Author: Tom Karpik
Editor: Howard Ha
Publish Date: Wednesday, February 15th, 2006
Originally Published on Neoseeker (http://www.neoseeker.com)
Article Link: http://www.neoseeker.com/Articles/Hardware/Reviews/asus_a8n32-sli/
Copyright Neo Era Media, Inc. - please do not redistribute or use for commercial purposes.
With relatively little fanfare, NVIDIA announced in August of 2005 that they had updated their nForce4 SLI chipset. Dubbed the "nForce4 SLI x16", the key feature of this new chipset was the addition of another 16 PCI-Express lanes, finally allowing SLI operation with both cards running in x16 mode. Before the existance of this chipset, SLI was achieved by splitting up the 16 lanes originally intended for a single PCI-E x16 slot into a set of two 8-laned x16 slots.
Even though it was available for both the Athlon 64 and Pentium 4 platforms, most hardware manufacturers seemed to stay hush-hush about the nForce4 SLI x16. Dell was really the only manufacturer that showed some enthusiasm right off the bat -- they immediately snatched the chipset up for inclusion in their flagship XPS gaming line of desktop PCs.
Since then, a number of manufacturers have picked up the chipset, and we're finally seeing some definitive products making it to the shelves. The first of those products was ASUSTek Computer's A8N32-SLI Deluxe, and we are happy to say that ASUS has kindly sent us a retail box sample of the board.
Let's see what this motherboard, and its chipset, have in store for us!
So what makes the nForce4 SLI x16 tick? The secret is one that NVIDIA has not made use of on the AMD platform since the nForce2 days. Yes, the North bridge has once again been reunited with its slightly disabled cousin, the South bridge. We saw a similar configuration with NVIDIA's nForce4 for Intel chipset, but the AMD side of things has remained exclusively single-chip for the entire life span of the nForce3 and nForce4 chipsets -- until now, of course.
Bear with me here while I explain the chipset soup going on:
The AMD version of the nForce4 SLI x16 uses a C51-series North bridge, dubbed the C51D. The C51 series is also home to the C51G North bridge, which may ring a bell -- the C51G is what is dubbed as the nForce 410/430 chipset, and it features integrated GeForce 6100/6150 graphics. The C51D merely seems to be a C51G without those integrated graphics.
The South bridge is simply the very familiar nForce4 SLI MCP -- the same one found on all other AMD-platform SLI boards. It seems that NVIDIA has figured that putting "two chipsets" on a single board is more cost-effective than simply designing an entirely new one with the extra PCI-Express lanes. The two chips are linked together with the same 16-bit, 1 GHz HyperTransport bus that the chipset/CPU usually enjoy, resulting in 8 GB/s of bandwidth between the North and South bridge, as well as between the processor and the North bridge.
The following is a block diagram of the nForce4 SLI x16 chipset:

A nice little touch is the addition of RAID 5 support -- something that was sorely missing from the nForce4 up until now. If full x16 SLI operation doesn't bear any actual fruit, RAID 5 could be the nForce4 SLI x16's *only* new worthwhile feature.
Let's take a look at the ASUS A8N32-SLI Deluxe, and then on to some testing.
The Bundle
ASUS usually does quite an impressive job when it comes to their motherboard bundles -- and especially so for their "Deluxe" models. This time is definitely no exception, as you can see below:
An inventory of everything that you see in the photo is as follows:
Not too shabby at all. I particularly like the addition of five SATA data and power cables. It really speaks something about the manufacturer when they include enough cabling to make use of all the capabilities of their product right out of the box, rather than having to hunt around in computer stores for some kind of cable or other. That said, let's see what the rest of the board brings.
The Board
At first glance, the A8N32-SLI Deluxe seems to be a winner in the layout department. Not only does it seem to lack clutter issues, but the aesthetics of it are also appealing. This is probably in no way unrelated to ASUS' innovative single-heatpipe cooling design.
The problem with a motherboard that sports two complete chipsets is one of heat. Our DFI Lanparty nF4 SLI-DR has a single chip, and it features an active cooling design. While this does the job, such fans are not only noisy, but prone to failure. ASUS has opted to go with the more engineered option, and has implemented a single heatpipe system that spans across both chips. In the grand scheme of things, this system allows the board to work silently, all while looking more professional.
Moving on, we have to make the standard mention of whether or not the memory slots are obstructed by video cards, and all that other jazz. Fortunately, that is not the case, though there is something that has the potential to be even more annoying.
Consider the following: The typical overclocker makes very good friends with his/her CMOS reset jumper, and usually wastes absolutely no time in doing so. Some of these overclocker types like to put high-end video cards into their systems -- and dare I say, maybe even two such cards. Once that is the case, the CMOS reset jumper becomes useless -- obscured by the second video card in its slot. This is a small oversight on ASUS' part, but it has the potential to be infuriating if you need to reset your CMOS often -- and who doesn't?
In any case, I can't spot any other layout deficiencies. The ATX and Molex power connectors could possibly be placed a tad higher up on the board, so as to prevent the fat ATX cable from draping over the CPU socket region, but that's really starting to get nit-picky.
As far as features are concerned, ASUS has done a great job with some of them. We've already mentioned the interesting cooling design, so let's add a few more:
It's apparent that ASUS is targeting the enthusiast market with this motherboard. With loads of room for expansion, and ASUS' excellent overclocking capabilities to boot, this board ought to be a favourite.
One thing I would have been absolutely thrilled to see is an integrated Creative SoundBlaster Live!/Audigy, rather than the usual Realtek ALC850 codec. A few motherboards on the market already feature such an extravagant audio solution, and I would be delighted to see ASUS make that move as well.
The physical side of things is looking delightful for the ASUS A8N32-SLI Deluxe. What does the BIOS have in store for us?
As we're starting to see on an increasing basis, ASUS is making use of an AMI BIOS on the A8N32-SLI Deluxe. Personally, while I am impartial to the brand of BIOS used, I find AMI BIOSes a touch faster to navigate, due to the navigation bar up top, as opposed to a root menu as can be found on Award BIOSes.
The central point of interest for tweakers is going to be the Advanced menu, so we'll go straight there.
First off, ASUS' exclusive "LAN Cable Status" screen is a good tool for checking LAN connectivity and the length of your cable. Here it can be seen reporting an average 7.6 meter length -- which I found to be more or less correct.
The JumperFree Configuration menu is home to a number of frequency/voltage adjustments. Most of these are invisible until a manual mode is selected, at which point the crown jewels are revealed.
The adjustable ranges and modes of the options on this screen are as follows:
Next up is CPU Configuration. There is nothing terribly interesting here, with the exception of the 1T/2T memory timing option, but it does lead into the Memory Configuration screen.
And here is the Memory Configuration screen.
The adjustable ranges and modes of the options on the Memory Configuration screen are as follows:
The Chipset menu contains the last few options that will be of interest to most of us. They are as follows:
This is a very robust BIOS -- no doubt about it. I would even go so far as to say that it is on-par with the DFI Lanparty nF4 Ultra/SLI series BIOSes, but with a few extra kicks. It takes a little bit of time to get used to the menu layout, since most of us are more familiar with the Award layout, but once you get past that hurdle, it's smooth sailing.
Let's get this out of the way quickly!
Our benchmark system for the standard set of motherboard tests was as follows:
The competitor for our ASUS A8N32-SLI is the DFI Lanparty nF4 SLI-DR -- a mature, proven, and established motherboard that is rock-solid and consistently puts out very impressive performance numbers. The ASUS A8N32-SLI Deluxe is up for a tough challenge, but going by what I have seen so far, I feel confident that it will not leave us hanging.
Our benchmark system for the SLI testing was identical, with the exception of the CPU and memory timings. For our standard motherboard testing, the Athlon 64 3700+ was used at stock (200 MHz HT, 11x mult., 2200 MHz) settings, while the RAM was being pushed at 2.5-3-3-8 1T timings. Our SLI testing was done with the same processor, but at overclocked (300 MHz HT, 9x mult., 2700 MHz) settings, while the RAM had to be loosened to 2.5-4-4-10 2T timings.
A list of benchmarks performed follows:
* denotes SLI testing
Drivers used were ForceWare version 81.98, nForce4 AMD/Intel x16 version 6.85 (for ASUS A8N32-SLI), and nForce4 AMD version 6.70 (for DFI Lanparty nF4 SLI-DR).
That's it, enough stalling. It's time for some pretty graphs!
Business Winstone 2004

The DFI board has a very minor lead over the ASUS board, but the results were consistent in this manner. This is obviously a non-issue, as a 0.4 difference in scores will not be noticeable even to a hardcore business application user.
Multimedia Content Creation 2004

The MMCC scores paint an identical picture for the ASUS board -- it is just slightly slower than the DFI in these tests.
SiSoft Sandra Memory Bandwidth

The DFI board leads our ASUS board by approximately 40 MB/s in both integer and floating point memory bandwidth. Since both boards were tested with virtually the same OCZ memory at identical 2.5-3-3-8 1T timings, I'm sure this slight lead can be explained by minor tweaks that DFI has made to their board's BIOS/chipset. In any case, memory bandwidth scores for both motherboards are still very impressive.
WinRAR Compression

The WinRAR compression algorithm is extremely sensitive to memory timings, as Bill has noted in past reviews. We can see here that whatever gave the DFI board a lead in Sandra's memory bandwidth benchmark is doing the same for WinRAR.
HDTach SATA
Since our standard motherboard benchmark platform now includes a Seagate Barracuda 7200.9 SATA2 drive as part of the HDTach SATA testing, do not be surprised by the extremely high burst read results. SATA2 has a maximum theoretical bandwidth that is twice that of the older SATA standard, and while this increase does not necessarily translate to real-world performance increases for the most part, it does allow for much higher burst read possibilities.

We can see here that between both motherboards, burst and average read scores are identical, but the ASUS A8N32-SLI pulls in those performance numbers with a 7% CPU utilization, while the DFI handles it without any measurable increase in CPU utilization. Overall, I'd have to call the DFI the definite winner here.
HDTach IDE

The ASUS board completely eats our DFI for breakfast when it comes to IDE performance. Not only are burst read speeds significantly higher, but it is all done at a lower CPU utilization. Good stuff.
HDTach USB 2.0

Both of our boards have very good USB 2.0 implementations, though the ASUS is just under 2 MB/s faster.
HDTach has shown us that it's a toss-up between the ASUS A8N32-SLI Deluxe and DFI LanParty nF4 SLI-DR -- one is great for IDE, while the other is great for SATA. You be the judge.
TMPGEnc MPEG2 Encoding
Note that we are now publishing results for timed benchmarks in pure seconds, rather than minutes (dot) seconds.

Both boards have identical MPEG2 encoding scores, when you take into account even a miniscule margin of error. Nothing to see here.
XviD Encoding

Again, the margin of error eliminates any chance of conclusive interpretation. Both of these boards are face-to-face when it comes to video encoding.
LAME MP3 Encoding

We can draw the same conclusion about MP3 encoding as we did with video -- in other words, there are no performance concerns. Both motherboards seem to be taking advantage of the available hardware equally as well.
RightMark Audio (2D)

Two-dimensional audio implementations seem to be identical between these two boards. Again, moving along.
RightMark Audio (3D)

I didn't expect to see anything different when moving to hardware-accelerated 3D -- and I didn't see anything different. Perhaps 3D+EAX performance will finally show us something interesting?
RightMark Audio (3D+EAX)

Nope. Same old.
I think it is interesting to note that the ASUS A8N32-SLI Deluxe, while always being within a hair of matching the DFI board, has consistently shown itself to be just a tad slower at all tasks up until now. Admittedly, the differences between results are nearly immeasurable, but they are there, and they are almost always in the DFI board's favour. I'm willing to say that this is probably due to the DFI being a mature product that has undoubtedly enjoyed many a BIOS tweak from its mommy, and nothing more.
Call of Duty

Call of Duty scores are as close to each other as you can get without getting suspicious. Let's see whether throwing high-quality EAX 2.0 sound into the mix changes things around at all.

... Not one bit. Then again, this was to be expected, after seeing how close our RightMark Audio scores were to each other.
Doom 3

Doom 3 is more of the same thing -- even down to the slight DFI lead.
Comanche 4

No comments are necessary.
Halo

And again, there isn't anything of interest here.
Jedi Knight: Jedi Academy

Consistency is a great thing, isn't it?
Unreal Tournament 2004

Our final motherboard test ends in an uneventful whimper as well.
So what have we learned? We have learned that the ASUS A8N32-SLI, as a standard motherboard, performs just as well as our established, mature favourite -- the DFI LanParty nF4 SLI-DR. The DFI did seem to have a lead in most of our tests, but I attribute this to slight improvements in the BIOS (in the form of timings, possibly) on DFI's part.
We are now going to move on to a few high-end game tests in SLI in order to compare dual PCI-E x16 performance to regular dual PCI-E x8 performance. Will it be more of the same old thing, or will the ASUS finally pick up and stretch its legs?
Doom 3

The DFI board seems to have a very slight lead over the ASUS board in all test configurations. It is more important to note that true dual PCI-e x16 did not help the ASUS out here at all.
Half-Life 2

These results really are too close to call, with the only possible exception being plain 1600x1200. Again, the dual 16-lane PCI-E did not improve performance one bit over the standard nForce4 SLI chipset.
Serious Sam 2

The DFI board started out with a significant lead here, and then blended in with the ASUS. On the other hand, the ASUS seems to have a delicate lead in 1024x768 with 4x AA/8x AF, but that doesn't last long either.
FEAR

The ASUS board races out of the gate with a decent lead, but then falls back to stay in line with the DFI. So far we haven't seen any definitive victories (or losses) for either of our two boards.
Splinter Cell 3: Chaos Theory

In an almost spiteful fashion, both boards have scored virtually identically to one another here.
I think the concluions that we can draw are obvious -- either the modern day crop of video cards simply do not require more bandwidth than what 8-lane PCI-Express provides, or modern day games do not fully saturate even 8-lane PCI-Express with texture/geometry data. In either case, having dual 16-lane PCI-Express slots is currently, at best, a marketing gimmick. Based on our simple (but conclusive) analyzation, we can say that a PCI-E x16 SLI solution simply does not offer any performance advantages over regular PCI-E x8 SLI at this point in time.
The ASUS A8N32-SLI Deluxe has lived up to my expectations, and surpassed them in a few cases. ASUS should be commended on a job well done as far as motherboards go. The board was not only rock-solid during testing (with the exception of stability problems on a certain day caused by dirty power in our lab), but actually managed to keep our old favourite, the DFI Lanparty nF4 SLI-DR, in close sights the entire time.
Features-wise, the board is packed with expansion and storage capabilities. To further distance it from run-of-the-mill nForce4 motherboards, ASUS even implemented a very creative single-heatpipe passive cooling solution -- a most welcome addition for any silent-PC freaks. "Alright", you say. "The board is neat, it's fast, and it's stable ... so what's wrong with it?"
What's wrong with the ASUS A8N32-SLI Deluxe is the fact that its main selling feature is really just a marketing gimmick. That's not to say its dual 16-lane SLI capabilities are entirely useless, because we never know what kind of bandwidth requirements future video cards/games will have, but at this point in time, all it is is a marketing gimmick that earns you some bragging rights. In our testing, it was revealed that true 16-lane SLI did not improve performance one bit.
In fact, we expected as much, seeing as how NVIDIA stated this themselves when SLI was first launched. They explained that reducing the lane width to both cards to only 8 lanes would have a non-existant to very minimal impact on performance. NVIDIA was absolutely correct about that, yet they've still released the nForce4 SLI x16 chipset. Extravagant? Definitely -- but that stops no one.
All in all, I'm very pleased with the ASUS A8N32-SLI Deluxe. Whether or not bragging rights are important to you, this motherboard still has things to offer to any potential SLI user. If it's any consolation to those who were expecting magnificent gains in performance from this motherboard's SLI-on-steroids capabilities -- you still get RAID 5, which is something those boring, old regular nForce4 SLI users don't have.
Thanks for reading!
Please do not redistribute or use this article in whole, or in part, for commercial purposes.