More
benchmarks
with
explanations, where possible
| Quake 2: using default settings, 640x480, OpenGL |
 |
|
In Demo1, the Rage 128 hit the wall badly, so I added the crusher test scores, where the machines scale more evenly, or namely scale at all. As with most 3D games, bus and memory bandwidth really matter a lot here. Interesting to see that the G4 450/466 is always faster than a G3 500, even without Altivec optimisation. (well, apart from OpenGL). Even hit the 60 FPS barrier with the Voodoo3 and the G4 533. Nice. |
| Quake 3 1.27h: using demo001, 640x480, 16bit |
 |
|
While
in normal both Macs seem to be struggling with the code, the B/W
pulls off quite a bit once the overhead is removed (using my settings
file: download here ) Again,
the G4 450/466 is simply faster than a G3 500.
|
| Unreal Tournament: cityintro, 640x480, medium, 45FPS desired |
 |
|
Note
the huge difference between Glide (Voodoo3) and RAVE mode (Rage
128), I didn't bother to test with the 550 MHz overclock, its obvious
the card is hitting a wall. I guess it would be around 35.5 FPS,
not more.
|
| Unreal: castle flyby, ultra high quality, 640x480 |
 |
|
Showing
minimum, average and maximum FPS scores. I really don't know what
I did wrong with the Rage 128 (RAVE), or if it just couldn't put
up with curved surfaces etc, this is really strange. The Voodoo
(Glide) shows nicely how you can still get a lot out of those Beige
G3s.
|
| G4Timedemo: Amazing quality |
 |
|
This
also shows how important Altivec can be. If Quake 3 made such a
jump once it's got that altivec enhancement, we'd surely all rejoice:)
Couldn't test this on the Voodoo, it doesn't work.
|
| The old classic: MacBench 5.0 |
 |
|
This
benchmark definitely needs an update to take advantage of Altivec.
It shows nicely how much better the G4 FPU is compared to the G3
FPU, though. Allow a variation of about 10-25 points.
|