Re: CCL:Molecular Modeling hardware

[Sandra Antunez <antunez : at : chem.cinvestav.mx>]

 A couple of years ago in usenet and at the HotWired site, there was some
 discussion as to whether SGI could survive the Wintel onslaught.
 Technical issues such as system bus speed and polygon processing, as well
 as the ability of big-time users to have the actual operating system
 modified to suit their needs made it obvious that wintel boxes were
 precluded from the high end.
 At the time I expressed the worry that the low and medium ends were
 exposed, and would soon succumb. OpenGL is native in NT these days, and
 $1000 will get you some fairly reasonable polygon processing power. At
 this point, I suspect the main hurdle to be overcome in porting
 traditional SGI sw to NT is not graphics, but issues peculiar to UNIX
 such as process management, signals and other low-level programming thigs
 that differ between Unix and NT.
 NT 4.0 is the first "decent" MS operating system, which can support
 most
 of the functionality you expected from IRIX or Solaris. Intergraph and
 others now sell big-time graphics software on NT. SGI-based developers
 will have to do the same. I suspect the market will go to wintel over the
 course of two or three years to do both mainstream modelling and
 parallelized numbercrunching. Toy Story is doable on wintel boxes, the
 only thing missing is a port of the software. A note of caution, though.
 From a strictly visceral point of view, NT "feels" bulkier or klunkier
 than unix variants. It seems you need quite a lot more memory to get the
 same job done.
 Heinz Hemken
 hhemken : at : franchiseloan.com