[Chimera-users] Question on Chimera
gregc at cgl.ucsf.edu
Mon Mar 6 11:58:47 PST 2006
On Fri, 3 Mar 2006, Seiha Thorng wrote:
> My name is Seiha Thorng and I work for BioMarin Pharmaceuticals, Inc.
> I'm looking into UCSF Chimera as a promising software for our research
> purposes. I would like to ask you what equipment you would recommend
> for optimal performance of this software. The answer I'm looking for is
> in terms of ram, platform, processor, disc space ....etc. We would like
> very much to be able to run this program at the best possible
> performance. Thank you.
The answer to your question depends on what size of the systems you are
going to be viewing/modelling, and what other software you will be using.
You should create a representative set of sample data and the operations
you wish to perform on them, and then get several vendors to supply you
with a demo system, so you can test how well they work for your data. It
also helps to have a budget -- you can easily spend $10,000 or more for
an "optimal" system if wanted. And maybe you have other constraints,
sometimes companies only buy computers from one vendor to simplify support
Some high-end examples:
(1) An Apple Quad 2.5GHz PowerPC G5 with 2GB ECC SDRAM, 250GB disk, Quadro
FX 4500, 30" Cinema HD display -- $7,948.
(2) An HP xw9300 Dual Dual-core 2.6GHz Opteron, 2BG ECC, 74BG 10K-rpm
disk, Quadro FX 4500, 23" widescreen monitor -- $9,603.
The HP system would be more optimal than the Apple system, but both would
be great systems. If you wanted to view your molecules in stereo, you
would have to buy a separate CRT that could handle the higher refresh
rates (see <http://www.stereographics.com> for stereo hardware help). You
could make the HP system better with 15K-rpm SCSI disks and a second SLI
Of course, depending on the size of your data, you might not need a
workstation. Smaller systems work well on home computers, especially
ones designed for gaming. You really need to use your own data on real
computers to be able to tell for sure.
Rules of thumb:
More than 2 cpus (or one dual-core cpu) doesn't help chimera (yet!), so
you could scimp on that. Faster cpus are better because ball&stick
visualization is partially cpu limited. Even one cpu will work, but the
computer will be more responsive with two.
Chimera uses a lot of memory, so 2GB is reasonable for 2 cpus -- if you're
displaying volume data, you will want more, if your displaying huge
molecules, you'll want more. You'll also want more memory if you have
other programs running at the same time. Memory is fairly cheap, so get
Chimera doesn't need much disk space, so if you data doesn't either, buy
the faster disk.
If you want to view your molecule in stereo, you have to buy a
workstation-class graphics card -- an ATI FireGL, NVidia Quadro, or 3DLabs
Wildcat. The Quadro FX 4500 in the examples above is the current best
graphics card in our benchmarks, but we have not benchmarked the new
3DLabs Wildcat Realizm 800, nor the ATI FireGL V7100. See
<http://www.cgl.ucsf.edu/chimera/benchmarks.html> for the benchmark
numbers we have. My hunch is that the 3DLabs card is as good, if not
better than, the NVidia one, but I don't know for sure.
Hope this helps, good luck,
UCSF Computer Graphics Lab
More information about the Chimera-users