wescott@LNIC1.HPRC.UH.EDU (Andrew M. Wescott) (04/25/89)
Well I don't know how familiar you are with the DN 10000, but it does vector operations very well with the functional unit approach used by Cray and a few others as opposed to having "explicit" vector processing boards. We have benchmarked the machine against the Ardent Titan and have found them to be equal in "vector" mode and the DN 10000 is about twice as fast in scalar mode. In regard to the graphics, have you seen Apollo's literature on the DN 10000VS series? This configuration is available with either 40 or 80 planes of graphics memory, and achieves 1.1 million 3d vectors/sec as opposed to the 300,000/sec claimed by Ardent. The updated versions of GMR, GSR, and GPR are still in beta, but this is going to be a serious 3d rendering machine (that's why we are upgrading our existing DN 10000). I don't know much about image processing in and of itself, but I have heard that several vendors were developing "ip" boards which plug directly into the 10000's X-bus (150 MB/sec bandwidth). So I think you could accomplish something with that. I suggest you talk in more detail with your local SE. I'll let you play with our machine if you are in Houston rather than Linden. Sincerely, Andrew Wescott University of Houston Department of Chemical Engineering