ImageJ

ImageJ benchmarks

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

18 messages Options

Albert Cardona

ImageJ benchmarks

ImageJ's Plugins - Utilities Benchmark on the Mandril sample image (RGB
512x512):

(Best of 3 runs, ImageJ freshly restarted)

MacOSX 10.4 java 1.6.0-beta 32-bit on a powerpc G5 2.0 Ghz 1.5 Gb
SDRAM: 22.8 seconds

Ubuntu 7.10 java 1.6.0 64-bit on an intel core 2 quad 2.4 Ghz 4 Gb
DDR3: 0.741 seconds

Ubuntu 7.04 java 1.6.0 32-bit on an intel core 2 duo 2.2 Ghz 2 GB DDR2:
0.802 seconds

The difference is staggering. That G5 is about 2 years old only.

I would appreciate if people posts benchmarks on their machines,
to get an idea on what is ImageJ (and TrakEM2) going to be run on.

Thanks!

Albert

--
Albert Cardona
http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

Joachim Wesner

Antwort: ImageJ benchmarks

OK, let´s start

No longer up-to-date AMD-XP-3000+ (2.16GHz), WinXP, 768 MB-DDR

3.2sec

Centrino 1.6GHz (Single-Core) Subnotebook, WinXP, 512 MB-DDR

5.0 sec

Joachim Wesner

Albert Cardona
<[hidden email]
S.ETHZ.CH> An
Gesendet von: [hidden email]
ImageJ Interest Kopie
Group
<[hidden email]. Thema
GOV> ImageJ benchmarks

23.01.2008 18:27

Bitte antworten
an
ImageJ Interest
Group
<[hidden email].
GOV>

ImageJ's Plugins - Utilities Benchmark on the Mandril sample image (RGB
512x512):

(Best of 3 runs, ImageJ freshly restarted)

MacOSX 10.4 java 1.6.0-beta 32-bit on a powerpc G5 2.0 Ghz 1.5 Gb
SDRAM: 22.8 seconds

Ubuntu 7.10 java 1.6.0 64-bit on an intel core 2 quad 2.4 Ghz 4 Gb
DDR3: 0.741 seconds

Ubuntu 7.04 java 1.6.0 32-bit on an intel core 2 duo 2.2 Ghz 2 GB DDR2:
0.802 seconds

The difference is staggering. That G5 is about 2 years old only.

I would appreciate if people posts benchmarks on their machines,
to get an idea on what is ImageJ (and TrakEM2) going to be run on.

Thanks!

Albert

--
Albert Cardona
http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email
______________________________________________________________________

lechristophe

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

MacOSX 10.4.11 java 1.6.0 beta on an iMac 2.16GHz Intel Core 2 Duo (2Go 667
MHz DDR2 RAM):

1.025s

On Jan 23, 2008 6:27 PM, Albert Cardona <[hidden email]> wrote:

> ImageJ's Plugins - Utilities Benchmark on the Mandril sample image (RGB
> 512x512):
>
> (Best of 3 runs, ImageJ freshly restarted)
>
>
> MacOSX 10.4 java 1.6.0-beta 32-bit on a powerpc G5 2.0 Ghz 1.5 Gb
> SDRAM: 22.8 seconds
>
>
> Ubuntu 7.10 java 1.6.0 64-bit on an intel core 2 quad 2.4 Ghz 4 Gb
> DDR3: 0.741 seconds
>
>
> Ubuntu 7.04 java 1.6.0 32-bit on an intel core 2 duo 2.2 Ghz 2 GB DDR2:
> 0.802 seconds
>
>
>
> The difference is staggering. That G5 is about 2 years old only.
>
> I would appreciate if people posts benchmarks on their machines,
> to get an idea on what is ImageJ (and TrakEM2) going to be run on.
>
> Thanks!
>
> Albert
>
> --
> Albert Cardona
> http://www.mcdb.ucla.edu/Research/Hartenstein/acardona
>

Jon Harman

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

Windows XP Intel Core 2 Quad 2.4 Ghz .735 sec

I'm surprised at the similarity between the core 2 quads and duos.
Perhaps the mandrill image is too small. Image display may be a big
part of the benchmark

Jon
Albert Cardona wrote:
> ImageJ's Plugins - Utilities Benchmark on the Mandril sample image
(RGB 512x512):
>
> (Best of 3 runs, ImageJ freshly restarted)
>
>
> MacOSX 10.4 java 1.6.0-beta 32-bit on a powerpc G5 2.0 Ghz 1.5 Gb
SDRAM: 22.8 seconds
>
>
> Ubuntu 7.10 java 1.6.0 64-bit on an intel core 2 quad 2.4 Ghz 4 Gb
DDR3: 0.741 seconds
>
>
> Ubuntu 7.04 java 1.6.0 32-bit on an intel core 2 duo 2.2 Ghz 2 GB
DDR2: 0.802 seconds
>
>
>
> The difference is staggering. That G5 is about 2 years old only.
>
> I would appreciate if people posts benchmarks on their machines,
> to get an idea on what is ImageJ (and TrakEM2) going to be run on.
>
> Thanks!
>
> Albert
>

ctrueden

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

Processor: 2 x 2.66 GHz Dual-Core Intel Xeon
Memory: 5 GB 667 Mhz DDR2 FB-DIMM

Mac OS X 10.4.11, ImageJ v1.39l, java 1.5.0_13 : 0.96 seconds

Same machine, using Parallels v2 build 3214 with 512 MB RAM:

Windows XP, ImageJ v1.39f, java 1.6.0_02 : 1.352 seconds
Ubuntu Gutsy, ImageJ v1.39l, java 1.6.0_03 : 1.014 seconds
Ubuntu Gutsy, ImageJ v1.39l, java 1.5.0_09 : 2.733 seconds

On Jan 23, 2008 11:27 AM, Albert Cardona <[hidden email]> wrote:

Albert Cardona

Re: ImageJ benchmarks

In reply to this post by Jon Harman

> Windows XP Intel Core 2 Quad 2.4 Ghz .735 sec
>
> I'm surprised at the similarity between the core 2 quads and duos.
> Perhaps the mandrill image is too small. Image display may be a big
> part of the benchmark
>

As far as I know, the benchmark is single-threaded, which would then
make it dependent on the max speed of a single core. Also the graphics
card, front bus, RAM type, etc. play a role.

Albert

--
Albert Cardona
http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

drix

Re: ImageJ benchmarks

In reply to this post by Jon Harman

On MacBook Pro 2 GHz Intel Core Duo and 2 GB 667 MHz DDR2 SDRAM Mac OS
X 10.4.11 (8S2167)
with ImageJ 1.38x and java version "1.5.0_13".

1.5--1.6 s

Drix

On Jan 23, 2008 1:03 PM, Jon Harman <[hidden email]> wrote:

> Windows XP Intel Core 2 Quad 2.4 Ghz .735 sec
>
> I'm surprised at the similarity between the core 2 quads and duos.
> Perhaps the mandrill image is too small. Image display may be a big
> part of the benchmark
>
> Jon
>
> Albert Cardona wrote:
> > ImageJ's Plugins - Utilities Benchmark on the Mandril sample image
> (RGB 512x512):
> >
> > (Best of 3 runs, ImageJ freshly restarted)
> >
> >
> > MacOSX 10.4 java 1.6.0-beta 32-bit on a powerpc G5 2.0 Ghz 1.5 Gb
> SDRAM: 22.8 seconds
> >
> >
> > Ubuntu 7.10 java 1.6.0 64-bit on an intel core 2 quad 2.4 Ghz 4 Gb
> DDR3: 0.741 seconds
> >
> >
> > Ubuntu 7.04 java 1.6.0 32-bit on an intel core 2 duo 2.2 Ghz 2 GB
> DDR2: 0.802 seconds
> >
> >
> >
> > The difference is staggering. That G5 is about 2 years old only.
> >
> > I would appreciate if people posts benchmarks on their machines,
> > to get an idea on what is ImageJ (and TrakEM2) going to be run on.
> >
> > Thanks!
> >
> > Albert
> >
>

Michael Schmid

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

Hi Albert,

your OSX result is very strange!

On a similar machine (also 2 year sold) I got 3.5 seconds:
OSX 10.4.10, 2.1GHz PowerPC G5, 1GB DDR2 SDRAM, ATI Radeon X600 XT
(128 MB)
256 MB reserved for ImageJ and a lot of other applications open.
I have Java 1.5.0_07.

Maybe your Mac has a very slow graphics card? OSX performance
is known to depend significantly on the graphics performance,
and Benchmark.java contains a lot of updateScreen(imp);
commands.

On my OSX G5 machine, as an external plugin with
boolean showUpdates=false;
i.e. no image updates, it needs 2.6 s.

Concerning the remark of Jon about the similarity of Core Duo
and Quad, none of the number crunching routines called by the
benchmark is multithreaded and can take advantage of multi-
core CPUs.
Currently, ImageJ plugins use multithreading only for
processing stack slices (if the PARALLELIZE_STACKS flag is set -
it is for the majority of ImageJ built-in filter operations).

Michael
________________________________________________________________

On 23 Jan 2008, at 19:03, Jon Harman wrote:
> Windows XP Intel Core 2 Quad 2.4 Ghz .735 sec
>
> I'm surprised at the similarity between the core 2 quads and duos.
> Perhaps the mandrill image is too small. Image display may be a
> big part of the benchmark

On 23 Jan 2008, at 18:27, Albert Cardona wrote:

> ImageJ's Plugins - Utilities Benchmark on the Mandril sample image
> (RGB 512x512):
>
> (Best of 3 runs, ImageJ freshly restarted)
>
>
> MacOSX 10.4 java 1.6.0-beta 32-bit on a powerpc G5 2.0 Ghz 1.5 Gb
> SDRAM: 22.8 seconds
>
>
> Ubuntu 7.10 java 1.6.0 64-bit on an intel core 2 quad 2.4 Ghz 4 Gb
> DDR3: 0.741 seconds
>
>
> Ubuntu 7.04 java 1.6.0 32-bit on an intel core 2 duo 2.2 Ghz 2 GB
> DDR2: 0.802 seconds
>
>
>
> The difference is staggering. That G5 is about 2 years old only.
>
> I would appreciate if people posts benchmarks on their machines,
> to get an idea on what is ImageJ (and TrakEM2) going to be run on.
>
> Thanks!
>
> Albert
>
> --
> Albert Cardona
> http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

Albert Cardona

Re: ImageJ benchmarks

Hi Michael, hi all,

> your OSX result is very strange!

You are right, and others have pointed that out too. Wayne suggested
that 1.6.0-beta is not doing JIT-compilation on the G5.

I know now that I have to fix that machine somehow, if anything with
1.5.0_13. Why in the world java is not installed from a third party like
in all other OSes is beyond me: Apple always lags one or two years
behind (and never mind that there are no means to unistall. An apple
document online states that one needs to reinstall the _operating
system_ to do so.)

Thank you everyone for the contributions so far!

Albert

--
Albert Cardona
http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

Jon Harman

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

Knoppix 5.1.1CD Java 1.5.0_10 IJ 1.38x running under qemu on Windows XP,
Intel Core 2 Quad 2.4 Ghz 3 Gb: 3.66 sec.

Interestingly this benchmark takes 6.66 seconds using the jre included
in the ImageJ Linux distribution. Other routines (my own plugin for
example) take about the same time for Knoppix jre vs the one in the
ImageJ directory.

Albert Cardona wrote:

Joachim Wesner

Antwort: Re: ImageJ benchmarks

In reply to this post by Michael Schmid

Hi

The sourcecode of Benchmark.java has a boolean that controls display
updating, if you set it to false, I get 2.9 vs. 3.2 sec on my "fast"
machine, only 10% difference and I use
a comparatively slow graphics card.

Reagarding the G5 results, my slowest systems is my Linux fileserver, which
is based on a pretty old low-power 700 MHz VIA C3 (x386 clone), a CPU
notoriously know for pretty bad
float performance. It nevertheless runs the benchmark (graphics update off)
in 36 secs. (java 1.5.0x)

Mit freundlichen Grüßen / Best regards

Joachim Wesner

Michael Schmid
<[hidden email]
N.AC.AT> An
Gesendet von: [hidden email]
ImageJ Interest Kopie
Group
<[hidden email]. Thema
GOV> Re: ImageJ benchmarks

23.01.2008 19:27

Bitte antworten
an
ImageJ Interest
Group
<[hidden email].
GOV>

Hi Albert,

your OSX result is very strange!

On a similar machine (also 2 year sold) I got 3.5 seconds:
OSX 10.4.10, 2.1GHz PowerPC G5, 1GB DDR2 SDRAM, ATI Radeon X600 XT
(128 MB)
256 MB reserved for ImageJ and a lot of other applications open.
I have Java 1.5.0_07.

Maybe your Mac has a very slow graphics card? OSX performance
is known to depend significantly on the graphics performance,
and Benchmark.java contains a lot of updateScreen(imp);
commands.

On my OSX G5 machine, as an external plugin with
boolean showUpdates=false;
i.e. no image updates, it needs 2.6 s.

Concerning the remark of Jon about the similarity of Core Duo
and Quad, none of the number crunching routines called by the
benchmark is multithreaded and can take advantage of multi-
core CPUs.
Currently, ImageJ plugins use multithreading only for
processing stack slices (if the PARALLELIZE_STACKS flag is set -
it is for the majority of ImageJ built-in filter operations).

Michael
________________________________________________________________

On 23 Jan 2008, at 19:03, Jon Harman wrote:
> Windows XP Intel Core 2 Quad 2.4 Ghz .735 sec
>
> I'm surprised at the similarity between the core 2 quads and duos.
> Perhaps the mandrill image is too small. Image display may be a
> big part of the benchmark

On 23 Jan 2008, at 18:27, Albert Cardona wrote:

______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email
______________________________________________________________________

Michael Schell

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

Mandrill wars:

Mac OSX 10.4, dual G4 1.25 GHz 1.3 GB of 2.0 GB DDR allocated to IJ
1.37: 4.2 sec

Mac OSX 10.4, G5 2 GHz 1.2 of 2.0 GB DDR2 allocated it IJ 1.38: 3.5
sec

Mac OSX 10.5, Core duo 1.83 GHz 1.3 of 2.0 GB DDR2 allocated to IJ
1.37: 1.4 sec

XP32 dual Xeon (single cores) 3.6 GHz, high-end Quadro graphics
card, 1.4 of 4.0 GB DDR2 allocated to IJ 1.38: 1.02 sec

Same XP machine as above running XP64, with 5.0 of 8.0 GB DDR2
allocated to IJ 1.38: .812 sec

On Jan 23, 2008, at 12:27 PM, Albert Cardona wrote:

Michael J. Schell, Ph.D.
Assist. Professor
Dept. of Pharmacology
Uniformed Services University
4301 Jones Bridge Rd.
Bethesda, MD 20814-3220
tel: (301) 295-3249
[hidden email]

Mathieu Marchand-2

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

since I have access to a lot of different configs, here is a few
benchmarks (all with ImageJ v1.39n).
I minimized the windows on my servers as the remote access / slow
graphics were slowing down everything.
the new (and much anticipated) Apple JDK 1.6dp8 (only for x86_64 and
10.5 so far) looks pretty fast...

- imac Core2Duo 2.16Ghz - 667Mhz bus - 2x1GB DDR2 667 - Nvidia 7600GT
- OSX 10.5.1 - Apple JDK 1.6dp8 x86_64
0.85 sec
same with leopard's default Apple JDK 1.5.0_13
1.11 sec

- macpro Xeon 5160 3GHz- 4GB RAM (4x1GB FBDDR667) - OSX 10.5.1 - ATI
X1900 - Apple JDK 1.6dp8 x86_64
0.54sec
same with WinXP pro SP2 32 bits (bootcamp) - JRE JRockit 6 R27.4 32 bits
0.61sec

- Dell PowerEdge 1950 - Xeon 5160 - 8GB RAM (4x2GB FBDDR667) - RHEL4
x86_64 - Java 1.6.0_2 64bits
0.55sec - with window minimized
same than above but with jrockit-R27.4.0-jdk1.6.0_02 x86_64
0.51sec - with window minimized

- Dell Precision 690 - Xeon 5160 - 16GB Ran (8x2GB FBDDR667) - Windows
2003 Server x64 - jrockit-R27.4.0-jdk1.6.0_02 x86_64
0.51sec
same as above with Sun JRE 1.6.0_3 32bits
0.93sec

- AMD Opteron 2212 2GHZ - 4GB DDR2 - Centos4 x86_64 -
jrockit-R27.4.0-jdk1.6.0_02 x86_64
1.13sec - with window minimized

- AMD Opteron 856 2.8GHz - 16GB DDR - Centos4 x86_64 -
jrockit-R27.4.0-jdk1.6.0_02 x86_64
0.90sec - with window minimized

On Jan 23, 2008 12:27 PM, Albert Cardona <[hidden email]> wrote:

--
Mathieu Marchand
--
Bio-Imaging Resource Center, The Rockefeller University
1230 York Avenue, box 209, New York, NY 10021
http://www.rockefeller.edu/bioimaging .
+1-212-327-7487 (7489 for fax)
+1-646-457-0430 (cell)

Michael Doube

Re: ImageJ benchmarks

In reply to this post by Michael Schell

>> I would appreciate if people posts benchmarks on their machines,
>> to get an idea on what is ImageJ (and TrakEM2) going to be run on.

Ubuntu 7.10 64-bit java 1.6.0_03 IJ 1.39p on an Intel Core 2 Duo 2.2 GHz
2GB DDR2 (Sony Vaio SZ650): 2.635 seconds

NeLaS

Re: ImageJ benchmarks

Laptop Toshiba Satellite A100
Intel Core Duo 2.0 GHz
1.0 GB RAM
Ubuntu Studio 7.10 (32bit)
kernel 2.6.22-14-rt
java 1.6.0_04
IJ 1.39p

1.625s

On Jan 25, 2008 10:32 AM, Michael Doube < [hidden email]> wrote:

> >> I would appreciate if people posts benchmarks on their machines,
> >> to get an idea on what is ImageJ (and TrakEM2) going to be run on.
>
> Ubuntu 7.10 64-bit java 1.6.0_03 IJ 1.39p on an Intel Core 2 Duo 2.2 GHz
> 2GB DDR2 (Sony Vaio SZ650): 2.635 seconds
>

--
Bruno C. Vellutini
organelas.com | ccnelas.org | desertoresdaescada.com

Centro de Biologia Marinha (CEBIMar)
Universidade de São Paulo
Av. Manoel H. do Rego km 131,5
11600-000 São Sebastião, SP, Brasil
http://www.usp.br/cbm/

suendermann

Re: ImageJ benchmarks

a little late...

AMD Athlon 64 X2 Dual Core Processor 3600+
Microsoft Vista 32 bit
2,0 GB RAM
ImageJ 1.38s
Java 1.6.0_01

1.705 sec

Albert Cardona

Re: ImageJ benchmarks

In reply to this post by Albert Cardona

Dear all,

Thank you for your contributions (and time!)

As computers have improved a lot since the benchmark was put together,
there is a need for a new benchmark for ImageJ and java systems in general.

We have noted the following:

1 - since chips are now so fast, the benchmark depends a lot more on the
screen refreshing rate -and thus the graphics card- than on the CPU itself.

2 - the benchmark doesn't last enough to trigger dynamic scaling of some
CPUs, particularly on laptops.

3 - the benchmark does not consider multithreading, very relevant now
with multicore CPUs being the norm.

Currently, there is no proper way to test how well will a computer
perform with java applications, particularly java imaging apps like ImageJ.

Eventually "we" (Stephan Preibisch, Johannes Schindelin, myself) will
put together a new one that tests integer performance, floating-point
performance, screen updating etc. separately.

If anyone has or knows of such a benchmark already, I would appreciate
email notice of it.

Thanks again,

Albert

--
Albert Cardona
http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

ctrueden

Re: ImageJ benchmarks

Hi Albert,

I thought it was worthwhile to preserve these numbers in some form, so I
created a FAQ entry on the wiki:

http://imagejdocu.tudor.lu/imagej-documentation-wiki/faq/are-there-performance-statistics-to-compare-against-my-system

If you create a better benchmark, it should be easy enough to update the
information there, too.

-Curtis

On Jan 28, 2008 10:05 AM, Albert Cardona <[hidden email]> wrote:

> Dear all,
>
> Thank you for your contributions (and time!)
>
> As computers have improved a lot since the benchmark was put together,
> there is a need for a new benchmark for ImageJ and java systems in
> general.
>
> We have noted the following:
>
> 1 - since chips are now so fast, the benchmark depends a lot more on the
> screen refreshing rate -and thus the graphics card- than on the CPU
> itself.
>
> 2 - the benchmark doesn't last enough to trigger dynamic scaling of some
> CPUs, particularly on laptops.
>
> 3 - the benchmark does not consider multithreading, very relevant now
> with multicore CPUs being the norm.
>
>
> Currently, there is no proper way to test how well will a computer
> perform with java applications, particularly java imaging apps like
> ImageJ.
>
> Eventually "we" (Stephan Preibisch, Johannes Schindelin, myself) will
> put together a new one that tests integer performance, floating-point
> performance, screen updating etc. separately.
>
> If anyone has or knows of such a benchmark already, I would appreciate
> email notice of it.
>
> Thanks again,
>
> Albert
>
> --
> Albert Cardona
> http://www.mcdb.ucla.edu/Research/Hartenstein/acardona
>