|
Athlon
XP 2800+, AMD Hit 333!
|
|
Author : Wayne
Date : 1st October 2002
|
Test System :
ASUS A7N8X nForce2 motherboard
Epox 8K3A+ motherboard
Abit TH7II [i850]
AthlonXP 2800+(333MHz)
AthlonXP 2100+ Thunderbird core
Intel Pentium4 2.53GHz
512MB Corsair XMS3200
ATi Radeon9700 Pro 128MB (Cat 7.76)
WindowsXP Pro
A Look at Performance:
You can call me an old cynic but despite of my
admiration for AMD and all they've achieved over the years,
when they suddenly announce a product that they know users have
wanted for at least the past twelve months I get rather cautious.
I want to be certain that the process used to deliver said product
was the approved application of industrial strength black magic
and not that they've taken an axe to the architecture just to
get the thing to run to spec. With this in mind I decided the
first thing I should do was to drop the FSB to 133MHz and effectively
turn this 2800+ into a 2200+ (13.5x133).....is nothing sacred!
This doesn't really cover every possible technique AMD could
have used to get the chip running reliably with a 166MHz FSB
but it should at least give us an idea.
Sandra Report 2800+ @ 2200+
I'd originally planned to plug in a 2200+ to use
as a comparison but SiSoft's Sandra saved me the trouble, not
to mention the fact that the 2200+ was in the workshop across
a dark and rainswept back yard. In both the arithmetic and multimedia
tests the hobbled 2800+ matched Sandra's 2200+ reference scores
incredibly closely but I have to admit in every case it did
come in a fraction lower.
SiSoft Sandra CPU Arithmetic
| |
ALU
|
FPU
|
|
2800+ @ 2200+
|
4995
|
2472
|
|
2200+
|
5007
|
2505
|
SiSoft Sandra CPU Arithmetic
| |
Integer
|
Floating Point
|
|
2800+ @ 2200+
|
9961
|
11015
|
|
2200+
|
10016
|
11102
|
Either way this is enough to convince me that
there's been no axe swinging going on, it was the black magic
that did it!
Just to double check I ran 3DMark on a genuine
2100+ that was lying around the desk and on a declocked 2800+.
Once again the declocked 2800+ came in behind the real deal
despite now running a full 66MHz faster (133MHz faster internally).
I won't publish the whole run of tests I laboriously sat through
but I can say that in every case the 2800+ came in second, often
by fractions but sometimes by a little more than a fraction.
Without more testing I'm not sure precisely why this is, possibly
a cache, latency or prefetch issue. My main concern is that
this minor performance shortfall will magnify as the FSB increases
and so we won't quite be getting the kind of increases we were
all expecting from the boosted FSB speeds.
2800+ @ 2200+
2200+
Time to flip the settings back to where they rightly
belong and let the 2800+ flex a little muscle. To begin with
we'll, see what SiSoft Sandra has to report :
Sandra Memory Bandwidth :
2800+ on ASUS A7N8X nForce2 Motherboard
2800+ on Epox 8K3A+ KT333 Motherboard
Sandra CPU Arithmetic :
2800+ on ASUS A7N8X nForce2 Motherboard
2800+ on Epox 8K3A+ KT333 Motherboard
Sandra CPU Multimedia :
2800+ on ASUS A7N8X nForce2 Motherboard
2800+ on Epox 8K3A+ KT333 Motherboard
Two things become clear here, the first is that
the XP2800+ is a very impressive performer. Combined with NVIDIA's
nForce2 we're seeing numbers we could only have dreamed of previously
and this potent combination looks set to buy AMD some time in
their ongoing rivalry with Intel who have hit on an architecture
that seems infinitely flexible and almost limit free where operating
speeds are concerned. Of course AMD's architecture is far more
efficient but it was beginning to look like AMD were set to
run into a MHz brick wall.
<<<
Back to Review Index | Page
3 - Performance >>>
Home
