QuakePPC, Warp3D and more..

[rvo_nl]

I really want to see how fast I can get Quake (yes, 1!) to run on my 'new' PPC setup. I used to get around 20fps using QuakePPC5 back in the days, on my 166mhz 603e on 320x200. But now there are so many versions and compiles of Quake for Amiga PPC, I really needed to make a quick overview to catch up with them:
 
1) BlitzQuake (v1.0, Aminet, Christian Michael)
-architecture: 68k, Warp3D
320x200 56.3 fps
800x600 8.3 fps
-architecture: WarpOS, Warp3D
320x200 fps (WarpOS exception, doesnt work with BVision? Then why does 68K+Warp3D work?)
 
2) GL/QuakeWOS (v2.3, Aminet, Frank Wille & Steffen Häuser (& Massimiliano Tretene?))
-architecture: WarpOS, CGX
320x240 21.9 fps (CGX)
800x600 5.8 fps (CGX)
-architecture: WarpOS, Warp3D
320x240 19.4 fps (CGX)
800x600 fps (CGX) (can not change screenmode)
 
3) QuakePPC (v5.0, (aka FastQuake 10), unknown author (who knows?), 'illegal' Aminet release)
architecture: PowerUP, CGX
320x200 23.15 fps
800x600 4.87 fps
 
4) WarpQuake (v0.7, Aminet, Jarmo Laakkonen)
-architecture: WarpOS, CGX
320x200 22.3 fps
800x600 5.4 fps
 
5) AwinQuake (0.9, Aminet, Peter McGavin)
-architecture: 68K, CGX
320x200 ~5fps (only realtime fps counter)
-architecture: WarpOS, CGX
320x200 ~22 fps
800x600 ~5 fps
-architecture: PowerUP, CGX
320x200 fps (crash)
 
6) 'Legacy' releases
-version by Clickboom
-the orginal 'hack' (would be great if someone could send this to me)
 
I was a bit dissapointed by the results. BlitzQuake is most promising but wont work on (my?) BVision and leaves out lots of particle and lighting effects to gain speed. Funny though that the 68K+Warp3D version has excellent performance. GLQuakeWOS is prefered by many but I cant change to a higher resolution and still I think its not as fast as it could be. Also, there's lots of weird rendering bugs like partly transparent walls and such.
 
When I look around on youtube and on threads on this site, I get the feeling my system is not performing as well as it should. This is also apparent when I run warp3D/MESA/Glut demo's, if this is in any way related. I've fiddled with the different tooltypes of the Bvision monitor and CyberGraphX, but that didnt help. For example, the demo 'skyfly' does not run smooth at all in 640x480.
 
I am using:
 
WarpUP 5.1
Warp3D 4.0
CGX4.3 rc5
the 'correct' 040/060 libraries
latest flashupdate
bppcfix if possible
 
1) is the above overview correct? Any more releases?
2) are these results 'okay' for my hardware config (see below) ?
3) any things I could try to increase performance?
 
BTW, I know I'm a bit crazy trying to get this ancient game to run on my ancient Amiga, no need to remind me..

[rvo_nl]

great ideas, tips and results. especially happy with that link from zipper. will try out some stuff over the weekend.

[rvo_nl]

Okay, I did some test today. But first: when I look at the results on amigaspeed.de.vu I am noticing a funny thing:
 
Quake 1 running in 800x600 gives the following results:
 
Michael Lanser, 160mhz ppc, 040 + bvision: 10.6fps
Torsten Hees, 240mhz, 060 + bvision: 8.3fps
 
Can someone explain this to me? I get 8.4fps using the same test, so I really need to know what Michael Lanser did to get it to run this fast. It explains also why all ppc/warp3d apps run a bit sluggish on my setup. There must be something wrong..

[rvo_nl]

small update so far: Ive noted quite some speed difference between warp3d 3.x and warp3d 4.0. And with that I mean the old version is much faster. Ive tested this using the skyflyppc demo.

[rvo_nl]

My question is probably even more silly, but since I've been out of this 'scene' for so long, does Quake 3 run on classic Amiga's running OS4? If not, why not? I dont mind the fps, I just like getting stuff like this to run and probably wont play it at all.

[rvo_nl]

Finally, I've got everything sorted now! The only problem is that I dont really know what I did to fix some of the issues.. The MESA/GLUT performance problem was fixed by installing WOSprefs and fiddling with the parameters. (especially the one labelled FAST, which I remember from way back). The Skyfly demo is now finally moving at a steady pace, same goes for the other GLUT/MESA stuff.
 
However, I was able to increase performance a lot more by installing the Warp3D 3.1 libraries over the 4.0 libraries. Apparently Warp3D 3.1 is a bit faster than 4.0. (and 4.2 aswell, but that one is bugged, too!)
 
So, with all the tweaks I did I was really eager to give glquakewos another try. Much to my surprise it was a lot faster than I expected. I was afraid I wouldnt get anywhere near the results by this Michael Lanser guy, who is using a 160mhz ppc. But fortunately I did and am getting near Voodoo 3 speed just using my Bvision. (I guess those with a Voodoo 3 have some configuring to do, Im not fooling myself).
 
Take a look at: http://www.alinea-computer.de/amigaspeed/q***e.html
 
320x240:17.6fps
My result: 25.5fps
 
800x600:10.6fps
My result: 12fps
 
So, not a lot, but still it shows the difference between 160mhz and 240mhz. With the overclock tooltypes activated it went a bit faster, but no more than 1 or 2 fps.
 
I still hope someone can send me the original Quake 'hack', aswell as some more info on the authors of certain QuakePPC ports. And can anyone answer my previous question regarding Quake 3? Google doesnt help me at all

[rvo_nl]

I've noticed those black rectangles around light sources using blitzquake aswell. Thought it was a way to increase speed, doh! Thanks for the info, now I know what to look for.
 
@ Piru: you dont have to tell me his name.. But nationality would be a start! Is this the same guy that was working on Quake 2 before Quake 1 was even released on Amiga? And yeah, I meant Q3 for classic. Would be nice to try it out.

[rvo_nl]

However, I was able to increase performance a lot more by installing the Warp3D 3.1 libraries over the 4.0 libraries. Apparently Warp3D 3.1 is a bit faster than 4.0. (and 4.2 aswell, but that one is bugged, too!)

No, no, NO! That is not true. It's just QuakeWOS that gives unpredictable results. Stick with 4.0 and find a better Warp3D application to do benchmarks.
 
(This was just to keep the information in this thread correct..