This comparative is interesting:
- As for CPU test, G5 is definitely inferior to the G4 for some tests such as Altivec Basic, but it regains advantage for floating point calculations.
Yer in the overall, G5 doesn't frag the G4 at all.
- On the other hand, memory tests show that DDR is now fully addressed by G5s, which wasn't the case with G4s. The G5 bandwidth is roughly double, except for the "fill" test where it is quite surprisingly divided by 2.
- As expected Radeon 9800 Pro is faster than Geforce Ti 4600, but only by 25%.
- The disk test looks disappointing, as Serial ATA seems to be of no impact.
- The User interface test, which we consider as the most important, reveals a 15% gain only. It allows us to have an idea of the G5 overall reactivity under OS X.
One has to be careful though, as Xbench isn't really reknowned for its reliability with results and their reproduction. But here's the first conclusion one could draw from this first test : the owners of dual 1.42 G4s shouldn't feel frustrated by the G5 arrival !
In our views the G5 will show its true potential only with Panther (OS X.3).
[Upd] Here's Fabrice interesting comment:
As I have quite a thorough experience with AltiVec, the results given by XBench do not surprise me much. The G5 units aren't as high-performant as G4 7455, at least in bold results. I suspect XBench test uses L1 cache (on the G4), which would explain the performance rating. The G5 probably works with its memory in regard to this test, which would explain its poor performances. Memory performances in AltiVec streaming on the other hand seem to augurate of the overall performances of a G5 as compared to a G4 when working with memory in real everyday use... So. Don't take this test too seriously. :) |
