Today on review we have the MSI
K7N420 Pro motherboard, a SocketA motherboard using DDR memory.
This motherboardboard is the first non-reference board using NVIDIA's
"Crush" or nForce chipset. Why is the nForce chipset
so interesting? Well, for a start, it's the first attempt at a
motherboard chipset by NVIDIA, who have over the last two years
established a strong domination of the video chipset industry.
The way the motherboard chipset industry works is very similar
to their approach to video cards. NVIDIA design and produce the
chipsets and sell them to other companies who produce usable products
around them. So, given their experience we'd expect them to have
some success in this new market. The development of this new chipset
has been watched with much anticipation and it's finally here.
This first page is going to be
inescapably wordy. If you don't want to read a tonne of text about
the internals of the chipset and just want to get into the meaty
stuff, skip this page and go to the next
one where we look at the board itself in detail.
Chipset features:
As with most modern chipsets, nForce is divided into two physical
parts. Most manufacturers refer to these as the North Bridge and
the South Bridge, but nVidia call them the IGP (Integrated Graphics
Processor) and MCP (Media & Communications Processor) respectively.
This is a little confusing because both chips contain more functions
than are described by their names. Anyway, lets take a brief look
at what's in each one, so we understand what we're looking for
in the benchmarks later on. NVIDIA have provided some technical
fact-sheets on their website, but they are a little overburdened
with marketing speak, so I'll try to give some plain-english explanations.
TwinBank Memory Architecture:
It's normal for the North Bridge to control memory access. In
this case, where most PC chipsets deal with memory sticks individually
(and therefore send 64 bits at a time), NVIDIA claim their chipset,
if two sticks of ram are present, will talk to them simultaneously
allowing 128-bit access. Note that only the nForce 420D chipset
(as used on this MSI motherboard) has the TwinBank architecture,
in the form of 2 separate 64-bit memory controllers. The nForce
220D has only a single 64-bit memory controller, like most other
chipsets. Anyway, this feature will potentially double the (already
DDR) bandwidth of the SDRAM system, with the theoretical maximum
being in the order of 4.2GB/second. Of course, it has to be noted
that this figure is considerably higher than anything else in
the system is capable of using. Even the AthlonXP CPU, with its
266MHz (DDR) FSB, is only capable of pulling 2.1GB/sec from the
memory subsystem. The real advantage is expected to be seen when
another memory-hungry device such as a video card is also trying
to pull lots of data from memory - in this case, the twin memory
controllers can service the CPU and the GPU simultaneously. In
this chipset this is important, because it contains an:
Integrated GeForce2MX GPU:
It surprised nobody when nVidia, with their dominance of the
graphics chipset market, announced that their new motherboard
chipset would contain integrated graphics. However, the prospect
of a powerful, genuinely 3D-capable onboard adapter was very
interesting. Previously, onboard graphics earned the instant
contempt of any game players or performance enthusiasts. Indeed,
in my earlier review of Transcend's TS-ASL3 based on Intel's
i815E chipset, after benchmarking I said I was going to draw
some fancy graphs to show the difference, but instead just imagine
a really short bar next to a really long bar, that's the i815e's
onboard video compared to pretty much any other video card you
can buy. This has unfortunately been true for all onboard
graphics to date and reviews usually end with an apologetic remark
about the motherboards being good for your grandmother's PC.
In this case however, NVIDIA include a full GeForce2MX core running
at 175MHz with all that entails including their 2nd generation
T&L engine, their Shading Rasterizer - it's fully buzzword-compliant.
It's apparently using an internal AGP 8X bus, the first we've
seen - this is according to the NVIDIA documentation, but in
the specs for this particular motherboard only AGP4X is mentioned
for the onboard graphics and normal AGP slot. An interesting
aspect of this integrated GPU is that it uses the system memory
for a frame buffer. Previous integrated chipsets generally had
a small amount of dedicated RAM inside the North Bridge but this
GF2 uses up to 32MB (selectable amount in BIOS) of your normal
system DDR ram. For that reason it's fairly hard to quantify
exactly where it fits, performance wise, into the normal spectrum
of GF2 cards as you may be running PC2100 or PC1600 RAM. Anyway,
we'll see the performance of the onboard unit later in this review.
As hinted at earlier, it has a normal AGP slot too, for future
more high-powered video cards. As they are both on the same AGP
bus, you can only use the onboard OR an add-in board, not both
at the same time.
Dynamic Adaptive Speculative
Pre-Processor (DASP):
I'm not going to even pretend to understand in any detail how
this mouthful of an acronym really works. From what I've read
about it, it's essentially a smart L3 cache inside the North
Bridge. If that's the case then we should see, as with any cache,
a general performance increase across the board with large increases
in certain (cache-friendly) benchmarks and no increase in cache-defeating
ones. Given that it's on the other side of the CPU's front-side
bus, you have to wonder how much of a performance boost there
will really be due to the latency involved with going off-core.
Still, it does some pre-processing which will mean less requirement
to go all the way to main memory for data, which is obviously
good. However, since there's no way that I know of to turn DASP
off, it's hard to quantify exactly how much benefit this provides.
It will be a factor in the overall performance of the chipset
which we will be looking at later in this review.
Communications:
This chip contains controllers for 10/100 Ethernet, HomePNA 2.0
(phone-line networking), 56k/sec modem and support for 6 USB
ports. Of course, it's up to the motherboard manufacturer to
decide if they want to provide the physical layer (plugs etc)
for those controllers. In the case of the MSI board, we have
an RJ-45 Ethernet jack, 6 USB connectors and an optional modem
connector (not included with the review unit).
On-Board Audio:
As well as the fairly standard AC97 audio codec, this chip also
contains a Dolby Digital 5.1 controller capable of driving up
to 6 separate speakers - which should considerably enhance watching
DVD's or surround-enabled games. This audio processor is the
same unit as used in the Microsoft XBox and NVIDIA claim it is
capable of processing 256 simultaneous audio streams. I'm not
going to be covering the audio in any great detail today.
StreamThru:
This is another buzzword which is hard to find any simple explanation
of. NVIDIA describe it as an optimized pipeline for streaming
data and claim it will smooth the flow of broadband, streaming
video and audio etc. There is a technical brief about StreamThru
here if you really want to know the specifics.
HyperTransport
Most existing chipsets use the PCI bus to connect the two bridge
chips. The PCI bus is not exactly cutting-edge any more and doesn't
really have the bandwidth a modern chipset requires. AMD's HyperTransport,
which I have to presume NVIDIA have licensed from AMD, is a new
connection between the bridges that allows 800MB/sec and frees
up the PCI bus for other work.
Ok. Now that your head is about
to explode from the theory of all that is nForce, let's jolt back
to reality and take a look at the physical layout and features
of this particular nForce board from MSI, the K7N420 Pro.
All content copyright 1999-2002 James Rolfe.
All rights reserved. No reproduction allowed without written permission. Interested in advertising on OCAU? Contact us for info.
