4200 MAMAXIMXIMUUMMPPCC FEBRUARY 2007JANUARY 2007
Quad FX FilesQuad FX Files
B
oth Intel and AMD scoff at the use
of single-threaded applications to
test quad solutions, but in doing so,
they’re ignoring the obvious: The
majority of today’s games and consumer
apps are single-threaded. To disregard
a machine’s performance with existing
software is shortsighted at best. To give
you an idea of how the quads stack up,
we ran both single- and multithreaded
tasks. Some of the benchmarks are able
to exercise the quad cores properly:
Valve’s particle test and map-compil-
ing benchmark, for example, work all
four cores, as do the workstation apps
Newtek’s Lightwave 9 and Maxon’s
Cinebench. Bibble Labs’ Bibble Pro run-
ning simultaneously with PictureCode’s
Noise Ninja also works over all four cores
pretty well. Others, such as Premiere Pro
2.0, Quake 4, and Nero Recode 2.0, are
optimized for dual cores but don’t really
stress all four CPUs. As a fi nal test, wealso tossed in some “mega-tasking”—a
term AMD coined for the Quad FX—in
which we ran two multithreaded appli-
cations simultaneously.THE FINAL ANALYSIS
Fire up the Quad FX and your heart
warms with the sound of all that hardware
spooling up. If you’re of the belief that
noise equals power, then the sound of
the Quad FX rig’s two CPU fans and two
chipset heatsink fans certainly connotes
muscle—that is, until you see the bench-
mark numbers. While the pair of dual-
core 3GHz Athlon 64 FX-74s are indeed
fast, the single 2.66GHz Core 2 Extreme
QX6700 eats their lunch in the major-
ity of our tests. From single-threaded
games and applications to multithreaded
programs to even mega-tasking, Intel’s
QX6700 pulled out ahead, and generally
by a sizeable 10- to 33-percent margin.
At face value, that amounts to adefeat for AMD, especially when you con-
sider the physical aspects: Two Athlon 64
FX-74s are slower than one Intel Core 2
Extreme QX6700? That’s just not a good
message to send.
Still, we have to admit as power
users that AMD is on to something.
Just as we’re drawn to the concept of
two videocards or two hard drives, two
CPUs has sex appeal to us. Of course,
if we had our druthers, we’d run two
Core 2 QX6700 CPUs, not two Athlon
64 FX-74s.
What the Quad FX is really about,
though, is faith. Faith that AMD will have a
competitive quad-core CPU next summer.
Faith that AMD can somehow integrate
an ATI graphics chip into the platform.
And faith that, well, AMD has the atten-
tion span to continue the platform even if
it’s not a huge success. You have to ask
yourself whether you’d rather have faith or
a faster system right now.Benchmarking the Quad FX
Best scores are bolded. Both systems were outfitted with two GeForce 7900 GTX cards running in SLI, 2GB of Corsair DDR2/800, a Western Digital 400GB 7,200rpm hard drive, and Windows XP Pro with Service Pack 2 installed.Benchmarks
3GHZ AMD ATHLON 2.66GHZ INTEL CORE 2
64 FX-74 (PAIR) EXTREME QX6700 PERCENT DIFFERENCE
SYNTHETIC GAMING
3DMARK06 OVERALL 10,293 11,268 -8.7%
3DMARK06 CPU 3,678 3,895 -5.6%
3DMARK05 CPU 8,390 8,964 -6.4%
VALVE PARTICLE TEST 73 83 -12.1%
GAMING
FEAR AT LOW RESOLUTION (FPS) 180 198 -9.1%
QUAKE 4 (FPS) 153 176.3 -13.2%
COMPANY OF HEROES AT LOW RESOLUTION (FPS) 82 83 -1.2%
WORKSTATION
SCIENCEMARK 2.0 1,507 1,496 0.7%
PREMIERE PRO 2.0 HDV (SEC) 1,935 1,454 -24.9%
NOISE NINJA (SEC) 1,783 1,831 2.7%
NEWTEK LIGHTWAVE 9.0 (SEC) 28.6 26 -9.1%
CINEBENCH 9.5 (SEC) 1,361 1,412 3.8%
VALVE MAP COMPILE TEST (SEC) 182 160 -12.1%
BIBBLE & NOISE NINJA (SEC) 2,283 1,830 -19.9%
ENCODING
NERO AVC ENCODE (SEC) 1,837 1,896 3.2%
MULTITASKING
QUAKE 4 W/VALVE MAP BUILD (FPS) 121 145 -16.6%
VALVE MAP BUILD W/QUAKE 4 (SEC) 228 191 -16.2%
PREMIERE PRO W/NOISE NINJA (SEC) 2,979 2,845 -4.5%
NOISE NINJA W/PREMIERE PRO (SEC) 2,165 1,974 -8.8%