|
TESTING METHODOLOGY:
To
help explain our benchmark scores, we feel it is necessary to
explain how a system was setup before running
the benchmarks. On all of the boards, I started off by
manually optimizing the
BIOS settings to the most aggressive RAM
settings and system options. The memory
frequency was manually set to DDR266 for the
i845 boards, but since the SiS648 fully
supported DDR333, we chose to do so to show the
benefits of the higher memory clock speed (Keep
in mind the i845PE is not available, which also
officially supports DDR333). The hard drive was then
formatted, and Windows XP Professional with
Service Pack 1 was installed. After the Windows
installation was complete, we installed the
chipset drivers and used the latest download for
the Intel Application Accelerator on the i845
systems. Next, we installed the drivers for the
rest of the components.
Auto-Updating and System Restore were disabled,
and we set up a 512MB permanent page file.
On these test systems we set the visual quality
to "best performance" in Windows XP's system performance
menu, as
well as in the video driver settings. Lastly, I
installed all of the benchmarking software,
defragged the machine, and rebooted one last
time.
Overclocking the 648 Max
was a chore, to say the least. Using a
256MB stick of Corsair PC3000 DDR, we felt safe using the "fast"
setting in the BIOS. The AGP/PCI speed was locked
at 33/66MHz, as it removed a possible source of
problems with devices running out of spec. Fuzzy
Logic 4 was a Windows based software utility we found on the CD which was
supposed to help us find the maximum CPU
overclocking value by simply clicking on "Auto".
Every time we did this, however, the system locked up.
Believing the program to be at fault, we started
changing the bus speed manually from within the BIOS,
which is the preferred method. We
did not have much luck, however. We were only
able to reach a maximum FSB of 139MHz. We were able to
get Windows to load with a 143MHz FSB, but almost all
of the benchmarks failed at one time or another.
We were only able to complete them all successfully at
139MHz.
 |
|
The Hot Hardware Test Systems |
|
Intel on
the inside, hard crunchy coating on
the outside |
|
|
TEST BOARDS:
MSI 648 Max Pentium 4 Motherboard (SiS 648)
IWILL
P4ES Pentium 4 Motherboard (i845E)
DFI
NB-76EA Pentium 4 Motherboard (i845G)
COMMON
HARDWARE:
Intel
Pentium 4 2.26GHz CPU (533MHz FSB)
256MB Corsair PC3000 DDR
Asus V8420 Deluxe GeForce 4 Ti 4200
On-board AC'97 audio
Western Digital WD200BB ATA100 7200rpm 20GB Hard
Drive
Creative Labs 52x CD-ROM
Windows XP Professional with Service Pack 1
Intel Chipset Drivers v4.00.1013
Intel
Application Accelerator v2.3.2144
SiS
AGP Driver 1.11
NVIDIA Detonator Drivers v40.41
|
 |
|
Performance Comparisons with SiSoft
SANDRA 2002 |
|
Sandra,
Sandra, Sandra! |
|
|
First
up on the list of testing software was SiSoft
Sandra 2002 Professional. It's a quick and
easy way to compare results
from any system against an internal database of similar systems. These benchmarks are theoretical
scores, and can't necessarily be measured in real-world
terms, but they do provide a good way to make
comparisons amongst like components. We ran a set of tests at both
our CPU's default
2.26GHz, and then at the overclocked speed of 2.36GHz.
Here are the results:
CPU Arithmetic
at 2.26 GHz
 |
CPU Arithmetic
at 2.36GHz
 |
At
2.26GHz, the CPU scores were in line, between
the 2GHz and 2.4GHz Pentium 4 reference
system. The score is actually much closer to
the 2.4GHz score, only off by less than 100MIPS in
the Dhrystone test and by 150 MFLOPS in the
Whetstone test. It still falls slightly behind
the two Athlon reference score, however.
Overclocking the system gets us slighter better
than 2.4GHz performance, and overtakes the
Athlon XP 2000 as well. The Dhrystone score
is still well behind the Athlon XP 2200+.
CPU
Multimedia
at 2.26 GHz
 |
CPU
Multimedia
at 2.36 GHz
 |
In the
next test, the MSI board performed well with the Integer calculations score
coming in right behind the Pentium 4 2.4GHz CPU
and the Athlon XP 2000. Floating point
calculations are right up near the top of the
charts, falling short of the 2.4GHz score by only
500 points. With the FSB set to 139MHz,
the scores just miss the Pentium 4 2.4GHz in both
tests.
Memory
Bandwidth
at 2.26GHz
 |
Memory
Bandwidth
at 2.36GHz
|
Here
was the MSI 648 Max's chance to shine. We
had expected the memory bandwidth scores to be
impressive, and we weren't let down by what we
saw. While it was hard to find proper
comparisons in Sandra's database, the scores we
obtained easily outclassed other boards using
DDR200/266. While this should be obvious
since we were using DDR333, check out the small
difference between the 648 Max's score, and that
of a Intel 850E reference board with PC800 RDRAM.
Both of our scores are only 150MB/s off the RDRAM
scores. That's quite impressive considering
the considerable difference in price.
 |
|
Performance Comparisons with PC Mark
2002 |
|
CPU,
Memory and Hard Drive Testing |
|
The
next series of tests we ran were with MadOnions
PCMark 2002. PCMark 2002 performs a series
of tasks, such as image compression, text searches
and audio conversion, to give us three scores:
CPU, Memory, and Hard Disk Drive (HDD).
It is a relatively quick process for comparing the
performance of two or more systems. We
compared the MSI 648 Max at stock clock speeds,
and while overclocked, to an Iwill board powered
by the i845E and an i845G based motherboard from
DFI.
What
we've got here are some darn close scores, with
the MSI 648 Max coming out on top by a mere
point. Even though we had limited success
while overclocking, the 6MHz increase in FSB raised our benchmark score by
almost 250 points, far more than we
expected in this test. This resulted in a 4.5% increase
in performance.
With
the memory performance module, the extra bandwidth
provided by running the memory at DDR333 speeds,
helped the MSI 648 Max
pull away from the two boards. It beat the IWILL board by 350
points and the DFI board by more than 400, about a
7-8% difference. Interestingly, the
difference is not near the 20% difference between
the speed of a DDR333 module and a DDR266 module.
Overclocking brought us another 127 points.
In the hard drive
test, the MSI and IWILL
boards were neck and neck, with the DFI score
falling well
below the others. The MSI and Iwill boards were
both using ATA133 ports, while the DFI board was
using ATA100 though.
|
|
Gaming & The Winstones
 |
|