the results you most want to see. So it
really comes down to which benchmark-
ing approach you most respect, ours or
MacAddict’s. You have the information,
reader. It’s on page 52. You make the call.
As far as we’re concerned, PCs take
the Photoshop throne. We can say the
same for InDesign 2.0 PDF conversion,
in which the AMD and Intel chips were
23 and 25 percent faster than the G5,
respectively.
In the Bibble/MacBibble RAW conver-
sion tests, the G5 nailed an unambigu-
ous victory, finishing its conversion
roughly 47 percent faster then the
Athlon 64 FX and 32 percent faster then
the P4EE. As Keanu would say: “Whoa.”
So how does the G5 go from being the
dunce in SETI@Home and Mathematica
to the head of the class here? We have a
few theories. The first is that the Bibble
PC code may not have received enough
developer attention. Is it possible that
Altivec instructions are supported in
MacBibble , but no similar SIMD instruc-
tions (that is, SSE2) are supported in
Bibble? And if SSE2 is supported by the
app, is it possible that it’s not properly
detected by the Athlon 64 FX, whichshowed the poorest performance?
The other factor likely at play is the
multiprocessor support— Bibble and
MacBibble both support multiproces-
sor machines, such as the G5. So let’s
take an inventory. Our AMD box has
only once processor. Our Intel box
also has only one processor, but that
processor supports Hyper-Threading,
which provides multiproc capabili-
ties on a cycle-by-cycle basis. As for
our Apple box, it sports two physical
processors—thus it has a decided
advantage in the MacBibble test.
The upshot: Apple would do well to
encourage all developers to code for
multithreading, because it clearly
makes a difference.
As for our final test—DV-to-MPEG2
conversion using two different
apps—we’re finding it difficult to
give much credence to the results.
As you can see, the two PCs blew
by the G5 in the tests that were set
up for lower-quality MPEG2 encoding,
and in high-quality mode, the tables
turned. But, again, we c
the obvious: If the high-
ProCoder cranks out becompressed video than the high-quality
mode in Compressorr , then the PCs areAND NOW (^) FOR THE
MINORITY OPINION...
We asked Rik Myslewski,
MacAddict’s editor-in-
chief, to weigh in on our
testing and analysis.
While we here at
controller and the Pentium
4 Extreme
Edition’s 2MB L3 cache, and w
hile we also
grudgingly concede the lead to you PC
folks in game performance,
we’re pleased
as punch that it took two exp
ensive, just-
released, specialized procs (una
vailable
to normal consumers searching
the web
sites of Dell, Gateway, HP, and othe
edge rs) to^
out a standard, off-the-shelf
2GHz dual-
Power Mac G5. In tests that mim
work ic
performed on non-hot-rod PCs by
normal working shlubs like,
oh, about 99
percent of the personal-com
puter market,
the G5 (née IBM PowerPC 970)
rocks. We
look forward to testing the G5 ag
ainst the
garden-variety P4s that litter the com
ing land put-
scape.
After looking at the official test results that
appear on page 52, and after considering
the overall performance of all three plat-
forms, we have to declare Intel’s Pentium
4 Extreme Edition the winner by the thin-
nest of margins. It looks like the P4EE’s
large L3 cache is giving the basic P4 core
a big performance boost in not only mul-
timedia apps but also games. The P4EE’s
Hyper-Threading is also showing more of
its value, and is responsible for some of the
performance wins you see in this article.
The Athlon 64 FX-51, however, is so
damned close, we’re reluctant to call it
a “second-best” CPU. Yes, it won fewer
benchmark rounds in this competition,
but its overall performance was just a hair
behind the P4EE’s, and in some “unof-
ficial” tests that we ran on just the Intel
and AMD machines, the FX-51 showed
superiority in games. For example, in
Halo , the FX-51 beat the P4EE by about
six frames (62.4fps to 56.5fps). It also tri-
umphed in the UT2003 Botmatch demo
(96.9fps to 93.8fps). We think you can
ler for these wins.
support. In apps that aren’t optimized
for the G5’s unique benefits, the Mac’s
performance is a travesty. We’d even bet
that in some apps, the G5 is no faster
same fate.
Regardless, we can finally say that
Apple has introduced a worthy competitor
to the PC platform. n
And the Winner Is...
DECEMBER 2003 MAXIMUM PC 55