MaximumPC 2004 03

In the Lab A behind-the-scenes look at Maximum PC testing

SYSmark 2004: More is Better

Our favorite benchmarking app continues to prefer fast CPUs, lots of RAM,

and memory bandwidth. The big change: More detailed reporting

60 MAXIMUMPC MARCH 2004

W

hen BAPCo introduced the 2001 version of

SYSmark , some benchmarkers were disappointed

that the company eliminated a great deal

of the detailed analysis that the program previously

reported. While SYSmark 98 used to tell you how long it

took for, say, Premiere to run, SYSmark 2001 threw away

most of this data. That’s because individual application

performance could no longer be measured with the

benchmarking suite’s new model of concurrent use,

or running multiple applications at once. Similarly

spare, SYSmark 2002 reported just three scores: Overall,

Internet Content Creation, and Office Productivity.

While the concurrent-use model worked fine, the lack

of data resolution left testers guessing which program

was performing poorly. Thankfully, with SYSmark 2004 ,

BAPCo returns us to the good old days of detail.

Although you still can’t tell specifically whether it

was Dreamweaver MX or Premiere that hurt (or helped)

performance, at least you can now narrow your guess

down to two or three apps and the type of usage behind

the scores. This allows us more insight for analysis when

a PC system throws up poor numbers.

In addition to the Overall score, SYSmark 2004 now

reports 3D Creation, 2D Creation, and Web Publication

as aspects of the Internet Content Creation score.

SYSmark 2004 also breaks down the Office Productivity

score into three divisions: Communication, Document

Creation, and Data Analysis. Our benchmark chart will

continue to report the overall SYSmark 2004 score, but

we’ll report details of the individual test runs when

relevant.

Following are details of SYSmark’s Internet Content

Creation test:

 The 3D Creation score indicates how fast the tested

machine can render a 3D model in 3ds max 5.1 while

preparing a web page with Dreamweaver MX.

 The 2D Creation score is based on how long a

system takes to use Premiere 6.5 to create and export a

movie. This test also uses Photoshop 7.01 and Adobe’s

After Effects 5.5.

 The Web Publication score is derived by using

WinZip 8.1 to extract a compressed file while Flash

MX is used to open an exported vector file. A movie is

encoded with Windows Media Encoder 9 and scanned

with VirusScan 7.0.

The Office Productivity test includes:

 A Communication score based on how well the

system performs in Outlook 2002 while VirusScan 7.0 runs

in the background. Internet Explorer is also used to view

BAPCo’s corporate web site and browse files created by

SYSmark.

 The Document Creation section involves editing a

file in Word 2002 and using Dragon NaturallySpeaking 6 to

dictate a document. The document is then turned into

a PDF using Acrobat 5.0.5. Finally, PowerPoint is used to

create a presentation based on the document.

 The Data Analysis portion of the Office Productivity

test queries a database in Access 2002. These results

are imported into Excel 2002 and used to generate a

graphical chart.

Great, but what are the results?

We ran SYSmark 2004 on our zero-point Athlon 64 FX

equipped with an ATI Radeon 9800 XT card to generate

a score that can be compared with other review systems.

We also wanted to see the impact of memory bandwidth

on our system, so we ran it in a single-channel

configuration and with the memory cranked down to

200MHz instead of 400MHz.

The results surprised us. The Office Productivity

section of SYSmark 2004 proved to be extremely sensitive

to memory bandwidth. This manifested itself in the

single-channel run as well as when we ran the main

memory at DDR200 rates instead of DDR400.

While the Office Productivity score affected

bandwidth and hard disk I/O, the Internet Content

Creation score is less sensitive to bandwidth, but scales

nearly perfectly with clock speed. Decreasing by half the

speed of a 3.2GHz P4 Prescott to 1.6GHz, for example,

netted us a 50 percent drop in performance for the

Internet Content Creation portion, while the Office

Productivity dropped by just 39 percent.

CPU

2.2GHz

Athlon 64

FX-51

2.2GHz

Athlon 64

FX-51

2.2GHz

Athlon 64

FX-51

1.6GHz

Pentium

4E

3.2GHz

Pentium

4E

SYSmark 2004

Dual

Channel

DDR400

Single

Channel

DDR400

Dual

Channel

DDR200

Dual

Channel

DDR400

Dual

Channel

DDR400

Overall 172 166 157 98 185

Internet Content Creation Overall 200 198 190 100 213

3D Creation 193 192 188 104 203

2D Creation 244 239 225 101 257

Web Publication 170 168 161 94 185

Office Productivity Overall 148 140 129 97 160

Communication 143 143 143 101 143

Document Creation 177 172 159 96 178

Data Analysis 127 112 94 94 161

SYSmark 2004 reverses BAPCo’s trend of providing less granular

detail for each of its scores, with several new scoring subcategories.