Machines for 3D modelling... which???

Rick Venable rvenable at deimos.cber.nih.gov
Mon Dec 12 22:07:54 EST 1994


On 12 Dec 94 14:31:56 +1200 gjacobs at chmeds.ac.nz pontificated:
> Q:  Does anyone what machines are worth investigating (beyond SGIs, that is)
> for molecular modelling? What are their (comparative) performances for
> 3D work? 

HP 9000/7?5, DEC Alpha, IBM RS6000/5?? at least; both SGI and some IBMs have
*hardware* to do GL quickly, while HP and DEC machines have better raw number
crunching power.  For software using X windows, especially PEX, GL hardware
may not offer much advantage to SGI and IBM machines.  Your relative
performance may also be depend a good deal on the size ot the molecular
system, who wrote the software you are using, how well tuned and/or busy
the machine is, and not necessarily depend solely on who made the machine.

You may wish to consider the ease of obtaining service and/or replacement
parts and peripheral devices in your locale, as well as performance.  You
may wish to evaluate software first, then get the machine it runs best on.

> Note that some other companies do have the licence for the Silicon Graphics
> (SGI) graphics language (GL) - but I don't know which! Any ideas??

There is also a freely distributed software package for X11 which emulates
the GL calls on *any* X windows system, but possibly not with anything close
to the speed of hardware GL.  If the rate limiting factor is floating point
calculation speed, graphics rendering speed doesn't matter so much, though.

> In particular, is there anyway to have top-end 3D grpahics on a DEC Alpha?

Kubota makes an add-on graphics engine for DEC Alpha, especially for rendering
volumetric surfaces rapidly.  Contact Kubota or your DEC salesperson.

> Grant Jacobs
> ----------------------------------------------------------------

--
Rick Venable                  =====\     |=|    "Eschew Obfuscation"
FDA/CBER Biophysics Lab       |____/     |=|
Bethesda, MD  U.S.A.          |   \  /   |=|  / Not an official statement \
rvenable at deimos.cber.nih.gov       \/    |=|  \  or position of the FDA.  /



More information about the Molmodel mailing list