Another take home message is that if you want chimera to run faster drawing molecules, get a faster CPU. This is true for all platforms chimera runs on (Mac OS X, Windows, Linux, ...). For example, both the Quadro system and the HD 2600 systems mentioned had 2.8 Ghz CPU's and Apple sells 3.0 and 3.2 Ghz systems for $800 and $1600 additional cost respectively. I would expect the improvement to be roughly linear with the CPU speed change, so +7% and +14%. But we'll have to wait for someone to do the benchmarks to be sure. Being CPU limited is not a feature, but to fix it will take a rewrite of the chimera internals and we're still evaluating how best to do it. Greg Couch UCSF Computer Graphics Lab On Tue, 4 Mar 2008, Tom Goddard wrote:
Here's a comparison of Chimera graphics benchmarks for the current standard PowerMac graphics versus the top-of-the-line graphics:
Radeon HD 2600 - no additional cost Quadro FX 5600 - $2850 additional cost
Surfaces and meshes: 1.8x faster
Volume solid style: size 443**3 vs 325**3. Quadro system was limited 32-bit main memory address space.
Molecule: Neglible difference for 34000 atom model in wire, stick, ball and stick, ribbon, sphere styles.
The Quadro card has 1.5 Gbytes of memory and will probably have significantly higher performance on solid style volume rendering if the Chimera volume display code is optimized. Probably adding more main memory will not help because the limitation is the 32-bit address space (a limit of Mac OS graphical applications).
The Quadro card did not correctly display volume data in solid style with 3d textures (solid rendering 2d texture option turned off) -- displayed a chessboard pattern. Did not test 3d textures on Radeon card.
Tom