MaximumPC 2007 04

32 MAXIMUMPC APRIL 2007

B

uying fl ash memory is like trying to pick

out a toothbrush. The products make tons

of promises, but it’s tough to measure

actual performance gains. And just like the

promise of whiter teeth, does the expensive

extra speed really get you anything? To test

that theory, we put a basic 2GB SanDisk SD

card ($70, http://www.sandisk.com ) against a 2GB

SanDisk Ultra II SD card ($80, http://www.sandisk.

com ). The standard card has no speed rating,

but the Ultra II promises 9MB/s reads and 8MB/s

writes. As a control, we also tossed in a 2GB

Extreme IV Compact Flash card that claims read

and write performance of 40MB/s, or 266x. The

Extreme IV’s performance, however, is gated

by the fact that the camera or card reader

must support UDMA mode 4 or Ultra ATA 66

speeds. Few do, and in fact, we ended up using

a SanDisk card reader capable of only 20MB/s

speeds for part of our testing. For a real-world

test, we used Canon’s EOS-1D Mark II N, which

is capable of shooting 8.2MP images at 8.5fps

for 48 frames. At full tilt, this camera sounds like

an M-249 SAW. We also used straight fi le-copy-

ing tests on a quad-core machine and ran some

synthetic benchmarks to back up the tests.

The result? If taking

pictures in a hurry is impor-

tant to you, memory cards

with higher ratings are worth

the extra cash. The Ultra II

SD card, which cost about

15 percent more, cleared the

write buffer on our Canon

EOS from 58 percent to 100

percent faster than the standard card. On our PC

tests, the faster SD card was about 48 percent

faster in writes, but interestingly, read speeds

were a wash. On the PC, the Extreme IV card

really unleashed pain on both SD cards by writ-

ing 2GB of data almost 300 percent faster than

the slow SD card and 170 percent faster than the

Ultra II SD card. The upshot is that it’s worth pay-

ing for some speed. However, with even today’s

fastest camera unable to write beyond 10MB/s,

it doesn’t make sense to pay for bleeding-edge

speed until the camera technology catches up.

W

hen it comes to graphics fi repower, can

you ever have too much? The answer

depends on two factors: the size of your

display and the power of your CPU.

Nvidia doesn’t recommend even benchmark-

ing a pair of 8800 GTX cards running in SLI at

less than 1600x1200 resolution because the CPU

becomes a performance bottleneck, leaving the

GPUs to twiddle their thumbs waiting for work. In

fact, Nvidia suggests that even single-card 8800

GTX benchmarking be conducted on a 30-inch

panel at 2560x1600 resolution. But some of us

fi nd gaming on such large displays a disorienting,

neck-straining experience—especially with shoot-

ers. Our current test panel is a 23-inch ViewSonic

VP2330wb, which delivers 1920x1200 pixels

with a 16:10 aspect ratio. Is it

worthwhile to run two 8800 GTX

cards at that resolution?

To fi nd out, we tested a

single 8800 GTX ($575, http://www.

asus.com ) card and then

paired it with an Nvidia refer-

ence-design 8800 GTX. We

installed both cards in an EVGA

nForce 680i SLI 775 mother-

board with 2GB of memory

and an Intel Core 2 X6800 CPU

stock-clocked at 2.93GHz.

The videocards ran at their

stock clock speeds (576MHz

core, 900MHz memory). Benchmarks consisted

of Quake 4 (ultra quality),

FEAR (soft shadows on), and

3DMark06’s Shader Model

3.0 games. We tested every-

thing at 1920x1200, 4x AA,

and 16x aniso.

Running in SLI

boosted frame rates by 80 percent or more in our

3DMark06 and FEAR benchmarks when com-

pared to our machine’s single-card performance,

but we saw “only” a 56-percent improvement in

our Quake 4 score. That renders SLI—even with

the screamin’-fast 8800 GTX—a worthy invest-

ment in our book. If your rig is outfi tted with a

lesser CPU, on the other hand, you’ll likely be bet-

ter off making that upgrade fi rst.

2GB ULTRA II SANDISK SD VS. 2GB STANDARD SANDISK SD

PREMIUM VS. STANDARD FLASH

SINGLE 8800 GTX VS. 8800 GTX SLI

8800 GTX SLI: A GOOD INVESTMENT?

BENCHMARKS

PERFORMANCE

SINGLE 8800 GTX 8800 GTX IN SLI DELTA

3DMARK06 GAME 1 (FPS) 25.8 23.7 80%

3DMARK06 GAME 2 (FPS) 46.4 43.8 85%

QUAKE 4 (FPS) 92.1 143.5 56%

FEAR (FPS) 69 126 83%

Best scores are bolded.

BENCHMARKS ULTRA II SD STANDARD SD EXTREME IV CF

TIME TO WRITE 184 RAW ON PC (SEC) 363 534 135

TIME TO READ 184 RAW ON PC (SEC) 187 190 110

CAMERA TESTING

WRITE 50 JPG ISO 100 (SEC) 17 27 18

WRITE 26 JPG ISO 3200 (SEC) 19 38 22

WRITE 22 RAW (SEC) 20 37 21

Best scores are bolded.