And the Winner Is...
benchmarks
Best scores are bolded.18 MAXIMUMPC january 2007
Testing Methods
C
omparing notebook CPUs is more complex than testing
desktop chips. You not only want to factor in how fast the CPU
is, but also how much power it eats. To get a direct comparison
between the CPUs, we got an HL-80 notebook from iBuyPower.
The machine came equipped with a 2.16GHz Core 2 Duo T7400, a
GeForce Go 7600, 2GB of DDR2/667, a 5,400rpm hard drive, and an
Intel 945PM chipset.
When we finished testing the Core 2 T7400, we cracked the
machine open, installed a Core Duo T2600, and reran the tests. The
bulk of the benchmarks we use target processor performance more
than graphics performance, but we did run some gaming benchmarks
to see the impact of the CPU swap. The ancient 3DMark2001 SE
benchmark is more of a CPU and chipset performance test these days,
as today’s graphics chips are far beyond it. The newer 3DMark03 and
05 are heavily constrained by GPU performance, but the CPU tests
lean heavily on the processor. We also ran our standard Quake 4 demoat a low resolution, so the videocard’s performance is out of the equa-
tion as much as possible. For media encoding, we used Nero Recode 2
to encode a video to H.264/AVC format.
We also tossed in a couple of new tests from Valve that stress
multicore performance. One is a particle test using Half Life 2’s Source
engine. Source isn’t currently multithreaded, but a patch due early next
year will add quad-core support. Valve also provided the media with a
map-compilation benchmark. It’s certainly not something a consumer
would undertake on a lazy Sunday afternoon, but it’s a good predictor
of workstation-level tasks. We’ve found the test to be multithreaded
and heavily processor limited. For battery tests, we used BAPCo’s
Mobile Mark 2005, which lets you test how long a notebook will run in
a real-world situation. Mobile Mark fully discharges and recharges the
battery before the benchmark runs. Although Mobile Mark 2005 also
measures the computer’s performance, the applications it uses are
quite old and not a good predictor of modern computer performance.
More important to us is its battery life test.C
ore 2 Duo on the desktop impressed us by blowing away the
Pentium D and Athlon 64 in just about every single bench-
mark we could find. The Core 2 Duo doesn’t destroy the Core
Duo, but it’s definitely faster across the board. We saw Intel’s
performance claims backed up with benchmarks deltas of 13 to
25 percent—and even a shocking 48-percent difference in our
video-editing test. Some of the Core 2 Duo’s advantage comes
from doubling the L2 cache to 4MB, but the rest is attributable to
tweaks Intel made to the Core microarchitecture. It’s simply fasterclock-for-clock than its sibling.
Although we have concerns about Mobile Mark 2005’s valid-
ity when testing multicore notebooks, it is the industry standard
way to measure battery life. According to our Mobile Mark bat-
tery rundown, you get faster performance with just a minor hit
to battery life by moving to Core 2 Duo. That’s truly an amazing
accomplishment when you think of it—more performance for the
same battery consumption. It amounts to a big win for consum-
ers and a big win for Intel.Core Duo T2600 Core 2 Duo T7400 PerCeNT DIFFereNCe
3DMArk2001 SE 18,614 21,118 13.5%
3DMArk03 ovErAll 7,456 7,642 2.5%
3DMArk03 CPU 992 1,254 26.4%
3DMArk05 ovErAll 3,123 3,130 0.2%
3DMArk05 CPU 5,793 7,514 29.7%
QUAkE 4 640x480 (fPS) 100 104.9 4.9%
NEro rECoDE AvC ENCoDINg (MIN:SEC) 55:56 45:53 21.9%
vAlvE PArTIClE PErforMANCE 32 40 25%
vAlvE MAP BUIlD (MIN:SEC) 6:42 5:52 14.2%
PrEMIErE Pro (MIN:SEC) 66:34 44:58 48.1%
PhoToShoP CS2 (MIN:SEC) 6:20 5:23 17.6%
MoBIlE MArk BATTEry lIfE (MIN) 201 189 -2.7%head 2 head Two Technologies enTer, one Technology leaves