Message-ID: <34F9E0CF.FA9802A3@gbrmpa.gov.au> Date: Mon, 02 Mar 1998 08:27:27 +1000 From: Leath Muller Organization: Great Barrier Reef Marine Park Authority MIME-Version: 1.0 To: James W Sager Iii CC: djgpp AT delorie DOT com Subject: Re: The future of graphics programming References: <6cvn02$6t5 AT netnews DOT hinet DOT net> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Precedence: bulk Pet peeve time... > Um, 20,000 polygons per frame? > with a minimal speed of 24 fps, this is 400,000 polygons per sec. The latest Voodoo II architecture (yes, HW accelerated) does 3 million poly's a second - easily... so the figure of 400,000 is not high. I have a demo here which runs on a 3Dfx Voodoo 1 card at 1.16 Million poly's per second - and the CPU does all the transforms. > A fast pentium processeor does about 2-4 mil instructions per second. Ummm... no... :) It depends completely on what your doing, but for basic integer work which is properly pipelined you can get well in excess of 600 MIPS on P2's - you can get 300 MIPS on a crappy P166. (Even though MIPS are the worst presentation of performance anywhere) Quake performs perspective correct texture mapping every 8 pixels with affine mapping in the sub-spans to run at about 7.5 cycles per pixel. I get about 34 FPS on my P166 which is theoretically impossible by your figures. > So like if you could put a polygon down 1/10 as fast as you could do a basic > additoion problem this is feasable. But with with current hardware > constraints, we can do maybe 1/10000 as fast... 3 orders of magnitude. > I don't forsee computers becoming that much more fast in the next couple > years. I mean I'd like faster software than allegro. I mean, who doesn't > want faster software, the simple fact is that no one has done it, so I > just work with what I got. Your must be working on a 386 @ 8 Mhz... ;) Leathal.