Quake 3 117, Demo001

Mac OS 9.1 without Sound (s_initsound "0")

Quake 3 117, Demo001

[Update 06/16/2001: I corrected the Radeon 32bit scores. I accidentally wrote the sound ON scores in there first. -Oliver]

Mac OS X Final (s_initsound "0")

Quake 3 117 (port by Omnigroup), Demo001

Update - Quake3 Pointrelease 1.29f for Mac OS X and 9

Notes:
- All tests below with GeForce3 only
- Resolution switching is slower now, it's kind of the same as it in Mac OS 9 (Omnigroups build was faster)
- The q3config.file is no longer in a hidden directory ~/.q3a/baseq3/ – it's in ~/Library/Application Support/Quake3/baseq3/
- r_smp is ENABLED by default (r_smp 1)
- Physics!!! Circlejumps (for example the 25-health to 25-health jump in the "core" of q3tourney4) are now possible on your own Mac-listen server – I've never been able to do it on my own server until now, only when I was connected to a Wintel PC or Linux Box. I love that!
- With the new network code comes a new demo-format: there's a new demo called four.dm_66, which is the one i tested below.
- Update 06/22/2001: I replaced the scores with newer ones from the second release of Quake3 1.29f for Mac OS X, from now on called 1.29f-2.
- Update 06/22/2001: Only sound ON tests were repeated as sound OFF results were pretty much the same.

As you see, sound is extremely "processor-intensive", with sound disabled I get more than twice the framerates (with r_smp enabled and at resolutions up to 1024x768).
That's why I changed the s_mixahead value (which basically says how far in time should sound-prediction look ahead and preload sounds in ram – or something like this). Its default value is 0.2. "s_mixahead 0.03" is the lowest value at which sounds were playing normally (at 640x480!). Going to 0.02 or below produced stuttering or distorted sound. At higher resolutions or with higher quality settings the value needs to be higher. The lowest possible value is different for each setup, so try it out for yours.
Values between parentheses mean that sound was stuttering there, so s_mixahead should have been higher.

r_fastsky had to be changed to 0 because otherwise the picture was flickering wildly. r_finish was changed to 0 so the fps could go above my screen's refresh rate (which is 160hz at 640x480, i believe – so i'm not sure what that 174fps value is doing up there). And finally cg_draw3dicons 1 has proven to be quite a speed boost when smp was enabled (see above).

When I saw the basic-locki-scores I was pretty disappointed. Only 14fps faster than with default config?! Anyway, with modifying the s_mixahead value one can squeeze almost 3 times more fps out of Quake3. Again, values between parentheses mean that sound was stuttering there, so s_mixahead should have been higher.
Update 06/22/2001: The scores with basic-locki are not as bad with 1.29f-2.
The graph below shows only the values with the modified config – not with the basic locki.

Now finally it leaves me wondering why the 1.29f beta pointrelease has so much of an fps hit if sound is enabled when Omnigroup's build seemed not to have one at all...
Update 06/22/2001: The slow sound issue was much improved in 1.29f-2.

Mac OS 9.1 – Q3 129f beta:

All Tests below were done with 32bit color & textures(r_colorbits 32 and r_texturebits 32) only.

At 32bit color and textures i was not able to switch resolutions to 1280x1024 or 1600x1200. If I selected to change resolution to 1280x1024 or 1600x1200, it would do exactly as it would when really switching resolution, only that afterward the resolution is still at 1024x768 (easily visible because of the size of the console-text). I had the same problem before back when I was using a Geforce2MX and Q3 1.27h. Changing to 16bit usually enabled me to switch to resolutions higher than 1024x768.