How We
Configured Our Test Systems:
As
mentioned on the outset of our preview, our test
board had some trouble loading Windows XP on
either a Western Digital Raptor or Maxtor 250GB
Serial ATA drive. Support representatives
from VIA confirmed that the problem had been
encountered before, and we simply moved on to
use a 30GB IBM parallel ATA hard drive instead.
The rest of our testing went relatively
smoothly. Understandably, we're looking to
evaluate platform performance here. And
so, to minimize the biases that might favor an
ABIT board over an AOpen board (clearly, one is
going to be more "tuned" than the other), we
tested using relatively automatic settings.
For example, the Corsair "Pro" series is tuned
to run at 2-3-2 settings. Thus, we let
each motherboard operate according to the SPD
modules on the memory. Further, each board
was set to utilize comparable options, like a
256MB AGP aperture. And because the PT880
board wouldn't function with a Serial ATA drive,
we used the same IBM parallel ATA drive across
all three platforms.
For
each set of tests, the drive was reformatted for
a fresh installation of Windows XP with Service
Pack 1, DirectX 9.0b, and all of the latest
drivers. Accessory features, such as
System Restore and automatic update, were
disabled for consistency. We also disable
the screen saver, all power saving features, and
the aesthetic options available under the
"Advanced" tab in the "System Properties"
applet.
|
HotHardware Test Setup |
VIA
goes head-to-head with Intel |
|
Platforms Tested:
VIA PT880 Reference Board (VIA PT880/VT8237)
AOpen AX4C Max (Intel 875P/ICH5-R)
ABIT AI7 (Intel 865PE/ICH5-R)
Common
Hardware:
Intel Pentium 4 Processor
3.2GHz / 800MHz FSB
1GB of Corsair TwinX1024-3200LLPRO
Sapphire RADEON 9800 XT 256MB
IBM 30GB 7200RPM P-ATA HDD |
Software & Drivers:
Windows XP with Service Pack 1
DirectX 9.0b
ATI CATALYST Driver 3.9
VIA 4-in-1 Driver 4.49
Benchmark Software:
PC Magazine
Content Creation 2003
PC Magazine Business Winstone 2002
SPECviewperf 7.1
MadOnion.com PC Mark 2002
SiSoft Sandra 2004
Futuremark 3D Mark03 v.330
AquaMark 3 Standard Measurement
id Software Quake III (1.32) 'four' Demo |
|
SiSoft Sandra 2004 |
Synthetic Benchmarking |
|
The first test
is comprised of the Dhrystone and Whetstone benchmarks.
They are both multi-threaded 32-bit tests and the Dhrystone
test contains a sample of numerical operations employed by
various applications, while the Whetstone test contains the
sort of operations you'd expect to find in scientific and
engineering programs. While not significantly slower,
VIA's PT880 does turn in the lowest numbers in both
metrics. The difference is slight, though, varying by
less than three percent, which is in the benchmark's margin
of error.
The second test
generates an 860x750 picture of the Mandelbrot fractal using
255 iterations for each data pixel in 32 colors. Like
the previous test, it's multi-threaded for up to 64
processors and because we're using a Pentium 4, both the
integer and floating-point benchmarks run in SSE2 mode,
measuring performance in iterations per second. This
time, the results are even closer together than before,
indicating that, at least for multimedia performance,
platform selection is far less important than your processor
of choice.
Finally, the
SiSoft Sandra 2004 Memory Bandwidth benchmark attempts to
quantify the amount of sustainable throughput from a
system's main memory. It's like the old STREAM
benchmark, in a way, but it employs dynamic data rather than
static. When we first started talking to VIA about
PT880, representatives were most excited about the
platform's memory bandwidth numbers. After a quick
look over our results, it's easy to see why. Not only
is the PT880 able to edge out Intel's 865PE, but it also
tops the 875P, purportedly tweaked out in its own regard
with Performance Acceleration Technology. Do those
higher bandwidth numbers translate into better real-world
performance, though?
Winstones and PC Mark 2002
|