Dirk, Your results are consistent with my experience.
SGIs fell behind PCs in price/performance long ago, and have been way behind PCs in raw performance for at least the last couple of years (except for high-end multiprocessing). This applies to both CPU and graphics performance. The same can now be said for Suns. In my view, the only intrinsic argument for purchasing traditional brand-name Unix hardware is to run existing applications that for some reason can not be run under Linux, OSX, or Windows. Buying SGIs or Suns to run PyMOL would be a waste of money (and end-user time). If you want a supported Unix hardware setup, then buy a fast Mac. Otherwise, buy a dual-processor PC with lots of RAM and a fast graphics card and then run Linux or Windows. For the price of a comparable SGI, you could buy a whole cluster of these machines. Stereo now runs almost anywhere. Cheers, Warren On Thu, 27 Feb 2003, Dirk Kostrewa wrote: > in your experience, how does the graphical performance of pymol (or any other > graphical program) on a PC running under Linux with a good stereo-capable > graphics card (Nvidia Quadro, ATI FireGL) compare to a stereo-capable SGI > workstation (O2, Octane, Fuel, Onyx)? I've got an old Pentium III PC with 450 > MHz and a (non-stereo) Nvidia GeForce2 MX440 graphics card side-by-side with > an SGI O2 with a R12000 processor: a simple pymol (0.86) movie of a protein > sketch at 800x800 resolution and maximum display quality requires on the SGI > 3 minutes for 180 frames, i.e. 1 frame/sec, whereas the same movie requires > on the PC 20 seconds, i.e. 9 frames/sec. I wonder how this can be and would > like to hear about your experiences. > I've compiled pymol from the source file with the mips pro compiler from SGI. > Results were comparable with the previous pymol version for which I've > installed directly the binaries. Thus, I don't think that the speed > difference is due to a compiler problem on the SGI. I've got another graphics > program for protein modelling, Moloc, that runs on both platforms. There, > I've observed similar subjective speed differences in favour to the PC. But I > couldn't measure any benchmarks. > > Best regards, > > Dirk Kostrewa. > >
