5 MAXIMUMPC DECEMBER 2006
Putting Quad to the Test
Our benchmarks reveal some interesting things about multicore computing
T
o test the new Core 2 Extreme QX6800,
we selected a raft of real-world CPU-
intensive tasks—video editing, image editing,
and video encoding—as well as a handful of
games and graphics tests at low resolutions
to isolate CPU performance from GPU per-
formance as much as possible. We paired the
proc with a new Intel D975XBX2 motherboard
with a beta BIOS and drivers so we could com-
pare it with the Core 2 Extreme X6800 in an
older Intel D975XBX motherboard. Both sys-
tems were equipped with a GeForce 7900 GTX
card, a 400GB Western Digital SATA hard drive,
and a PC Power and Cooling 1KW power sup-
ply, as well as the same graphics drivers run-
ning under Windows XP Professional with SP2
installed. We excluded AMD’s Athlon 64 FX-62
from this face-off because, well, why bother?
It’s just not in the same league.
As we alluded to earlier, the benchmarks
we had moved to for their multithreaded
performance aren’t multithreaded enough
for quad. From Premiere Pro 2.0 to Nero
Recode 2 to Bibble to Noise Ninja, we didn’t
see the quantum leap forward with QX6800.Sure, when the two procs were tasked with
converting 2.7GB of digital RAW fi les to JPEG
using Bibble and then removing noise from
the images with Noise Ninja, the QX6800 was
about 19 percent faster. Pretty good but not
quite the boost we expected from two extra
cores inside.
Despite the QX6800’s 10 percent slower
clock speed, it was as fast as or a bit faster
than the X6800 in a majority of the applica-
tion benchmarks. It gets even better for the
QX6800 with apps that are more multithread-
ed. Intel has been pimping Sony’s new Vegas
7.0 for reviewers, and we see why. During
encoding, the app works dual and quad cores
to their full potential. We also saw signifi cant
performance benefi ts when we ran the mul-
tithreaded Quake 4 while encoding an H.264
video in the background.
In straight gaming, however, quad core
isn’t as strong. As we mentioned previously,
the higher-clocked dual-core X6800 is faster in
today’s games, which are mostly single-thread-
ed. And even Quake 4, which is multithreaded,
favors the X6800 because apparently the gameis optimized for two cores, no more.
As a fi nal test, we overclocked our QX6800
to the same speeds as the X6800 and saw the
quad core close the gap in the benchmarks it
lost due to the clock defi cit.THE UPSHOT
The quad core isn’t the decisive winner in the
benchmarks because of the 300MHz clock
disadvantage. People who can only see into
next week will probably go for the dual core,
but we think that would be a mistake. If you’re
willing to perform a minor overclock of 10 per-
cent, you will all but erase the dual core’s
advantage. Even if you don’t, the quad core will
end up being faster once developers optimize
for it, while the dual-core’s performance poten-
tial won’t change. We liken it to the choice peo-
ple made between single core and dual core.
We suspect that many folks who bought sin-
gle-core CPUs are wishing they now had dual-
core procs. You’ll think the same about quad
core in 12 to 18 months. With QX6800 costing
the same as X6800, it isn’t a very hard choice
to make.Benchmarks
Bold denotes winner. Red scores are overclocked and not factored into overall win. ??? indicates “WTF?!”PERFORMANCE DELTA C2E QX6800
C2E QX6800 C2E X6800 BETWEEN QX6800 AND X6800 OVERCLOCKED
CLOCK 2.66GHz 2.93GHz 2.93GHz
APPLICATIONS
VEGAS 7.0 (SEC) 252 378 50% 238
PREMIERE PRO WMV9 (SEC) 1,848 1,835 -0.7% 1,680
PREMIERE PRO WMV9 1,088 1,291 18.7% 1,040
ADVANCED PROFILE (SEC)
RECODE 2 IPOD H.264 (SEC) 1,835 2,033 10.8% 1,832
RECODE 2 MPEG2 (SEC) 465 589 26.7% 440
BIBBLE 4.9 RAW CONVERSION 1,450 1,725 19.0% 2,431???
WITH NOISE NINJA (SEC)
PHOTOSHOP CS2 (SEC) 219 211 -3.7% 206
NOISE NINJA 2.11 (SEC) 1,457 1,604 10.1% 1,403
SYNTHETIC TEST
SCIENCEMARK 2.0 1,478.9 1,611.6 -8.2% 1,597.4
PCMARK05 7,590 7,861 -3.5% 7,638
PCMARK05 CPU 8,452 7,470 13.6% 8,455
3DMARK06 CPU 3,897 2,522 54.5% 2,585
3DMARK06 6,727 6,370 5.6% 6,828
GAMING
3DMARK05 CPU 11,320 11,407 -0.8% 11,148
3 DMARK05 10,711 11,277 -5% 11,007
QUAKE 4 (FPS) 173.3 195.0 -11.1% 198
FEAR 1.07 (FPS) 258 268 -3.7% 260
MULTITASKING
QUAKE 4 WHILE RUNNING AN AVC 97 145 -33.1% 162
RENDER IN THE BACKGROUND (FPS)Quad Power!Quad Power!Quad Power!