We also used the
synthetic Processor, Multimedia, Memory Bandwidth
and File System benchmarks built into SiSoft
Numbers With Sandra...
As expected, all of the
motherboards tested performed at virtually the exact
level. If you run Sandra's CPU test
repeatedly, the numbers will fluctuate slightly
each time. With this test, you cannot say
that any board was actually "faster" than another.
We saw more of the
same in the Multimedia tests. The numbers
are so close that no one board holds a clear
performance advantage over another.
Something that must be mentioned is that the Asus
A7V266-E, by default, is more aggressively clocked
than the other boards. While Sandra
reported the other boards to be operating at
1533MHz. (plus or minus 1MHz or 2), the A7V266-E
reported the CPU speed at 1541MHz. This
small clockspeed increase explained why the
A7V266-E consistently tested slightly faster than
the other boards in the round-up.
Before you see these
memory scores and say, "Hey! Those are way
to high!", let us explain something. The
latest version of Sandra has an updated memory
bandwidth test that now utilizes the data pre-fetch
and MMX/SSE/SSE2 instructions available in today's modern
processors. These scores are probably around
double what you're used to seeing, but they should be
more indicative of "real world" performance.
The new test still shows P4 / RDRAM equipped
systems dominating, but the gap between them and
Athlon / DDR systems is much smaller. For a
better explanation of what SiSoft has done,
check out the FAQ on their site.
As we mentioned
we entered the system
BIOS and set each board to their "Default High
Performance" settings, and then we set the memory
to CAS 2, 1T, with 4-Way bank interleaving, and
these are the scores we achieved. The Soyo
K7-DRAGON+ was the clear winner, with MSI's K7T266
Pro 2 once again lagging slightly behind the others.
Keep in mind though, there are quite a few other "tweakable"
memory settings in the system BIOS that could change
these scores significantly.
We found something
very interesting when testing the RAID 0
performance on each of these boards as well.
The Abit KR7A-RAID, outperformed all three of the other
boards tested by about 10% using the exact same
7200RPM UDMA/100 IBM hard drives. The Asus,
Soyo and MSI boards all used an ATA100 capable
Promise 20265 controller, while the Abit board
High-Point Tech. 372 controller.
It seems that High-Point Tech. has done a good job
with their newer parts and now hold a measurable
performance lead. A year ago, the tables
and Our Conclusion