MaximumPC 2003 12

DECEMBER 2003 MAXIMUMPC 3

then converts the whole shebang into an

MPEG2 file using the app’s built-in encod-

er. Results are timed with a stopwatch;

lower scores are better.

Like previous versions of the app,

Premiere Pro performance is almost entirely

tied to CPU performance, but if a system

is short on memory, the app will hit the

hard drive, and this will increase process-

ing times.

Photoshop 7.0.1

Next month marks the return of

Photoshop , the world-renowned image

editing app, to our benchmark suite—this

time with a new software version (7.0.1)

and a new action script. Instead of hand-

picking a small number of filters to run

(and thus inevitably favoring a particular

CPU platform), we decided to just run ‘em

all, and let God sort it out.

Before we run this benchmark, we

defrag the system’s hard drive, reboot, and

set Photoshop to use 75 percent of avail-

able system memory. We then load our

test file (a 2MB JPEG), and start timing the

action script. The script runs all but three

of the filters that come with Photoshop

by default, applies a series of color mode

changes, and then converts the JPEG to

TIFF. We would prefer to use all the filters

that ship with the app, but we found that

three of them cause our script to stop.

Photoshop performance hooks into all

PC hardware subsystems, and our test

likes a fast CPU, lots of RAM, plenty of

bandwidth, and a fast hard drive. The

test is measured in seconds, and lower

scores are better. Adobe is about to ship a

new version of Photoshop , but we’ll likely

stick with Photoshop 7.0.1 for the next six

months before updating the software.

MusicMatch 8.1

The MusicMatch 8.1 MP3 encoding and

music management app comes standard

on many OEM machines, and its encoder

is among the best. The software tends to

favor raw CPU clock speed, but that’s to

be expected, as we haven’t run in to many

MP3 apps that don’t favor the P4’s raw

frequency power. Also, MP3 encoding is

pretty predictable from a branch-predic-

tion standpoint, so the P4’s long pipeline

doesn’t have a terrible impact on encod-

ing times.

To run our test, we first copy an Audio

CD WAV file—track 15 from the Star Wars

soundtrack—to our desktop. We then set

MusicMatch’s processing level to “very

high” and convert the file to a 160kbps

MP3, timing how long the conversion

takes (lower scores are better). If you’re

wondering why we convert just a single

track, it’s because the “very high” setting

uses “quantization,” which is particularly

punishing. To quote MusicMatch officials:

“Quantization is done via a power-law

quantizer. In this way, larger values are

automatically coded with less accuracy,

and some noise shaping is already built

into the quantization process. The quan-

tized values are coded by Huffman coding.

As a specific method for entropy coding,

Huffman coding is lossless. This is called

noiseless coding because no noise is added

to the audio signal.”

Yeah, it sounds like a Star Trek answer

involving zeta particles, but it’s real stuff.

The bottom line is that “very high”

encoding involves a whole lot of time-

consuming number-crunching and results

in kick-ass audio quality.

Jedi Academy

Jedi Academy is a sequel to the popular

Jedi Outcast, which we previously used as

a benchmark until a bug in its code made

our test demo unusable in multiproc and

Hyper-Threaded machines. Jedi Academy

runs on a modified Quake III Arena engine,

which is based on OpenGL. Our Jedi

Academy benchmark is affected mostly by

CPU performance, but does pick up some

continued on page 94

How to Read Our Benchmark Charts

Maximum PC’s test beds also double as “zero-point” systems, against which all review systems are compared.

Here’s how to read the new benchmark charts, which will first appear next month.

� ��� ��� ��� ���

PERCENT FASTER

������

�����

SCORES

�����������

������������

�������������

��������������

������������

����

���

�������

�������

�������

��������

��������

����

���

������

����������������

�������

������� ���

�������

�������������� SCORES

Our zero-point system includes: an ASUS SK8N mobo with nForce3 Pro 150 chipset, an Athlon 64 FX-51 CPU, 1GB of

registered DDR400 memory, an ATI Radeon 9800 Pro XT videocard, and a Western Digital WD2500 hard drive.

The scores achieved by our zero-

point system are noted in this

column. They remain the same,

month in, month out, until we

decide to update our zero-point.

The actual scores achieved by

the system being reviewed.

The bar graph indicates

how much faster the review

system performed in respect

to the zero-point system. If a

system exceeds the zero-point

performance by more than 100

percent, the graph will show a

full-width bar and a plus sign.

The names

of the actual

benchmarks

used.

Every month
we remind
readers of our
key zero-point
components.