|Introductions, Specification & Bundle|
We certainly do evaluate and showcase our fair share of ASUS products here at HotHardware, and it's no great mystery because Asus is also typically one of the first out of the block with a new chipset technology. The company's new A8N32 SLI Deluxe motherboard is no exception. NVIDIA's new nForce 4 SLI X16 chipset has been out the door at NVIDIA for months now, but retail boards have yet to appear on the scene. We've heard ABIT and others may have boards on the horizon, based on this new high-end NVIDIA chipset, but the A8N32 SLI Deluxe was the first to hit our test bench, and it looks like it will be the first to hit retail sometime this week. NVIDIA has also launched the Intel version of the nForce 4 SLI X16, and we're also working with that board, dubbed the P5N32-SLI Deluxe, right now.
The new NVIDIA nForce 4 SLI X16 chipset takes the same basic feature set of the hugely successful nForce 4 SLI and expands upon its PCI Express lane configuration, offering two full X16 graphics slots for dual graphics SLI configurations. This adds another 16 lanes to the chipset, as well as 16 additional 2.5G PCI Express SerDes (serializer/de-serializer) interfaces, which in turn adds to chip real-estate and pin count. As a result, in this incarnation, the nForce 4 SLI X16 has become a discrete two-chip solution for north and southbridge functionality. We'll give you more on the chipset later; for now let's run down what the ASUS A8N32 SLI Deluxe has to offer.
We've also grown accustomed to ASUS delivering nicely appointed bundles with its "Deluxe" and "AI Life" branded boards, and the A8N32 SLI Deluxe's set up is pretty much cut from the same cloth. There are an abundance of SATA data and power cables to enable various configurations, as well as various USB, Game Port, Firewire, and serial IO back-plate modules. ASUS bundles WinDVD suite with the board along with the drivers and utilities disk. In addition, you get an SLI bridge connector, but it looks like ASUS has had to go back to the flexible ribbon cable version with the A8N32 SLI Deluxe. This isn't quite as robust, in terms of staying fastened to the cards, as the rigid PCB version. Unfortunately there wasn't much of an option here if you wanted to populate either of the two PCI slots in between the board's two PCI Express graphics slots. Which brings us to another small shortcoming of the A8N32 SLI Deluxe, which we'll discuss shortly.
|The A8N32 SLI Deluxe - Up Close|
The chipset that the A8N32 SLI Deluxe is built around is essentially a beefed-up version of the already beefy NVIDIA nForce4 SLI for the AMD platform.
In this version of the chipset, NVIDIA expanded PCI Express connectivity to a total of 32 lanes specifically dedicated to graphics for full support of two X16 PCI Express Graphics cards in SLI mode. No longer must the cards split a single X16 connection into a pair of X8 electrical connections in each slot. Now each PEG slot is fully connected with an X16 link to the root complex, in this case what is pictured as the "SPP" or System Platform Processor in the block diagram above. This is what can be classically thought of as the northbridge of the chipset, while the NVIDIA "MCP" or Media Communications Processor pictured above handles the I/O and Storage connectivity, for support of Serial ATA drives also in RAID 0, 1, 0+1 and 5 configurations, as well as USB2.0, Firewire, Gig-E, and NVIDIA ActiveArmor Secure Networking and Firewall features. Below is a table that details all of the various incarnations of the entire nForce 4 family, for your reference.
The A8N32 SLI Deluxe board itself is an impressive assembly of technology to behold. Chipset cooling on the board is a totally passive solution, employing the company's heat pipe technology along with its "Stack Cool" fanless design to cool both the SPP and MCP as well as the CPU power array. ASUS also offers "optional" cooling fans that clip on top of the power array heatsink fin assembly, but this is not a requirement for normal operation or even overclocking conditions with air cooling for the CPU. Rather, ASUS recommends the use of these fans in extreme cases where watercooling or some other air-free method of cooling, such as Vapor Phase Change, is used.
ASUS also built the A8N32 with an eight-phase power MOSFET power module array. ASUS claims their eight-phase or eight stage power array provides a cleaner, more ripple free signal to the CPU that also has the ability to react more quickly to voltage and current transient demands as heavy loads are placed upon the CPU.
However, the location of two of the board's standard PCI slots make it a very tight squeeze for anyone who actually runs this board with a Dual Graphics SLI setup. And isn't that the whole intent of the product, with its extra X16 PCIe connection? If you would actually like to use those two PCI slots for expansion, say a TV-Tuner card or something of that nature, air flow between the PCIe Graphics slots will be very restricted. In addition, with any dual-slot cooled graphics card (ASUS' own N7800 GTX Top for example), you'll be limited to a single half-height PCI card in between the PEG slots, and the end PCI slot on the board is also completely obstructed, as well. We were left wondering what the physical layout limitation may have been that caused ASUS to drop not one but two PCI slots in between the PEG slots, when perhaps they could have been placed at the end of the line where they could be more easily accessed and utilized.
|The BIOS & Overclocking|
The standard issue configuration and overclocking features are present, of course, along with what we think is one of the more important features: ASUS C.P.R. or "CPU Parameter Recall", in the event you set up a non-bootable configuration. If it is detected that the CPU locked up, timings will be set back to default parameters on the next boot attempt. This feature is also accompanied by a large selection of DRAM and CPU voltage settings, in .125V increments. Standard 1MHz CPU timings are present, as well as a full assortment of multiplier options for both CPU and DRAM.
There is, however, a very unique BIOS menu option called "DDR DRAM Skew," which basically is driven from an on-board PLL from ICS that allows programmable advance or delay of the DRAM clock skew in 150ps (picosecond) increments. Frankly, we're not sure exactly how useful this feature is, in that DDR DRAM timing has very tight skew and jitter characteristics that must be met to ensure stable performance at any speed. However, perhaps this setting, when trying to stabilize things in extreme overclocking scenarios, will have some sort of normalizing affect on the clock signal. We played a bit with this setting and weren't definitively able to determine if it bought us any more DRAM timing margin while overclocking. Still, more control is a good thing, and we plan on exploring this feature a bit more in the weeks ahead.
Overclocking and Health Monitoring utilies such as ASUS PC Probe and AI Booster are part of every ASUS board bundle and are offered on the all-in-one driver CD that comes with the A8N32SLI Deluxe. PC Probe II is a basic utility for monitoring system temps, fan speeds, and voltages, while AI Booster facilitates real-time dynamic overclocking depending on processor workload.
You'll notice in the screen shot above that our motherboard temp, which is typically a reading taken from the PWM circuit on the motherboard, is showing a somewhat toasty 44oC. Although this was taken with an open-air bench setup, it also was a reading we got while overclocking our Athlon 64 FX-57 processor to over 3GHz. Regardless, with the nature of the fanless heat pipe design of this motherboard, we do recommend a well-ventilated case.
Of course, our well-ventilated HotHardware test bench is a perfect environment for overclocking, so we dialed up the ever-capable BIOS menu options of the A8N32 SLI Deluxe and like others before it in its lineage, the board was as smooth as silk under pressure.
In the end we were able to realize a 260MHz system bus driven to the Athlon 64 FX-57 processor we used for testing. We then backed down the CPU multiplier to 12X and hit a top clock speed of 3.12GHz. This tops the speed that we were recently able to hit with the ASUS' A8N SLI Premium back in July of this year by about 20MHz. This was nothing earth shattering for the A8N32 SLI Deluxe, in terms of capabilities over its predecessor, but pretty darn spiffy nonetheless.
|Our Test Systems & SiSoft SANDRA|
How we configured our test systems: When configuring the test systems for this showcase, we first entered the system BIOS and set each board to its "Optimized" or "High-Performance Defaults." We then manually configured our system RAM to run at 200MHz (DDR400), with the timings set to CL 2,2,2,5 settings. The hard drive was then formatted and Windows XP Professional (SP2) was installed. Then we installed all of the necessary drivers and removed Windows Messenger from the system altogether. Auto-Updating and System Restore were also disabled, and we set up a 768MB permanent page file on the same partition as the Windows installation. Lastly, we set Windows XP's Visual Effects to "best performance," installed all of our benchmarking software, defragged the hard drive, and ran all of the tests. Our World Bench 5 tests were done on a completely clean install as the very first batch of each test we ran.
We began our testing with SiSoftware's SANDRA, the System ANalyzer, Diagnostic and Reporting Assistant. SANDRA consists of a set of information and diagnostic utilities that can provide a host of useful information about your hardware and operating system. We ran three of the built-in subsystem tests that partially comprise the SANDRA 2005 suite (CPU, Multimedia, and Memory). All of these tests were run with our A64 FX-53 processor set to its default clock speed of 2.4GHz (12x200MHz).
Sandra is more of a quick sanity check for us than anything else. The results shown here are pretty much as expected for the processor we chose, as well. What's perhaps somewhat interesting and impressive are the memory scores we took at DDR400 CAS 2,2,2,5 settings with the A8N32 SLI Deluxe in excess of 6000MB/sec. This is right in line with what we've seen on the A8N SLI Premium historically, as well. Moving along to the Sandra hard disk benchmark, we see a strong 40MB/sec driven from the WD Raptor drive, which again is right on top of what we've seen in previous versions of this board with the standard nForce 4 SLI single chip solution. It seems at least as far as this base level test goes, the extra latency of having discrete north and southbridge components with the new nForce 4 SLI X16 chipset has no immediately apparent effect on performance. However, there is still a lot more testing to be done so let's move on.
|PCMark05: CPU & Memory|
For our next round of synthetic benchmarks, we ran the CPU and Memory performance modules built into Futuremark's brand-new PCMark05 benchmark suite. Below is a brief description from FuturemMark on just what these modules run and measure in terms of performance metrics.
"The CPU test suite is a collection of tests that are run to isolate the performance of the CPU. The CPU Test Suite also includes multithreading: two of the test scenarios are run multithreaded; the other including two simultaneous tests and the other running four tests simultaneously. The remaining six tests are run single threaded. Operations include File Compression/Decompression, Encryption/Decryption, Image Decompression, and Audio Compression" - Courtesy FutureMark Corp.
So what does the above test tell us? Well, in short, an Athlon 64 FX-53 runs about the same on any nForce 4 SLI-based chipset board. the MSI board lagged the three ASUS boards by a hair, but the differential is minuscule.
"The Memory test suite is a collection of tests that isolate the performance of the memory subsystem. The memory subsystem consists of various devices on the PC. This includes the main memory, the CPU internal cache (known as the L1 cache) and the external cache (known as the L2 cache). As it is difficult to find applications that only stress the memory; we explicitly developed a set of tests geared for this purpose. The tests are written in C++ and assembly. They include: Reading data blocks from memory, writing data blocks to memory, performing copy operations on data blocks, random access to data items, and latency testing." - Courtesy FutureMark Corp.
We've specifically had issues getting the A8N SLI Deluxe version board to complete PCMark04 and PCMark05's memory test module, which is why you see the "WNR" for "would not run" in this chart. Beyond that again, both of the other A8N series boards put up nearly identical scores, leaving the MSI K8N Neo4 slightly behind by less than 3%.
|WorldBench 5: Photoshop & Office XP|
PC World Magazine's WorldBench 5.0 is a new breed of Business and Professional application benchmark, poised to replace the aging and no-longer supported Winstone tests. WorldBench 5.0 consists of a number of performance modules that each utilize one or a group of popular applications to gauge performance. Below we have the results from WB 5's Photoshop 7 and Office XP modules, recorded in seconds. Lower times indicate better performance.
Interestingly enough, in both the Photoshop 7 and Office XP SP2 WorldBench 5 tests, we see the new A8N32 SLI Deluxe fall behind the rest of the pack by a small margin, 1 - 2% to be exact. While this certainly doesn't constitute a significant and end-user perceivable variance, it does speak somewhat to what could be the slightly higher latency characteristics of NVIDIA's discrete MCP chip in the new nForce 4 SLI X16 chipset versus the totally integrated single-chip approach in the nForce4 SLI standard solution. Is it anything to quibble over? Absolutely not. Again, the 5 - 7 seconds overall, with the number of tests these benchmarks run, is nearly within the margin of error for any given test run.
|Windows Media Encoder 9 & LAME MP3|
We continued testing the ASUS A8N32 SLI Deluxe with the benchmark module based on Windows Media Encoder 9. PC World Bench 5's Windows Media Encoding test reports encoding times in seconds. Lower times indicate better performance.
Once again the A8N32 SLI Deluxe takes last place versus its other ASUS counterparts and the MSI board but again with only what could be considered an inconsequential shortcoming of approximately 1%.
In our custom Lame MP3 encoding test, we convert a WAV digital audio file to the MP3 format, which is a very popular scenario that many end users work with on a day-to-day basis to provide portability and storage of their digital audio content. In this test, we chose a large 223MB file (a never-ending Grateful Dead jam) and converted it into an MP3 file type. Processing times are recorded below. Shorter times equate to better performance.
Our Lame MP3 encoding test is largely dependent upon raw CPU horsepower and memory bandwidth. It takes zero explanation to note that, as far as raw audio conversion goes, all of these boards are to be considered created equal.
|KribiBench v1.1: 3D Modeling|
Next up, we ran Kribibench, a rendering benchmark produced by the team at Adept Development. Kribibench is an SSE aware software renderer. A 3D wireframe model is rendered and then animated by the host CPU, and the average frame rate is reported. We used two of the included models with this benchmark: an "Exploded Sponge" model consisting of over 19.2 million polygons and a gargantuan "Ultra" model that comprises over 16 billion polys...
Our Kribibench more or less shows all of these boards, including the A8N32 SLI Deluxe, about on equal footing with each other. The A8N32 SLI Deluxe does show a very slight edge over the other A8N series boards, as well as the MSI K8N Neo4. Because there is consistent tracking between both model tests here, we would consider the edge it has to be real albeit once again pretty much insignificant.
|Cinebench 2003 & 3DMark05: CPU|
The Cinebench 2003 benchmark is an OpenGL 3D rendering performance test, based on the commercially available Cinema 4D tool suite. This is a multi-threaded, multi-processor aware benchmark that renders a single 3D scene and tracks the length of the entire process. The time it took each test system to render the entire scene is represented in the graph below (listed in seconds).
We should note that because the host processor we ran for all of the motherboards in this test was a single-core Athlon 64 FX-53, the only option available was to test in Cinebench's single CPU thread mode. Multi-core processor performance, as one would see with an Athlon 64 FX X2, for example, would undoubtedly be faster.
Once again, whether we consider the single-chip implementation of the nForce 4 SLI or the two-chip nForce 4 SLI X16 with its additional 16 lanes of PCI Express on board, processing throughput from either solution is virtually identical.
3DMark05's built-in CPU test is a DirectX game engine performance metric that's useful for comparing relative performance between similarly equipped systems. This test consists of two different 3D scenes that are generated with a software renderer, which is dependant on the host CPU's performance. This means that the calculations normally reserved for your 3D accelerator are instead sent to the central host processor. The number of frames generated per second in each test are used to determine the final score based on a weighted average.
With 3DMark05's CPU test, the ASUS A8N32 SLI Deluxe took the lead position by a more comfortable margin. It's important to note that this was achieved with a single GeForce 6800 GT installed, so the board's additional X16 graphics slot isn't affecting the results in any way. What we're looking at here is a 5% differential from the fastest score turned out by the A8N32 SLI Deluxe and the slowest score offered by the A8N SLI Deluxe. Amongst the three nForce 4 SLI classic boards it's a virtual dead heat, but the A8N32 SLI Deluxe seems to have slightly better overall bandwidth in this test that mainly highlights CPU polygon throughput.
|Low-Res Gaming: UT & Doom 3|
To start our in-game testing, we did some low-resolution benchmarking with Unreal Tournament 2004. With this specific in-game test, we used "Low-Quality" graphical settings and low screen resolution, which almost completely isolates CPU and memory performance.
In UT2004 we're witness to a less than 2% variance in scores amongst all the motherboards we tested. The MSI K8N Neo4 Platinum SLI took the lead spot this time by a hair, most likely due to its use of a Creative SB Live 24 chip for its on-board sound processing. If we turned up the resolution and level of detail in this test, you would see all of these boards fall completely in line with respect to frame rate.
For our next game test, we benchmarked all of the test systems using a custom multi-player Doom 3 timedemo. We cranked the resolution down to 640 x 480 and configured the game to run at its "Low-Quality" graphics setting. Although Doom 3 typically taxes today's high-end GPUs, when it's configured at these minimal settings it too is also much more CPU / Memory-bound than anything else.
In this test, the tables turn in favor of the ASUS A8N32 SLI Deluxe, but again, the performance delta is so marginal that it's really not worth considering. So when looking back at things here, the moral of the story is that as far as standard usage models in general CPU and Memory bandwidth measurements go, the new A8N32 SLI Deluxe and its nForce4 SLI X16 two-chip set offers every bit of performance that the legacy single chip solution does today. So let's look at a couple of areas where the board may perhaps have more of a competitive advantage...
|SLI Performance: Doom 3|
With Doom 3's graphical settings configured for "high quality" mode and the screen resolution turned up to 1280X1024 and 1600X1200, the game engine and graphics rendering loads are a heavy burden for even one of the more powerful graphics processors on the market today, the GeForce 6800GT. Certainly the recently released GeForce 7800GTX would have a much easier time with this, but then again a pair of 6800GTs in parallel should offer nearly twice the performance.
Here we'll also have a look to see if a full X16/X16 PCI Express SLI Graphics setup has any tangible performance benefit.
Well now, we do see a distinct pattern here with our high-quality Doom 3 tests. Without AA or Aniso Filtering turned on, versus the MSI K8N Neo4 Platinum SLI, the ASUS A8N32 SLI and its dual X16 PEG slots takes a more significant 9% lead at 1280X1024 and approximately 4% at 1600X1200. Again this is without AA and AF turned on. Enable AA and AF, and the two boards are right on top of each other, with the A8N32 SLI Deluxe scoring a few extra frames at 1280 res. So it seems as if the more GPU bound we are, the less of a lead the new nForce4 SLI X16 chipset has, but as frame rates increase, dual X16 PCI Express graphics has its advantages. We asked NVIDIA about the results we were seeing, for a more complete explanation and this is what we were told.
From Nick Stam, Director, Tech Marketing - NVIDIA:
"You may see better scaling at 1x/1x (AA /AF) on x16 vs x8 SLI due to a few reasons, in descending order of most impact.
1) The higher frame rates obtained at 1x/1x (or even at lower resolution) actually requires more data to be exchanged between the two GPU cards across PCIe per unit time (more frames per second means the cards must exchange data more frequently). Resolution variations actually don't have as much impact across PCIe as the increased frame rates do in the GPU-GPU exchanges over PCIe. While much data does exchange through the SLI bridge connector, there is still some data that goes across PCIe between GPUs.
2) With higher frame rates, the CPUs can now pump more data across PCIe to the GPUs.
Future scenarios with lots more geometry being passed to GPUs, a lot more textures, SLI-AA, and so on, will see more improvements over time.
|SLI Performance: Splinter Cell Chaos Theory|
We chose the Shader Model 3 code path for our Splinter Cell Chaos Theory high-quality SLI testing but turned off all HDR features, so our GeForce 6800 GT pair had an easier time with this demanding game engine.
Splinter Cell Chaos Theory is a case where the game engine is so GPU shader-processing bound that we don't see any real advantage for the A8N32 SLI Deluxe and the nForce 4 SLI X16 chipset over legacy nForce4 SLI-based boards. However, let's not forget the scenario Nick Stam from NVIDIA described with SLI-AA enabled. We'll have a look at that shortly, but let's get a baseline on that topic first with some standard FarCry testing.
|SLI Performance: FarCry|
FarCry is an interesting game and game engine in that although it employs some of the more up-to-date, leading-edge DX9 pixel shader and lighting effects, it's still somewhat CPU-bound, as you'll see in the following tests. Well, at least when you're driving things with a pair of GeForce 6800 GTs, anyway.
Between 1280 and 1600 res, we see very little variance in performance overall between the two motherboard solutions we tested. The ASUS A8N32 SLI Deluxe does, however, have a slight advantage consistent with what we saw in our Doom 3 tests, without Anti-Aliasing or Anisotropic Filtering enabled. Again, this is due largely to the fact that there is more inter-GPU communication going on at these higher frame rates. With AA and AF turned on, however, the scores are completely leveled.
Next we'll plug in a pair of GeForce 7800 GTX cards and turn on SLI-AA to see if we can invoke even more GPU-to-GPU traffic over PCI Express, with the blending that is required in this type of AA rendering technique.
|SLI-AA Performance: FarCry|
Recalling what was explained to us by NVIDIA regarding potential performance advantages of the company's new chipset in SLI-AA scenarios, we set out to prove the theory ourselves. Below are tests with the 1.33 version of Far Cry with SLI-AA 8X and 16X modes enabled at 1280X1024 resolution. We compared frame rates of the new ASUS A8N32 SLI Deluxe nForce 4 SLI X16-based motherboard versus the legacy ASUS A8N SLI Deluxe nForce 4 SLI standard board.
** Article Updated 11/11/05 - BIOS Incompatibility Found **
When our article first launched on 10/18/05, we reported to you our findings based on the hardware and related driver versions we had in hand at the time. Unfortunately the Asus A8N SLI Deluxe motherboard we tested in comparison to the A8N32 SLI Deluxe, needed a new BIOS update to take advantage of NVIDIA's recently released 6.82 nForce chipset driver. This driver version was released in support of the nForce 4 SLI X16 chipset and we came to find out that the legacy A8N SLI Deluxe had performance issues with this new driver. Asus sent an updated BIOS for the A8N SLI Deluxe to us on 11/2 and we had to re-run our numbers in order to confirm our initial performance findings between the two motherboards. Here is what Asus reported to us with this BIOS release...
" A8N-SLI Deluxe 1015.006 --> Support the NVIDIA 81.82 graphics driver and nForce 6.82 driver to increase X16 performance. "
At the time we were testing with the 1013 version of the A8N SLI Deluxe BIOS. Our first pass benchmark findings can still be seen here but clearly they are erroneous due to the fact that the A8N SLI Deluxe's performance was hampered by this chipset/BIOS incompatibility. The following are our updated SLI-AA benchmarks with Far Cry.
Obviously, we do not see the 55 - 65% performance advantage we initially reported back on 10/18/05, for the A8N32 SLI Deluxe and NVIDIA's new nForce 4 SLI X16 chipset. In reality, there is a more modest 5 - 15% performance gain over the X8 SLI setup on the A8N SLI Deluxe, due to increased bandwidth for inter-GPU communications during frame blending operations.
|Benchmark Summary & Conclusion|
What can be said about the ASUS A8N32 SLI Deluxe that hasn't already been proven out under our large battery of test conditions? Frankly, there are a few caveats with this new ASUS motherboard, but beyond that, the numbers speak for themselves. First, those considering this motherboard should be very aware of its current mechanical limitations, should you want to run a pair of graphics cards in SLI mode. And again, we're not sure why you would opt for this motherboard if this wasn't at least a consideration for you in the future. So with that in mind, just make sure you only need two extra PCI slots (preferably one if you want things thermally comfortable inside your chassis) before taking the plunge on the A8N32 SLI Deluxe.
Mechanical challenges aside, our performance testing with the ASUS A8N32 SLI Deluxe started as a somewhat understated but precision-tuned melody and ended on a triumphant crescendo. Our thought process in this product assessment began with a ho-hum sort of feel, but we finished off whole-heartedly impressed.
In fact, if you're considering going the way of an SLI rig, we see no reason why not to spend a bit more on a full-up nForce4 SLI X16 solution, and the ASUS A8N32 SLI Deluxe, with an MSRP of around $195 USD, is a very good option to consider.
We'd just caution on the use of single-slot graphics cards and remind you to think about chassis real estate management first. However, for delivering in all our performance test scenarios and then taking the lead in certain NVIDIA SLI-AA driven modes, we're giving the ASUS A8N32 SLI Deluxe a 9 on the HotHardware Heat Meter.