The
benchmarks, charts, scores etc
with
explanations, where possible
| Altivec Fractal IP tests: |
 |
|
If here the advantage of the Altivec units (Velocity engine) isn't clear, then I can't help you. |
| Gauge Pro RAM performance |
 |
|
This
shows quite clearly how much advantage a faster system Bus and better
RAM controller chip can give.
|
| Unstuff Test using Stuffit Expander 5.5 |
 |
|
This
test is a bit harder to interpret. It shows that the G4 still has
an advantage over the G3, even without Altivec, and it clearly shows
how important 2nd level cache speed is to some applications, just
look at the score of the G4 550 with only 220MHz cache speed. Also
the harddrive speed and the bus speed are important here.
|
| iTunes: convert to MP3 |
 |
|
Again,
the altivec optimisation gives iTunes wings on a G4. The system
bus also matters, as does HD speed. This test speaks for itself.
|
| SoundJAM MP: convert to MP3 (again) |
 |
|
Unfortunately
refused to run on the G4 550 overclock (crashed), but again we see
how great optimised apps could be. SoundJAM is not only faster than
iTunes, its also less disk dependent, at least it seems so.
|
| Photoshop 5.5: 21 filter test totals |
 |
|
Unlike
the individual test scores, the totals actually make sense and show
that Photoshop cares little about bus or HD speeds, as long as it
has enough RAM (200MB assigned minimum). Altivec doesn't show a
huge advantage here, which probably means that the filters used
aren't all very optimized. Why the 533 is slower than the 500 XLR8
remains a mistery. You can see the detailed scores for each
individual filter here.
|
| CineBench 2000 |
 |
|
This
only shows that rendering aparently is more fond of MHz than of
Altivec and that the Voodoo3 is not designed for professional use,
OpenGL shading doesn't work right. (thus not tested)
|
I have also made a table of scores from the TechTool Pro benchmarks, if you'd like to see these, check them out here.