Do you see the difference on any sketch or only on some sketches? Ie. does the comparison change when running sketches in pure JAVA2D? P2D? P3D? OPENGL? Do the problematic sketches use some library?

I tried to isolate it to the simplest example:

1. void setup(){
     frameRate(120);
     size(600,600,P2D);
    // size(600,600,JAVA2D);
   }
   
   void draw(){
     ellipse(frameCount,frameCount,5,5);
     if(frameCount==2000)
       println(millis());
   }
   

When I run using P2D, the program prints 16682, when I use JAVA2D, it prints out 33947.

Also, I am noticing that up to frameRate(60), both are running at approximately the same speed but as I start going faster, JAVA2D pretty much stops getting any faster. In fact, I wouldn't be surprised if it turned out the JAVA2D was somehow getting clamped at 60 fps.

Uh? Your example compares two Processing modes, not two OSes. It is well known that these modes have different speeds, P2D was precisely made to be faster than JAVA2D, but it is less rich in some areas (limitations like no native fonts, quality...). But it wasn't the point of your initial message.

Sorry I am not being very clear. That previous post was an attempt at answering your question regarding what renderer. I am noticing this difference between the 2 OS's with both the P2D and the JAVA2D renderers, but JAVA2D is worst.

Here are the numbers printed out by my little program:

                  P2D            JAVA2D
win7          14020            13206
ubuntu      16682             33947

P2D is 19% slower on ubuntu but JAVA2D is 157% slower. I can live with the 19% but the 157% is so large I feel like I must be doing something wrong. Also, the fact that JAVA2D's actual frame rate seems to stop increasing above frameRate(60) while P2D continues to speed up makes me suspicious that something is not working right. (Maybe the two are unrelated)

I see, you had a good idea to test these two modes on both systems. I think then I have a possible explanation for the difference.
P2D (like P3D) mode is handled purely by Processing, which manipulates directly the pixels in the buffer, then use AWT only to display the result.
Java2D, on the other hand, is entirely managed by Java/the JVM. And I know that in recents versions of Java (starting with Java 5, but greatly improved in lastest versions of Java 6), it can use the native graphic pipeline/accelerator.
For example, in Windows, it uses DirectX, and you can set options like sun.java2d.d3d or sun.java2d.d3dtexbpp.
I don't know if they made such effort on Linux...
I found back an article: http://download.oracle.com/javase/1.5.0/docs/guide/2d/new_features.html
Ah, and another showing various options: http://download.oracle.com/javase/1.5.0/docs/guide/2d/flags.html

I used a nice little demo named JGears. Not the Web framework, seems to be a JOGL demo.
It was mentioned on http://acuriousanimal.blogspot.com/2007/04/ubuntu-edgy-ati-driver-and-java2d.html which, already at the time, mention speed issues with Java2D.
Ah, I found back the jar file (and project): http://trac-hg.assembla.com/jgears/ It can be a convenient tool to test the switches.
I kept the tests I did on my old computer (using a GeForce4 MX 440...):

> java -jar JGears.jar
Size: 512x512, 100 frames in 4250 ms = 23,529 FPS

> java -Dsun.java2d.opengl=false -Dsun.java2d.d3d=false -jar JGears.jar
Size: 512x512, 100 frames in 4172 ms = 23,969 FPS

> java -Dsun.java2d.opengl=false -Dsun.java2d.d3d=false -Dsun.java2d.noddraw=true -jar JGears.jar
Size: 512x512, 100 frames in 4172 ms = 23,969 FPS

> java -Dsun.java2d.opengl=true -jar JGears.jar
Size: 512x512, 100 frames in 3031 ms = 32,992 FPS

> java -Dsun.java2d.opengl=True -jar JGears.jar
OpenGL pipeline enabled for default config on screen 0
Size: 512x512, 100 frames in 2781 ms = 35,958 FPS

> java -Dsun.java2d.d3d=True -jar JGears.jar
Direct3D pipeline enabled on screen 0
Size: 512x512, 100 frames in 109297 ms = 0,915 FPS

> java -Dsun.java2d.d3d=True -Dsun.java2d.translaccel=true -jar JGears.jar
Direct3D pipeline enabled on screen 0
Size: 512x512, 100 frames in 88079 ms = 1,135 FPS

> java -Dsun.java2d.d3d=True -Dsun.java2d.translaccel=true -Dsun.java2d.d3dtexbpp=16 -jar JGears.jar
Direct3D pipeline enabled on screen 0
Size: 512x512, 100 frames in 82078 ms = 1,218 FPS

> java -Dsun.java2d.d3d=True -Dsun.java2d.d3dtexbpp=16 -jar JGears.jar
Direct3D pipeline enabled on screen 0
Size: 512x512, 100 frames in 71953 ms = 1,390 FPS
Size: 512x512, 100 frames in 9062 ms = 11,035 FPS <-- Deactivate anti-aliasing
Size: 512x512, 100 frames in 875 ms = 114,286 FPS <-- Deactivate translucent textured path
Size: 512x512, 100 frames in 688 ms = 145,349 FPS

> java -Dsun.java2d.opengl=True -jar JGears.jar
OpenGL pipeline enabled for default config on screen 0
Size: 512x512, 100 frames in 2828 ms = 35,361 FPS
Size: 512x512, 100 frames in 875 ms = 114,286 FPS <-- Deactivate anti-aliasing
Size: 512x512, 100 frames in 703 ms = 142,248 FPS <-- Deactivate translucent textured path
Size: 512x512, 100 frames in 672 ms = 148,810 FPS

It shows some combinations of such switches.
Hope that helps.

Great. That was it. With -Dsun.java2d.opengl=True, I get a comparable frame rate across OS's.
Thanks.

Now, for the next question, is there a way to setup processing so it uses this flag? I looked in the preference file but there doesn't seem to be an explicit java command.

If I read correctly (but quickly...) the code at processing/app/src/processing/app/debug/Runner.java, I would try to put the parameter in the run.options property of preferences.txt