64
in the lab REAL-WORLD TESTING: RESULTS. ANALYSIS. RECOMMENDATIONS
and chipset performance. A shortfall in any of these areas will impact
performance in the benchmark.Adobe Premiere Pro 2.0
We’re finally tossing out our ancient Premiere test (originally cre-
ated in Premiere 6.0 ) for something that stresses today’s power
rigs. To build our new test, we used a Sony HVR-Z1U Pro to shoot
a few scenes at 1440x1080. We then add titles, transitions, overlay
video, special effects, and a soundtrack in WAV format.
We encode the short video to Windows Media Video 9 at 720p
resolution, using the Adobe Media Encoder. Premiere Pro 2.0
is actually the third iteration of the Premiere Pro engine, which
better supports multiple processor cores and Hyper-Threading.
Premiere Pro 2.0 even uses the graphics card to render some
effects and transitions.
Our Premiere Pro benchmark remains primarily a CPU test.
While our older test tended to favor the Pentium 4 microarchitec-
ture over the Athlon 64, we found that AMD and Intel CPUs per-
form about the same in our new test. A 3.46GHz Pentium Extreme
Edition 955 was no faster than our Athlon 64 FX-60. Both Intel and
AMD CPUs run slow, but adding a second core helps a ton. Our
zero-point system turns in a score about 40 percent faster than the
single-core $1,000 PC we built for this issue (see page 46). Not too
shabby when you consider that the pure clock differences between
the CPUs are about 400MHz.
Either way, the test is brutal. The final video is just 2 minutes,
46 seconds long, but the process of rendering and encoding
takes almost an hour with the fastest desktop machines avail-
able. Because most home movies of your baby or cat will likely be
an hour or more in length, this is one area where you can’t have
enough speed.Adobe Photoshop CS2
We’re also jettisoning our old Photoshop script in favor of a new,
even more devastating routine. Our Photoshop CS2 test starts
with an 8.2MP image taken with a Canon EOS 20D in RAW for-
mat. We then apply nearly every filter available in CS2 as wellas other common Photoshop CS2 functions. The entire test
takes about 4:54 seconds to complete on our zero-point system.
Although CS2 is slightly better at multithreading than the previous
versions of Photoshop , this test primarily uses a single CPU core.
It favors faster CPUs and more RAM, but chipset and hard drive
performance also affect the score.Ahead Nero Recode 2.0
Our old VOB-to-Divx test was good in its day, but in the last
year, DVD-ripping tools have improved greatly. To modernize our
video-encode benchmark, we selected Ahead’s Nero 7 Recode
2 application to encode a DVD movie to a format that’s playable
on a Sony PSP.
To remove the optical drive’s ripping speed from the equation,
we copy about 7GB of VOB files to the local drive before tasking
Recode with transcoding it to a PSP-friendly MPEG-4 file. Recode
2.0 is multithreaded and favors dual-core processors and AMD.
On our zero-point, it took roughly 35 minutes to encode our file
using a single-pass. A 3.46GHz Pentium Extreme Edition 955 took
about 19 percent longer.Quake 4
For graphics- and game-performance tests, we use a DirectX-based
game and an OpenGL game. For this benchmark rev, we’re upgrad-
ing from Doom 3 to Quake 4 (which uses a modified Doom 3 engine).
Besides besting the former title with flashlight-on gun technology,
Quake 4’s 1.11 beta version runs more than one thread, whether
you’ve got a dual-core or Hyper-Threaded CPU, or multiple proces-
sors. We run Quake 4 at 1600x1200 resolution with 4x antialiasing
and 4x anisotropic filtering. Above 4x AA, we noticed inconsistencies
in performance and image quality among different vendors’ GPUs.
To keep things fair, we side-step the issue by sticking with 4x AA.
We initially had qualms about adopting Quake 4 as a benchmark
because Raven called the patch that enabled multithreading the
“Intel patch.” The subsequent 1.1 beta and 1.1 final patches carryContinued from page 63Continued on page 66 ËWe’ve moved from editing standard-definition digital video to a
high-def source, using Adobe’s new Premiere Pro 2.0. To really
push the hardware, we include a punishing number of transi-
tions, effects, and video overlays.Our new Photoshop CS2 test script starts with a RAW file from
a high-end digital camera, to push the PC harder than a mere
JPEG would.MA XIMUMPC JUNE 2006