Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wuarchive!uunet!mcsun!ukc!canon!laukee From: laukee@canon.co.uk (David Lau-Kee) Newsgroups: comp.graphics Subject: Re: Rendering performance Message-ID: <1991May6.114747.12681@canon.co.uk> Date: 6 May 91 11:47:47 GMT References: <1991Apr29.162712.1905@canon.co.uk> <1991May3.190648.13574@ucunix.san.uc.edu> <1991May3.224900.12807@dsd.es.com> Reply-To: laukee@canon.co.uk Organization: Canon Research Europe, Guildford, UK Lines: 55 rthomson@mesa.dsd.es.com (Rich Thomson) writes: >In article <1991May3.190648.13574@ucunix.san.uc.edu> > reher@ucunix.san.uc.edu (Jay R. Reher) writes: >> Silicon Graphics has an adapter that is supposed to display 3 million >>Gourad shaded polygons per second. >I know the VGX can do a million FLAT shaded polygons (triangle meshes >with 50 pixels per triangle, actually), but I've never seen anything >from them claiming 3 million "polygons" [have to be careful to say >exactly how big and what kind of polygon], especially not Gouraud >shaded polygons. >However, one of the parallel architecture machines (I can't remember >if its pixel planes or AT&T pixel machine -- I always get them >confused) does claim to be able to do 3 M Phong (note: not Gouraud) >shaded polygons. This is the point isn't it, "your polygons aren't the same as my polygons, and even if they were, would they do what I want them to?" The VGX does 1.1 million *GOURAUD* shaded, *PHONG* lighted, Z-buffered, 100 pixel, 3-sided polgons (triangle meshes) per second. But how useful is this? If you do some benchmarking on typical interactive manipulation of smallish (1k-2k) 3-D models you get around 35-60k polygons / second on a VGX. [Don't quote me, this is from memory.] With some new, highly "don't ask me any more, I've been issued with a cyanide capsule" secret techniques (and, errr... one or two caveats), we're getting around half this rate (though only FLAT shading) on a SparcStation 2. But this is *entirely in software* (no hardware support whatsoever), so the code will run happily on the Mac, the NeXT, high-end PCs, etc, limited only by the speed of the CPU and how fast it can move stuff to the framebuffer. If you want "near photorealistic" effects (like complex surface characteristics and "atmospheric effects") in real time then the VGX (and its ilk) are excellent machines. But if you want interactive 3-D applications for the masses then you have to realise that most people aren't going to want to fork out gobs of dosh for high-end, graphics specific hardware if they can get 90% of what they want as software-only solutions on "general purpose", industry standard, heh, workstations, for 10% of the price. The P*x*l Machine is the perfect example of how these huge "how fast you can do polygons" numbers don't always reflect how people want to use the machine. Is anyone actually using other than a "near top-end" p*x*l machine for interactive 3-D work? Would they like to quote their "typical" frame rates for the sort of work they are doing? ...And how much did you say you paid for that? ------------- David Lau-Kee, Canon Research Centre Europe, 17/20 Frederick Sanger Rd, Surrey Research Park, Guildford, Surrey, GU25YD, UK. laukee@canon.co.uk, laukee@canon.uucp, ..!mcsun!ukc!uos-ee!canon!laukee Tel: +44 (0) 483 574325 Fax: +44 (0) 483 574360