC vendors. Also, OS X doesn't manage memory
> well, so you will have to restart from time to time to clear it out if
> you're going through large datasets.

Anyone care to comment on this assessment of Mac/OSX?  I wonder whether
the machines were comparable -- or is this the older Mac vs. new PC
chestnut (viz. "I do most of my image processing on an old Powerbook...")?
Is OSX' memory management really worse than other Unix implementations?

  Best wishes,

   Mark

  Mark J. Chopping, Ph.D.
  Associate Professor -- Remote Sensing
  Earth & Environmental Studies ML350C
  Montclair State University, Montclair, NJ 07043
  Tel: [phone]  Fax: [phone]
  -----------------------------------------------
  http://csam.montclair.edu/~chopping/jornada/EOS

[Fernando Cassia-2]

On 10/19/06, Christophe Leterrier <[hidden email]> wrote:
>
> Hi all,
>
> I would like to get answers to the following question : what would be
> the best choice for a post-aquisition, image processing computer ?



You don't say if you need a desktop or notebook.

I did a quick search today, and was astonished of how low prices have gotten
lately on desktops... take this one for instance:

*HP Pavilion **a1450n **Desktop PC
AMD Athlon 64 X2 Dual-Core Processor 4200+,
2 GB RAM, 250 GB Hard Drive, LightScribe Double Layer DVD+/-R/RW Drive
*NVidia GeForce 6150LE Graphics with up to 256MB
Price?: $599
http://www.amazon.com/gp/product/B000F3HVIG?ie=UTF8&tag=showitem06-20&linkCode=as2&camp=1789&creative=9325&creativeASIN=B000F3HVIG

Oh my...

Too bad I can't discount the price of WinXX and get a blank formatted HD.
Otherwise I'll be paying the "Microsoft tax" for nothing just to then
install Fedora Core 5 or Ubuntu over it...

FC

[dscho]

Hi,

On Fri, 20 Oct 2006, Mark J. Chopping wrote:

> > The Macs are pretty, but if you intend to write your own code, they're very
> > hell to program on.... You'll get a really pretty computer for the price
> > from Apple, but you'll get every bit as much computer for half the price
> > from one of the PC vendors. Also, OS X doesn't manage memory well, so you
> > will have to restart from time to time to clear it out if you're going
> > through large datasets.
>
> Anyone care to comment on this assessment of Mac/OSX?  I wonder whether the
> machines were comparable -- or is this the older Mac vs. new PC chestnut (viz.
> "I do most of my image processing on an old Powerbook...")? Is OSX' memory
> management really worse than other Unix implementations?

In my experience, no. I did not comment on this originally, because I
imagine the original poster had some gripe with OSX, and I thought this
was obvious.

Having said that, I have Linux running on my iBook and probably will never
go back to OSX.

I think it is mostly a matter of taste (and habit: you can't change the
habits of a lifetime), not of technical aspects.

Ciao,
Dscho

[Kenneth Sloan-2]

On Oct 20, 2006, at 7:51 AM, Frederick Ross wrote:


>
> The Macs are pretty, but if you intend to write your own code, they're
> very hell to program on.

Really?  I just took delivery on a Mac Book Pro and have been happily
transferring legacy code to it.  emacs, gcc, and make, plus this  
newfangled
javac thingie...what more could any programmer ask for?


--
Kenneth Sloan                                          
[hidden email]
Computer and Information Sciences                        [phone]
University of Alabama at Birmingham              FAX [phone]
Birmingham, AL 35294-1170                http://www.cis.uab.edu/sloan/

[ctrueden]

Hi,

On 10/20/06, Mark J. Chopping <[hidden email]> wrote:

> > The Macs are pretty, but if you intend to write your own code, they're
> > very hell to program on.... You'll get a really pretty computer for the
> > price from Apple, but you'll get every bit as much computer for half the
> > price from one of the PC vendors. Also, OS X doesn't manage memory
> > well, so you will have to restart from time to time to clear it out if
> > you're going through large datasets.
>
> Anyone care to comment on this assessment of Mac/OSX?  I wonder whether
> the machines were comparable -- or is this the older Mac vs. new PC
> chestnut (viz. "I do most of my image processing on an old Powerbook...")?
> Is OSX' memory management really worse than other Unix implementations?

Well, since you asked... =)

First of all, I strongly dispute that Macs are "very hell to program
on." Since the release of Mac OS X (with its FreeBSD core), I have
personally witnessed a large number of developers switch to Mac,
especially in the laptop market (the MacBook Pros are awesome). I'm
curious why someone would think programming on a Mac is much more
onerous than Linux (since Mac OS X comes with nearly all the standard
*nix tools), or even Windows (which doesn't come with said tools).

As for memory management, as far as I know it is reasonable on OS X. I
have seen some criticism
(http://www.macintouch.com/readerreports/macosx10_3_8/topic2784.html)
but supposedly it's mostly people failing to understand how "inactive
memory" works (http://www.applelinks.com/articles/2003/03/20030303131234.shtml).

Regarding price, I'm not sure. The Mac Mini is pretty cheap.

As for Macs being slower than PCs, I'm not convinced (even in the G5
era: http://www.barefeats.com/macvpc.html). Some Google searches
should easily uncover more price/performance comparisons. Since Macs
use Intel chips now, it seems largely moot. I know my new dual-core
MacPro is obscenely fast, and ImageJ is blazing. I am using Parallels
with both WinXP and Ubuntu running on it, and ImageJ seems just as
fast or faster in OS X than on Windows or Linux (hard to say how much
of that is virtualization slowdown, but supposedly the hardware
virtualization support results in only a 2% performance hit).
Frederick Ross is right about Java2D being slower on the Mac
currently, but that will supposedly change with the next OS X release
with Java 1.6.

To the original poster: which machine you get depends on your needs.
As a Java developer, I really value having a MacPro with the ability
to test on all three major desktop OSes. But for image visualization
and processing, you surely don't need all three. Take a good look at
which software you need to run. If you have OS-specific visualization
software, you should probably use that OS. If you have software for
multiple OSes, you could consider the Parallels approach, or use
Bootcamp with an OS X box -- Bootcamp has the advantage of hardware
accelerated graphics on Windows or Linux, which could be crucial for
you depending on the software you use. Or if you don't need OS X, you
could just get a PC and dual-boot Windows and Linux -- I would say
whether to get a Mac depends on whether you need/want OS X vs Windows
and/or Linux.

-Curtis

[Gino Canessa]

When the 17" model came out (don't remember when), I purchased a MacBook Pro
with 2 GB RAM.  It is running 10.4.8, and has never run anything less than
10.4.

I typically have open XCode, Mail x 2 (Entourage - which is Outlook for Mac,
and Mail.app), Firefox and Camino, many terminals, OsiriX, Parallels (VM
Ware type software - usually running Win2K since I don't have a dedicated
box for it anymore), and many background procs (e.g., MySQL, Apache, etc).

I reboot my system about once every month or two, just suspending (via
closing the laptop) whenever is convenient (typically 2-5 times per day).

If you are constantly monitoring memory usage, you will see some 'funny'
numbers, but they appear to be just that (I believe the OS just does more
caching than typical, but it seems to free the cache without issue).

That said, I'm using the box for general office, storage-oriented
development/testing, and target testing from the Win2K VM.  My memory usage
profile would be fairly different from image processing/analysis, so YMMV.

As far as coding, it depends on what you are doing.  As far as IDE's, you
have your choice (XCode, Eclipse, etc), or you can use text editors with
build utilities.  OS X has its quirks (Aqua, Frameworks, etc), but I don't
recall anything being any worse than others.  Again, this mostly depends on
what you are doing (type of application, language choice, etc).

For instance:

-If you are using Eclipse and writing Java code (since ImageJ is, I'll
assume many are), there are few differences between _any_ operating system
(Windows/Linux/Unix).

-If you are using VI and the GNU Build Utilities (for command line apps),
Windows would be the "odd man out".

-If you are doing interface work with a windowing toolkit (e.g., wxWidgets,
etc), then again, there should be little difference between the three.

-If you are used to Microsoft, Linux and Unix will both have a higher hurdle
to jump.

And so on...


As far as price, my MBP was comparable in price to others (less than Dell's
standard price when I bought it, but Dell had coupons, so it would've been
cheaper).  The trick, again, is what you are buying and through which
department (education, healthcare, government, personal, small business,
etc).



So, for the original poster:

OS should be decided based on which applications you are using, and which
systems you are familiar with (even if a system is slightly less functional,
if you are more comfortable with it, you will get more use out of it).

To the 64-bit question specifically, I'll say that it depends on how large
your datasets are, how they are broken up, and what you are doing with them.
64 bit systems have more addressable memory*, so if you require more than is
typically available, that's a good route to go.

Keep in mind which apps you are using for processing, and look into any
limitations caused by them (for instance ImageJ is limited by the JRE on the
OS above and beyond any limits from the code itself).


*In order to use more than the standard 32-bit addressable memory, both the
OS and the application need to be able to handle it.  For instance, in OS X
the largest contiguous block of memory you can have is 2 GB, so if your
application is hard-coded to use a single block of memory, you're stuck
regardless of what you are running on.  Operating systems in general have
various limits for memory usage (e.g., how much a single
process/thread/block/etc can have), but that's a bit long-winded already, so
Google can fill in the blanks.


Hope this helps,

    Gino




On 10/20/06 8:32 AM, "Mark J. Chopping" <[hidden email]>
wrote:

>> The Macs are pretty, but if you intend to write your own code, they're
>> very hell to program on.... You'll get a really pretty computer for the
>> price from Apple, but you'll get every bit as much computer for half the
>> price from one of the PC vendors. Also, OS X doesn't manage memory
>> well, so you will have to restart from time to time to clear it out if
>> you're going through large datasets.
>
> Anyone care to comment on this assessment of Mac/OSX?  I wonder whether
> the machines were comparable -- or is this the older Mac vs. new PC
> chestnut (viz. "I do most of my image processing on an old Powerbook...")?
> Is OSX' memory management really worse than other Unix implementations?
>
>   Best wishes,
>
>    Mark
>
>   Mark J. Chopping, Ph.D.
>   Associate Professor -- Remote Sensing
>   Earth & Environmental Studies ML350C
>   Montclair State University, Montclair, NJ 07043
>   Tel: [phone]  Fax: [phone]
>   -----------------------------------------------
>   http://csam.montclair.edu/~chopping/jornada/EOS

--
Gino Canessa
 
Software Engineering Corporation (SENCOR)
7900 International Drive - Suite 305
Bloomington, MN 55425
 
web:      www.sencorsoft.com or www.sencor.com
 
email:    [hidden email]
voice:    952.854.7175 ext. 204
fax:      [phone]
support:  [hidden email]
 

CONFIDENTIALITY NOTICE
 
This message and any included attachments are from SENCOR and
are intended only for the addressee. The information
contained in this message is confidential and may constitute
inside or non-public information under international,
federal, or state securities laws.  Unauthorized forwarding,
printing, copying, distribution, or use of such information
is strictly prohibited and may be unlawful.  If you are not
the intended recipient, please promptly delete this message
and notify the sender of the delivery error by e-mail or
calling SENCOR's corporate offices in Bloomington,
Minnesota, U.S.A. at 952.854.7175.  Thank you.
 

[Jeff Brandenburg]

On Oct 20, 2006, at 8:51 AM, Frederick Ross wrote:

> The Macs are pretty, but if you intend to write your own code, they're
> very hell to program on.  I keep a Linux installation next to OS X on
> my Powerbook for this purpose, but I unfortunately use MATLAB quite a
> bit at the moment and their Linux version is only for x86
> architectures.  You'll get a really pretty computer for the price from
> Apple, but you'll get every bit as much computer for half the price
> from one of the PC vendors.  Also, OS X doesn't manage memory well, so
> you will have to restart from time to time to clear it out if you're
> going through large datasets.  Linux is in many ways a better option,
> unless you're attached to OsiriX.  2D graphics in Java at this point
> is much slower on OS X than on Linux, though I think that's supposed
> to be rectified soon.  ImageJ certainly feels faster on Linux.

To amplify previous replies, Java on a G5 or Core Duo Mac is very much
faster than Java on a G4.  I've just replaced my "old Powerbook" (a
667MHz G4) with a Core Duo 1.83GHz MacBook, and I'm seeing up to a
tenfold improvement in performance for heavily compute- or memory-bound
Java tasks.

OS X does cache very aggressively when you have a lot of RAM, so it can
look like free memory is going away faster than you expect.  I often
load a large MRI volume into IJ, reslice it, save the resliced stack,
and ship the new stack across the network.  Saving the stack is limited
by disk speed -- but then transferring it across the network requires
NO disk activity, because IT'S ALL STILL CACHED.  As someone else
pointed out, when a process needs more working memory, this cached data
can be discarded with no overhead.  So, yes, OS X often reports very
little free RAM -- but this is a *good* thing.  You still don't see
many page-ins/outs, and that's the critical metric.

It's also worth pointing out again that, while the cheapest PC is
certainly less expensive than the cheapest Mac, comparably-configured
Macs and name-brand PCs are quite close in cost.  And, for US$79, you
can buy a copy of Parallels, and then you can run OS X and Linux
simultaneously.  Or, if you really, really hate OS X, you could just
boot straight into XP. :-)
--
        -jeffB (Jeff Brandenburg, Duke Center for In-Vivo Microscopy)