|Introduction, Specifications, and the Board|
Hot on the heals of the official announcement of Intel's new dual-core Pentium Extreme Edition 840 processor, we have more news for those interested in Intel platform solutions. Today, NVIDIA is finally taking the wraps off of their nForce chipset for the Intel platform, a chipset now known as the nForce 4 SLI Intel Edition. For months there was speculation that this chipset would be called the nForce 5, but because it offers essentially the same feature set as its AMD supporting counterpart, NVIDIA thought it best to keep this chipset in the nForce 4 family.
The feature set of the NVIDIA nForce 4 SLI Intel Edition remains largely unchanged from other nForce 4 chipsets. However, since Intel's and AMD's architectures are quite different, NVIDIA had to design a memory controller to work with Intel's current (and future) processors. AMD's Athon 64 processors have their memory controller on-die with the CPU, and don't need a memory controller to be incorporated into the core logic chipset. NVIDIA's memory controller will largely determine the performance of their nForce 4 SLI Intel Edition platform, but NVIDIA seems confident that they were up to the task. So, let's fire up an nForce 4 SLI Intel Edition motherboard and see how she performs.
To test the nForce 4 SLI Intel Edition chipset, NVIDIA shipped us a reference motherboard with features that should be indicative of retail-ready products. We should note that the motherboard we tested does not support Intel's new dual-core processors, even though the nForce4 SLI Intel Edition was designed for both single and dual-core processors from the start. NVIDIA has informed us that support for dual-core processors is board dependent, and that top-tier manufacturers will have dual-core supporting nForce 4 SLI Intel Edition boards available shortly.
Unlike the single-chip nForce 4 SLI chipset for AMD processors, the Intel Edition is a two chip configuration. Since Intel's processors don't have an on-die memory controller like the Athlon 64, NVIDIA has designed their own Dual-DDR2 memory controller for the current Intel platform. The SPP (System Platform Processor) that incorporates NVIDIA's new memory controller, among other things like support for 20 PCI Express lanes, is built on a 0.13 micron manufacturing process and consists of roughly 61 million transistors. The companion MCP (Media & Communications Processor) which incorporates the features of a traditional Southbridge, is a .15 micron part comprised of about 21 million transistors. Together these two chips provide all of the functionality found in the AMD version of the nForce 4 SLI chipset, and then some.
|NVIDIA's nForce 4: Feature Overview|
The NVIDIA nForce 4 SLI Intel Edition shares many of the same features as the nForce 4 SLI for the AMD platform. The nForce 4 SLI Intel Edition has full support (plus more) for nVRAID, which has been re-branded "MediaShield" for this launch. It also incorporates NVIDIA's "ActiveArmor" and Firewall technology, along with support for NVIDIA's "nTune" performance monitor and tweaking application, and of course SLI. For more information on the nForce 4's main features, we suggest taking a look at this article. We cover the nForce 4 SLI's main features in that piece.
With the nForce 4 SLI Intel Edition, NVIDIA has beefed up their NVRAID technology with the ability to support bootable RAID 5 arrays. And with this new feature, also comes a new name - "MediaShield". As hard drives have become less and less expensive, we've gotten to the point where the data being stored on the drive is more valuable than the drive itself. Of course, this has always been the case in the enterprise, but on home PCs, in many cases, users weren't really storing much valuable data. Today, however, with huge MP3 and movie collections now commonplace, having the ability to secure personal files, and access them quickly, is very important. With MediaShield, users have the ability to create RAID 0, RAID 1, RAID 0+1, JBOD, cross controller RAID, spare disk allocation arrays and have a bootable RAID 5 array. And with NVIDIA's support software installed, users are given a graphical representation of their motherboard, in the event of a drive failure. The port that is highlighted on the alert, is associated with the drive that failed, so zeroing in a problem drive is relatively easy.
NVIDIA also points out that the drive controllers integrated into their MCP are more robust than some competing solutions. NVIDIA claims that the advantage of a dual-controller architecture, like the one in their MCP, is that it provides two separate and independent transfer paths to and from system memory, which will result in up to twice the bandwidth to disk. NVIDIA's MCP also supports SATA II (300MB/s) transfers and Native Command Queing (NCQ). Two features that aren't native to any other core logic chipset, with the exception of Intel's recently announced ICH7 Southbridge.
ActiveArmor / NVIDIA Firewall / Native Gigabit Ethernet:
The screenshots above represent just a few of the sections available to users in the nVFirewall interface. NVIDIA's Firewall is quite powerful and generally has more features than the average router/firewall many broadband subscribers currently use. The nVFirewall offers many advanced management features such as remote access, configuration, monitoring, command line interface (CLI), and WMI scripts. And on top of the firewall and antihacking features of ActiveArmor, it supports stateless and stateful packet inspection, Web-based management, predefined security profiles, port block filtering, and Intelligent Application Manager, remote administration, and an easy-to-use configuration wizard. The nVFirewall also incorporates some anti-IP-spoofing, antisniffing, anti-ARP cachepoisoning, and anti-DHCP server technologies.
|NVIDIA's nForce 4: Feature Overview Cont.|
nTune Performance Application:
For example, with nTune users can tweak their system to squeeze more performance out of it while gaming, or even clock down the system and lower fan speeds to keep the system quiet while playing a DVD or video file. With nTune users can alter FSB speeds, memory timings, and tweak key voltages. Changes are made within the application, while in Windows, eliminating the need to make changes via the BIOS or reboot the system.
Some of the key features of the NVMixer application include control over all playback and recording volumes, simple tweaking of speaker or headphone configurations, and integrated help information that's available on the same page as the actual feature in question.
SLI (Scalable Link Interface):
|A Closer Look at the Memory Controller|
Excluding the custom chipset used in the Xbox, the NVIDIA nForce 4 SLI Intel Edition is the first nForce core
The DualDDR2 memory controller incorporated into the nForce4 SLI Intel Edition's SPP interleaves the two memory channels in such a way that CPU access can be sent simultaneously to both channels. But the type of interleaving used depends on whether the two memory channels are populated symmetrically or asymmetrically. When the channels are populated symmetrically, with identical memory modules, the DualDDR2 memory controller uses a finer-grain interleaving mode. If the channels are populated asymmetrically though, the DualDDR2 memory controller drops down to a coarser-grain interleaving mode. For the highest performance, it's best to populate both channels with matched memory, because system performance will be somewhat lower when the two channels are coarsely interleaved. We should mention, however, that the memory controller in nForce 4 SLI Intel Edition always operates in a 128-bit wide mode, regardless of whether or not the two memory channels are populated with matched DIMMs. Some competing solutions drop down to 64-bit when not populated symmetrically.
NVIDIA's memory controller can operate the memory interface at 667MHz (peak 10.6GB/s sent in parallel) and higher data rates, and includes other features designed to improve performance. One enhancement is a dedicated address and command bus (also known as the address bus) for each DIMM. Providing a dedicated address bus to each DIMM, rather than sharing busses across multiple DIMMs, should improve performance because there are fewer shared resources and the memory controller can also operate with a 1T command rate, which reduces overall memory latency.
|Test Systems & SiSoft SANDRA 2005|
How we configured our test systems: When configuring the test systems for this review, we first entered their system BIOS and set each board to its "Optimized" or "High-Performance Defaults." The hard drives were then formatted, and Windows XP Professional (SP2) was installed. Then we installed all of the necessary drivers and removed Windows Messenger from the system. 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 drives, and ran all of the tests.
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). On the NVIDIA nForce 4 SLI Intel Edition reference board, we clocked the board's system memory at DDR2-533 speeds for one set of scores and DDR2-667 for another. Intel's i925XE (DDR2-533) represent the third set of scores.
The nForce 4 SLI Intel Edition's preliminary performance scores using SiSoft SANDRA looked quite good. As was expected, when using the same processor, clocked at the exact same speed, the CPU Arithmetic benchmark and CPU Multi-Media benchmark didn't show much variation from platform to platform. The nForce 4 SLI Intel Edition and i925XE performed similarly. SANDRA's Memory Benchmark, however, had the nForce 4 SLI Intel Edition well ahead of the i925XE, regardless of what speed we clocked the system memory. With the NF4 and i925XE both utilizing the same RAM, clocked at the same 533MHz (with the same timings), the NF4 had an approximate 300MB/s bandwidth advantage. And with the nForce 4's memory clocked up to 667MHz, its overall bandwidth scores were about 600MB/s higher than the i925XE.
|PCMark04: CPU & Memory Modules|
For our next round of synthetic benchmarks, we ran the CPU and Memory performance modules built into Futuremark's PCMark04. For those interested in more than just the graphs, we've got a couple of quotes directly from Futuremark that explain exactly what these tests do, and how they work.
"The CPU test suite is a collection of tests that are run to isolate the performance of the CPU. There are nine tests in all. Two pairs of tests are run multithreaded - each test in the pair is run in its own thread. The remaining five tests are run single threaded. These tests include such functions as file encryption, decryption, compression and decompression, grammar check, audio conversion, WMV and DivX video compression."
From this point forward, we've included some benchmark scores from an NVIDIA nForce 4 SLI based system, powered by an Athlon 64 4000+ for reference.
PCMark04's CPU performance module had all of the Intel powered systems finishing within 1 percent of one another, which is to be expected. And because this benchmark is influenced by the higher-clock speed and HyperThreading capabilities of Intel's processors, the Pentium 4 powered systems easily outpaced the Athlon 64 4000+ here.
"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."
Testing with PCMark04's memory performance module corroborated the results found with SANDRA's memory bandwidth benchmark. In this test, the nForce 4 SLI Intel Edition took first and second place. With the NF4 Intel Edition's memory clocked at 667MHz, it outpaced the i925XE by about 12%, and with its memory clocked at the same 533MHz the NF4 SLI Intel Edition still beat i925XE by a sizable margin. At 533MHz, the NF4 SLI finished this test with a score roughly 8.2% higher than the i925XE.
|Content Creation Winstone & WorldBench 5|
For this next batch of results, we used Veritest's demanding Content Creation Winstone 2004 suite. Before running this test, we patched the program to its latest version, shut-down any unnecessary background processes, and defragged the hard drive.
The Veritest Content Creation Winstone 2004 test utilizes the following applications in its benchmark routine, for more information about this test, see this page:
The NVIDIA nForce 4 SLI Intel Edition performed well in the Content Creation Winstone 2004 benchmark, outpacing the i925XE by a small margin. The Athlon 64 powered system was faster than all of the Intel based systems, but thanks to some fast RAM, the Intel machines were close behind.
PC World Magazine's WorldBench 5.0 is a new breed of Business and Professional application benchmark, that we're looking to with promise here on HotHardware. 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.
WorldBench 5.0's Photoshop 7 and Office XP SP2 performance modules also had the NVIDIA nForce 4 SLI Intel Edition besting the i925XE based system by a couple of seconds. In both the Photoshop and Office XP SP2 tests, running the system memory at 667MHz gave the nForce 4 SLI Intel Edition marginal advantages of 4 and 2 seconds in the respective tests. With NF4's memory clocked down to 533MHz, it was still faster than the i925XE in both tests, but this time by margins of only 2 seconds and 1 second respectively.
|Windows Media Encoder 9 & LAME MP3|
We continued our testing of the NVIDIA nForce 4 SLI Intel Edition reference motherboard with a video encoding benchmark using Windows Media Encoder 9. In this test, we ran the Windows Media Encoder 9 portion of the WorldBench 5 suite; encoding times were recorded in seconds. Lower times indicate better performance.
The nForce 4 SLI Intel Edition continued its winning ways in the Windows Media encoder benchmark, besting the i925XE based system by 4 seconds when the NF4's memory was clocked at 667MHz, and by 3 seconds when it was clocked at 533MHz. The Athlon 64 powered system lagged behind the Intel based systems by almost a full minute.
In our custom Lame MP3 encoding test, we convert a large 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 WAV file (a never-ending Grateful Dead jam) and converted it to the MP3 format. Processing times are recorded below. Shorter times equate to better performance.
Things continued to go the nForce 4 SLI Intel Edition's way in our custom LAME MP3 encoding benchmark. Here though, the nForce 4 Intel Edition was only 1 second faster than the i925XE when running the RAM at 667MHz, and it tied the i925XE when both system had their system memory clocked at 533MHz.
|KribiBench v1.1: Sponge Explode & Ultra Models|
Next up, we ran the Kribibench rendering benchmark produced by the folks at Adept Development. Kribibench is an SSE aware software renderer. A 3D model is rendered and 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 an absolutely gargantuan "Ultra" model that is comprised of over 16 billion polys...
Our KribiBench results went in favor of the i925XE based system in one test and in favor of the NF4 in another, but again, the performance deltas separating the platforms were quite small. When animating the Sponge Explode model, the i925XE beat the nForce 4 SLI Intel Edition by only 1 / 100th of frame per second. And when animating the Ultra model, the delta separating the platforms shrunk to only 1 / 1000th of a frame per second. We think it's safe to call this one a tie.
|Cinebench 2003 & 3DMark05: CPU|
The Cinebench 2003 benchmark is an OpenGL 3D rendering performance test, based on the commercially available Cinema 4D application. 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 ran two sets of numbers, one in single-thread mode, and another in the benchmark's multithread mode for our Hyper-Threading-enabled P4 test systems. Athlon 64s are only capable of running the single thread test, hence the "WNR" in the graph below.
Like the KribiBench results, the Cinebench 2003 results in both the single- and multi-threaded tests can by considered a tie between the Intel powered machines. The Athlon 64 4000+ came in with the fastest score in the single-thread test, but in the multi-thread test the Intel powered machines dominated.
It may not be an actual game, but 3DMark05's built-in CPU test is a "gaming related" DirectX metric that's useful for comparing relative performance among 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.
The nForce 4 SLI Intel Edition was back on top in 3DMark05's CPU performance test, where it outpaced the i925XE by 236 points (4.2%) when coupled with 667MHz RAM, and by 110 points (2%) with its RAM clocked at 533MHz. This test is influenced by the Pentium 4's HyperThreading capabilities as well, and as such the A64 based system trailed behind by a significant margin.
|Unreal Tournament 2004 & Doom 3|
To start our actual in-game testing, we did some low-resolution benchmarking with Epic's Unreal Tournament 2004. When testing with UT 2004, we use a specific set of game engine initialization parameters that ensure all of the systems are being benchmarked with the exact same in-game settings and graphical options. Like the other in-game tests in this review you'll see later, we used a "Low-Quality" setting with UT2004 that isolated CPU and memory performance.
The NVIDIA nForce 4 SLI Intel Edition also performed well in our custom Unreal Tournament 2004 benchmark. In this test, the NF4 / DDR2-667 combo was roughly 5% faster than the i925XE. When both platforms had their system memory running at DDR-533, the i925XE closed the gap a bit, but the NF4 SLI Intel Edition still came in a couple of frames per second ahead.
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's more CPU / Memory-bound than anything else...
We had a bit of an anomaly with our custom Doom 3 benchmark. Here, the nForce SLI Intel Edition was fastest when its system memory was running at 533MHz where it outpaced all of the other machines by a couple of frames per second. The margin of victory was quite small though, and falls well within the margin of error in this test.
|SLI Tests: 3DMark05 & Halo v1.06|
Before we wrapped up our testing, we ran 3DMark05 and a couple of popular games with the NF4 SLI boards (both Intel and AMD Editions) outfitted with a pair of GeForce 6800 GTs configured to operate in SLI mode. Please note, that in all of the previous tests, we were using a single GeForce 6800 GT.
% Improvement Intel = 63.6% | % Improvement AMD = 80.1%
As you'll see in the coming pages, we had some interesting results when comparing AMD NF4 SLI versus Intel NF4 SLI. When running 3DMark05's default test, we saw large performance improvements on both sides, but the AMD rig was faster all around; besting the nForce 4 SLI Intel edition in both actual score and percent improvement.
% Improvement Intel = 59.2% | % Improvement AMD = 55.9%
Things balanced out somewhat when we moved onto Halo for our SLI testing. This time around, both platforms showed roughly the same improvement, give or take a couple of percentage points. And once again, the AMD side proved why its currently the top choice for gamers with its roughly 10% higher framerates.
|SLI Tests: FarCry v1.31 & Doom 3|
% Improvement Intel = 51.5% | % Improvement AMD = 40.9%
With our custom FarCry benchmark, the nForce 4 SLI AMD Edition finished the test with the highest framerates in both the single-GPU and SLI tests, but the nForce 4 SLI Intel Edition posted a larger performance improvement.
% Improvement Intel = 80.7% | % Improvement AMD = 78.1%
Our custom multi-player Doom 3 demo ran equally well on both nForce 4 SLI platforms. The AMD based system posted marginally higher frame rates when configured to run in either SLI mode or single-GPU mode, but again the nForce 4 SLI Intel Edition gained more performance by adding in the second 6800 GT.
|Performance Summary & Conclusion|
Performance Summary: NVIDIA's nForce 4 SLI Intel Edition performed very well throughout our battery of benchmarks. Whether our nForce 4 SLI Intel Edition's system memory was clocked at DDR2-667 or DDR2-533 speeds, it performed on-par with, or slightly better than, a similarly equipped system that was outfitted with an enthusiast-class motherboard based on Intel's i925XE chipset.
Although Intel would obviously love to sell as many of their own chipsets as possible, we're sure there are many folks in Santa Clara that are pleased to see the release of NVIDIA's nForce 4 SLI Intel Edition. With this chipset, NVIDIA has made the Intel platform far more attractive to hardcore gamers and enthusiasts who were contemplating a purchase of an SLI setup. Of course, our gaming scores have shown that AMD's processors are superior in most gaming scenarios, but there are many situations where an Intel powered system would be a better fit. If you're the type of user who splits their computing workload equally between gaming, video / audio encoding, and general use (browsing, word processing, etc.), an Intel powered system makes sense, even more-so if you dabble in 3D modeling as well. And users who constantly have multiple applications open, and perform a few CPU intensive tasks simultaneously, like burning a DVD, while running Photoshop, and checking e-mail and chatting, for example, will appreciate the "smoothness" that comes with a HyperThreading enabled rig.
The bottom line is that the NVIDIA nForce 4 SLI Intel Edition, is a feature rich, high-performance chipset that further ads to the appeal of an Intel based system. In this initial look at the features and performance of the nForce 4 SLI Intel Edition, we were thoroughly impressed, especially considering all of our testing was done with a pre-production reference motherboard. Top-tier motherboard manufacturers are generally able to tweak performance even further with "retail-ready" products. With that in mind, we're looking forward to see what companies like Asus, Abit, MSI, and DFI, among others, have in store for this chipset.