Path: chuka.playstation.co.uk!news From: "del :)" Newsgroups: scee.yaroze.freetalk.english Subject: Re: Real PS speeds.... Date: 7 Sep 1997 21:06:48 GMT Organization: PlayStation Net Yaroze (SCEE) Lines: 70 Message-ID: <01bcbbd2$786d8160$da7f989e@nlights.demon.co.uk> References: <5tk1rp$3qt7@chuka.playstation.co.uk> <33FD5C48.27BF@cobradev.demon.co.uk> <33FED194.5D74@interactive.sony.com> <3403052F.1365@interactive.sony.com> NNTP-Posting-Host: nlights.demon.co.uk X-Newsreader: Microsoft Internet News 4.70.1155 Developer Support wrote in article <3403052F.1365@interactive.sony.com>... > Developer Support wrote: > > > > I thought I'd better correct some fallacies about the 'speed' difference > > between Yaroze and full PS libs. > > > > The 360k polys per second ( pps ) is a marketing number that represents > > the limit of the gpu hardware when drawing a single flat poly repeatedly > > ( no gourard / no texture ) - As such it is pretty unrealistic for > > programming purposes ( although it is strictly correct ) > > > > running several games through a proffessional program analyser the > > 'real' pps count ranged as below. ( I only did this because I was bored, > > and I had 15 minutes to spare - So I wont be producing a hughe list ) > > > > ( Note: These numbers don't represent maximums, just what was happening > > at that particular stage/point in the game ) > > > > Porsche Challenge 30,000 pps. > > Penny Racers. 64,000 pps. > > Soul Edge. 53,000 pps. > > Rage Racer. 45,000 pps. > > Terra Incognita(yaroze) 28,000 pps. > > Snave (yaroze) 40,000 pps. > > > > These represent the drawing throughput for these games. ( In all cases, > > the drawing time is less than the cpu time ) > > Although Penny racers seems to be drawing more pps, it actually has the > > slowest frame rate of all of the games. > > > > In my estimation the maximum speed up going from good yaroze code ( that > > uses all of the tricks possible with LibPs ) to good proffesional code > > (inline assembly and low level hardware access) is around 2 times. > > If you look at the differences between FTP test using the Yaroze, and > > RayStorm ( from Taito ) using the full devkits, there isn't really much > > technical improvement. > > > > Cheers, > > > > Colin. > > > > A quick addendum to this.... > > On the Yaroze, I tested two demo's this morning.. > > Lewis's tunnel demo ( May not be up yet ) 75,000 pps > Del's Engine 6 115,000 pps. > > Lewis's demo shows the limit ( probally ) of Yaroze at the moment, > while Del ( naughty ) wrote his demo using the full low level libs... > Del's is very quick because most of the poly's are untextured and > unshaded, - most games aren't written this way. > I have now optimized the engine6 demo to the max! I reckon I`m cranking up about 175,000 pps now.... hehe I`m still putting this demo into the yaroze libs so I can spread the source code, I've just got a few problems at the moment (converting my meshes to tmd etc), anyway lookout for the extra hot engine7 demo soon, maybe i`ll have the yaroze source too... :)