MaximumPC 2007 04

APRIL 2007 MAXIMUMPC 33

OPERATING SYSTEM BATTLE

T

he OS is fi nal. The drivers are done. It’s

time to put Microsoft’s new operating

system to the benchmarking test. We

set up one of our standard test beds with a

GeForce 8800 GTS board and the latest driv-

ers for both XP and Vista (97.92 and 100.59,

respectively) and then fi red up FEAR, Quake

4, Company of Heroes, and 3DMark06 to see

what was what.

We ran all tests at 1600x1200 with

all in-game visual quality settings maxed,

8x antialiasing and 16x anisotropic fi ltering

enabled, and hardware sound disabled in

XP. When testing in Vista, we ran with User

Account Control and Aero Glass enabled to

mimic a real-world Vista install.

As you can see in the chart below,

XP walloped Vista in every benchmark we

ran—by more than 10 percent in some tests.

Vista’s OpenGL performance, in particular,

was poor; our test rig’s Quake 4 performance

was a full 10fps slower in Microsoft’s new OS

than it was in Windows XP.

We were especially interested in the

performance of Vista’s Aero Glass user

interface, but unfortunately, our number

one tool for measuring desktop performance—

SYSmark—isn’t compatible with Vista yet. We’ll

revisit this test when we have a Vista-enabled ver-

sion of the desktop application suite.

We’re certainly not Vista apologists, but it’s

important to put these scores in context. We’re

comparing mature drivers based on a product that’s

been available for more than fi ve years with fi rst-gen

drivers for a product that’s been available for a couple

of weeks. That applies to not only the videocard but

the other hardware in the test rig as well—in Vista,

everything from the soundcard to the motherboard

is using either beta or fi rst-gen drivers, and these

results aren’t necessarily representative of the fi nal

performance of the operating system.

Still, if we’re playing the role of devil’s advo-

cate, we have to consider that Vista could just be

slower. When we look at the number of background

services running in Vista—from the search indexer

to SuperFetch—it’s easy to see why this might be

the case. The operating system’s just doing more.

And, while we usually choose a lean ’n’ mean PC,

we’re not averse to trading CPU cycles and memory

space for some of the additional functionality that

Vista provides.

So, where does that leave gamers? Unless

your rig is pushing heavy iron in the GPU depart-

ment—think SLI’d GeForce 7800s or better— you

might not want to face the frame-rate hit that a

switch from Windows XP to Vista will entail. If

you’re running new high-end cards, you might not

even notice the Vista performance penalty. We’re

also confident that performance will improve as

the drivers mature; however, we have no idea how

much (or how soon) things will improve.

And, of course, when DirectX 10 apps like Crysis

ship, Vista’s going to be the only operating system

they’re compatible with.

WHAT’S THE

PERFORMANCE HIT FOR

RUNNING BITLOCKER?

WINDOWS VISTA VS. WINDOWS XP

VISTA HARD DRIVE

ENCRYPTION

WINDOWS XP PRO WINDOWS VISTA ULTIMATE

FEAR (FPS) 61 56

COMPANY OF HEROES (FPS) 69.4 60.

3DMARK06 (FPS) 8,151 8,

QUAKE 4 (FPS) 84.4 74.

Best scores are bolded.

BENCHMARKS

UNENCRYPTED BITLOCKER ENCRYPTED

DRIVE DRIVE

FEAR (FPS) 56 57

3DMARK06 8,071 8,

PCMARK06 HARD DRIVE SUITE 6,133 4,

PHOTOSHOP CS SCRIPT (SEC) 303 330

Best scores are bolded.

BENCHMARKS

V

ista Ultimate and Enterprise

include BitLocker, a utility that

encrypts your hard drive in real time

using industry-standard 256-bit AES

technology. We love encryption, but

we haven’t spent a lot of time with

whole-disc encryption—instead,

we encrypt only sensitive files and

directories. After all, we don’t want

to slow down our entire system just

to encrypt our tax returns.

With that in mind, we set out

to fi nd out what BitLocker’s impact

is on different types of apps, and

our results were interesting, but

somewhat expected. In applications

that don’t require frequent hard drive

access, like games, there was virtually

no difference between the encrypted

and unencrypted drives. However,

when we busted out the hard-drive-

intensive benches, we immediately

saw as much as a 20 percent hit on

the encrypted system. We’re not sure

the PCMark06 slowdown represents

real-world performance, but our

Photoshop script, which includes

tons of hard-drive-intensive reverts,

showed a 10 percent hit as well.

What’s the bottom line? We’re

not going to enable BitLocker on

systems that don’t hold crucial, mis-

sion-critical data.