Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!mit-eddie!uw-beaver!cornell!batcomputer!andyrose From: andyrose@batcomputer.tn.cornell.edu (Andy Rose) Newsgroups: comp.graphics.visualization Subject: Re: Biomedical field? Message-ID: <1990Nov7.193654.7665@batcomputer.tn.cornell.edu> Date: 7 Nov 90 19:36:54 GMT References: <1470@vtsai2.sai.vtt.fi> Organization: Cornell Theory Center Lines: 98 With the news that Tektronix will no longer play a role in high end workstations a number of issues come to mind. If a seemingly strong commitment to graphics, if judged by the size of Tek's booth at SIGGRAPH, can vanish, where do graphics people put their "beans"? IBM (RS6000), Silicon Graphics, Stardent,souped up apples? I would be interested in hearing how buyers/developers deal with the changing market. Let's consider a few points of view... 1) Big Co. wants to set up a visualization lab. Let's say Big owns a few Convexen or some Crays. They are producing enormous amount of, let's say, pressure and temperature data in a three D irregular mesh of points using some boundary solving thingy. They have been using Big's inhouse cad system to build the meshes and design the objects. a. Big does not wish to develop a product for sale, rather it wants to visualize the data so that the scientists can better understand there science (Big is a big believer in sci vis). In these tight times, Big doesn't mind kicking down $$$ but wants a system to last into the future. Performancewise, Big wants to see its models rotate in realtime and see the colors change in realtime. Big also wants to interact with the science as it displays the graphics. True sci viz. Big definitely needs a good deal of processing power (tens of thousands of polygons shaded a second). Big can a. Buy a Stardent ($60-150K) and bang out graphics with AVS in a few weeks. b. Buy Silicon Graphics (same price), write custom code (assuming programming talent), and be groovy in a few months. c. Buy silicon graphics and some software (Wavefront Data Visualizer?) and maybe be happy in less than a few months (may also assume programming talent). d. Lease hardware (avoiding inevitable obselescence) and rent talent. What is Big to do? Can he be secure in his choice of SGI or Stardent, or should he go BIG BLUE and relax (???). Can he find programming talent? What software can he get off-the-shelf which will do this stuff? What about output? Does he need to record the animation? Digital or analog video? Can he get this kind of service out-of-house? 2. Small Co. develops graphics software. Small got in the game a few years ago writing plotting packages and supporting Telegraf for some Fortune 500s. Small just hired talent and wants to break into "scientific visualization." Small knows Big needs a package to display 3D models and data interactively with feedback to the science program (i.e. widgets). Small knows about AVS, apE, Khoros, Serpent, IKP, UIMX, Motif, X, etc... Small hates reinventing the wheel. Can Small add to an existing software product (AVS, apE) and repackage it along with support, training, doc, etc. for resale to Big? Small needs a machine, which one? Right now I don't think either of these guys can build a system which will sail them into the 21st century. It is painfully clear how quickly this stuff is changing (Tektronix XD88). SO. What is it about scientific visualization that is not changing? Besides all the inevitable goings ons in the hardware realm where is sci vis going? Since guys in suites are going to decide which platform to buy anyway, the software will be platform independent to protect the software vendors (ala Wavefront). Just how expensive is sci viz anyway? Rather than rambling on here are some figures. Dept. Vis. Cornell Theory Center serves supercomputer users with lots of data. In order to do this we rely on 2 full time animators cranking frames with Wavefront Tech Model/Preview/Image (Professional Visualizer) on Tek XD88 (2). Both are facile with c and write readers and other filters for each job. This setup is great for batch type animation but insufficient for interactive vis. For this we use a Stardent GS1000 ($140K 2 years ago) running AVS 2.0. This setup has successfully visualizaed 4,000 particles represented as spheres updated 2-4 times a second while the simulation ran (parameters effecting the science could be adjusted "on the fly"). This requires custom coding also to read science data and to design an interface for the scientist, however, with training the process of building AVS modules is on the order of weeks rather than months for building X or Motif interfaces. Color encoders allow output of Wavefront frames to an Abekas frame store for editing onto 3/4" video tape. Video setup costs $$$. Would like to convert the whole thing to digital soon. In many ways this is representative of setups at NCSA, SDSC, and other "production" facilites. Big, self-contained, labor intensive. As platforms become less expensive and everyone uses 24bit color and blasts 100,000 polygons/second, more mature software tools will be needed to take advantage of the resource. Video prices won't (read will but slower) come down so how does the small user output (one answer is shared resources or automated video dumping across the net)? enough for now... How do you vis? -- Andrew Newkirk Rose '91 Department of Visualization CNSF/Theory Center 632 E & T Building, Hoy Road Ithaca, NY 14583 607 254 8686 andy@cornellf.tn.cornell.edu