Hi Glenn, Did you have the Model Panel open while moving the molecule? We have found that due to an underlying bug in the Aqua versions of some windowing libraries, performance can be heavily impacted if the Model Panel is being displayed (as well as other tools that use the same kind of table as the Model Panel, namely: Axes, ModBase, ViewDock, Rotamers, and CASTp). --Eric On Jun 1, 2008, at 8:13 PM, Glenn Millhauser wrote:

Hi Folks, I just saw the mac aqua version at the download site and decided to give it a try. Layout looks terrific. However, it seems to move molecules much more slowly than the X windows version. I'm running on a macbook (intel) with inherent limited graphics capabilities, so perhaps that's the problem. Any comments? thanks, g

Glenn Millhauser Department of Chemistry & Biochemistry UC Santa Cruz Santa Cruz, CA 95064

831 459 2176 voice 831 459 2935 fax 831 566 3337 cell

glennm@chemistry.ucsc.edu http://chemistry.ucsc.edu/~glennm

_______________________________________________ Chimera-users mailing list Chimera-users@cgl.ucsf.edu http://www.cgl.ucsf.edu/mailman/listinfo/chimera-users

Hi Glenn, Your benchmark numbers on the MacBook with Intel GMA950 graphics suggest molecule display is 2-3x slower with Aqua Chimera vs X11 Chimera. But it appears you are using an older X11 Chimera because the ribbon benchmark frame rate is 73 which is too high to be right and older X11 Chimera versions gave incorrect high ribbon frame rates. I changed the benchmark code recently because the older Chimera versions gave inflated frame rates because they did not actually draw to the screen. If you tested the current X11 snapshot you will get smaller frame rates reported although actual performance has not changed. On your iMac with Radeon HD2600 graphics the Aqua version appears to be about 2x faster than the X11 version. That is unexpected too. I don't know why the Aqua and X11 OpenGL performance is so different on the machines you tested. Tom glenn millhauser wrote:

Hi Tom and Eric, I ran the benchmarks on my macbook and sent the results to Help (and included below). From my untrained eye, it looks as though there are performance differences. From the user perspective, as I noted in the previous email, rotating the molecule in the aqua version using the laptop feels jerky. However, I also did benchmarks on my imac, fairly new one with intel dual 2.8 GHz processors, and they are much closer to each other. These were also forwarded through Help. I did not have the Model Panel open. Hope this helps. all the best, g

macbook dual intel gma950 graphics

aqua Molecule 1f4h (33805 atoms): Null Wire Stick BStick Ribbon Sphere Ops 32.5 14.0 3.0 2.8 9.8 1.5 27.8

X11 1F4H (33805): 141.0 56.0 5.0 5.0 73.0 3.0 31.0

iMac dual intel ATI,RadeonHD2600 graphics

aqua Molecule 1f4h (33805 atoms): Null Wire Stick BStick Ribbon Sphere Ops 60.5 60.2 5.5 5.5 15.4 8.6 36.3

X11 1F4H (33805): 8.0 102.0 3.0 3.0 8.0 4.0 39.0

On Jun 2, 2008, at 11:29 AM, Thomas Goddard wrote:

...

Hi Glenn,

I compared the molecule rendering speed with Mac X11 and Mac Aqua Chimera snapshots (version 1.2524) and found them to be the same on an Intel Mac PowerBook with Radeon X1600 graphics and on a Power Mac G5 with Radeon ATI 9800 pro graphics. Here are scores for the G5 system

Aqua Molecule 1f4h (33805 atoms): Null Wire Stick BStick Ribbon Sphere Ops 59.8 59.8 3.4 3.7 11.1 6.0 13.4

X11 Molecule 1f4h (33805 atoms): Null Wire Stick BStick Ribbon Sphere Ops 70.8 86.3 3.4 3.7 9.4 5.9 14.9

(The aqua values are clamped to my display refresh of 60 Hz.) You could try this on your system using Tools / Utilities / Benchmark. You can click the "Show individual test controls" checkbutton and press the "Molecule" button to just run the molecule tests.

If the numbers do come out much different for X11 versus Aqua could you use Help / Report A Bug... to send us the numbers? Using that dialog will send us the technical specs for your graphics.

Tom

Glenn L. Millhauser

Department of Chemistry & Biochemistry

UC Santa Cruz

Santa Cruz, CA 95064

Hello, I used Chimera Aqua some time ago, I tried it for two-three days on a Macbook Pro with Leopard (at that time 10.5.2). I have to say that it worked for five minutes at maximum before it crashed. The same situation with a new Mac Pro. Then I used a X11 version, which is more stable, but also has a tendency to crash (sporadically). The biggest trouble was to load more than 8 volumes. On a Mac it was extremely slow and usually crashing. On a similarly configured, but Linux-based machine there were no problems encountered and it was clearly much faster. Good luck with it, thanks for all the effort! Best regards, Jean-Paul Armache

The current Chimera Aqua daily builds are much better and more stable than the older Aqua builds. We are getting reports that the new Aqua version is slower than the X11 version on Macs with Intel graphics, and comparable speeds on Macs with ATI or NVidia graphics. We are spending more time on the Aqua builds and hope to resolve that and other issues soon. With a little bit of luck, we will replace the Mac X11 version with the Aqua version by the next production release. And if not the next one, the one after that. So please try out a daily build at <http://www.cgl.ucsf.edu/chimera/alpha-downloads.html>, I think you will be pleasantly suprised. Greg Couch UCSF Computer Graphics Lab On Mon, 2 Jun 2008, Jean-Paul Armache wrote:

Hello,

I used Chimera Aqua some time ago, I tried it for two-three days on a Macbook Pro with Leopard (at that time 10.5.2). I have to say that it worked for five minutes at maximum before it crashed. The same situation with a new Mac Pro. Then I used a X11 version, which is more stable, but also has a tendency to crash (sporadically). The biggest trouble was to load more than 8 volumes. On a Mac it was extremely slow and usually crashing. On a similarly configured, but Linux-based machine there were no problems encountered and it was clearly much faster.

Good luck with it, thanks for all the effort!

Best regards,

Jean-Paul Armache _______________________________________________ Chimera-users mailing list Chimera-users@cgl.ucsf.edu http://www.cgl.ucsf.edu/mailman/listinfo/chimera-users

Hi Jean-Paul, Chimera cannot subtract two maps. Is that what you had in mind? or is it more complicated requiring determining equivalent normalizations for the maps? Do you have in mind that the maps have identical grids, or would one map need to be interpolated? There's underlying Python code to do all this in Chimera but no user interface. Tom

On Sun, 1 Jun 2008, Glenn Millhauser wrote:

Hi Folks, I just saw the mac aqua version at the download site and decided to give it a try. Layout looks terrific. However, it seems to move molecules much more slowly than the X windows version. I'm running on a macbook (intel) with inherent limited graphics capabilities, so perhaps that's the problem. Any comments?

So it turns out this is due to chimera asking the graphics to multisample each pixel to improve the image quality (it reduces the jagged edge effect, especially noticable on ribbons and surfaces). This was not implemented in Apple's X11 OpenGL, so we didn't have this problem before. Here are Chimera Aqua benchmark results for a MacBook 2Ghz Core Duo with Intel GMA 650 graphics running OS X 10.4.11: Volume benchmark scores surface mesh contour solid recolor 1 16 157 1 3 Molecule 1f4h (33805 atoms): Null Wire Stick BStick Ribbon Sphere Ops 29.8 7.5 1.6 1.4 4.3 0.3 26.3 And without multisampling: Volume benchmark scores surface mesh contour solid recolor 128 139 156 138 88 Molecule 1f4h (33805 atoms): Null Wire Stick BStick Ribbon Sphere Ops 72.2 41.8 14.1 14.1 26.5 10.2 26.7 So without multisampling, it's 6-100 times faster! Unfortunately, there's no way to give command line arguments when clicking on the chimera icon, so to turn off multisampling for now, you would need to start up a Terminal and type: ./Desktop/Chimera.app/Contents/Resources/bin/chimera --nomultisample That's assuming the Chimera.app is on your personal desktop. Or you could edit the Chimera.app/Contents/Resources/share/chimeraInit.py file so that nomultisample defaults to True instead of False on line 52. We will figure out an alternative solution that by default doesn't ask for multisampling for Intel motherboard graphics (same issue exists with Microsoft Windows). Greg Couch UCSF Computer Graphics Lab