Hi Grant,
    It's not exactly what you want, but there is a page that benchmarks Chimera on various graphics cards: http://www.cgl.ucsf.edu/chimera/benchmarks.html .  The 'Surface' column of the table in particular is directly related to how fast Chimera will draw surfaces and ball-and-stick.  Later this summer we will have a more detailed benchmark page with precisely what you want:  performance based on system size -- but we don't have that yet.  If you have access to any of the graphics cards you are considering, you can run the benchmark yourself -- it's in the Utilities category.
    I have a few tips for squeezing performance out for larger systems:

1) Use flat ribbons to depict the bulk of the system.  Show atoms only for the parts you are examining in detail.

2) Show the surface only on the parts you are examining in detail.  If you are using surfaces to get an idea of the overall shape of the molecule, try to use the low-resolution surfaces that Multiscale generates rather than the default high-detail molecular surfaces.

3) If you are using ball-and-stick because of visibility problems with wireframe rendering, try increasing the thickness of lines (which is under Actions->Atoms/Bonds).

                        Eric Pettersen

                        UCSF Computer Graphics Lab

                        pett@cgl.ucsf.edu

                        http://www.cgl.ucsf.edu



On Jul 13, 2005, at 4:52 AM, Grant Jacobs wrote:

Forgive me for asking a silly question, but can people here recommend what current GPUs are able to present large molecular systems in Chimera with smooth rendering of a surface or ball&stick rendering, without "jumping" on rotations, etc.  By large complexes, I mean the likes of the various multi-protein complexes; photosytem complexes, the ribosome and the like.

I'm currently reviewing GPUs for a system upgrade, but its fiendishly hard to translate the marketing information into something practical! (I'm sure everyone is familiar with that particular problem...) In particular, I would like to aim at a system where I avoid having to "trim" large complexes to a workable size (I'm aware there are RAM issues with this).

Currently I'm looking at a Linux-based solution, but would consider a G5 (Apple), if its capable.

Anyone's comments on their own experience would be appreciated.

It'd be lovely if there were a rotation speed vs. number of atoms vs. graphics card plot!  Somehow I doubt there is... :-)


Grant

-- 
-------------------------------------------------------------------
Grant Jacobs Ph.D.                                     BioinfoTools
ph. +64 3 478 0095  (office, after 10am)               PO Box 6129,
or  +64 27 601 5917 (mobile)                               Dunedin,
gjacobs@bioinfotools.com                               NEW ZEALAND.
   Bioinformatics tools: deriving knowledge from biological data
Bioinformatics tools - software development - consulting - training
Check out the website for more details: http://www.bioinfotools.com

The information contained in this mail message is  confidential and
may be legally privileged.  Readers of this message who are not the
intended recipient are hereby notified that any use, dissemination,
distribution or reproduction of this message is prohibited.  If you
have received this message in error please notify the sender immed-
iately and destroy the original message.  This applies also to  any
attached documents.
_______________________________________________
Chimera-users mailing list
Chimera-users@cgl.ucsf.edu
http://www.cgl.ucsf.edu/mailman/listinfo/chimera-users