It has been six months since NVIDIA announced their new 500 series chipsets. This past May the launch coincided with the release of AM2. Today NVIDIA launches a new chipset family, called the 600i family, with no mention at all of AMD and a launch date to coincide with the new Intel Core 2 Quad (Kentsfield).

Perhaps these two events, set just six months apart, best define the dramatic shifts that have occurred in the enthusiast market during this time. AMD was undisputed performance leader for the past couple of years, and enthusiasts didn't much care about Intel chipsets. With the launch of Core 2, however, the enthusiast world changed. Today Core 2 Duo and Quad are the undisputed performance leaders and AMD is once again the "value" chip. This will likely change again in the future, but for now Intel Core 2 is clearly the processor enthusiasts are demanding.

Of course, that has been the problem for NVIDIA. Where their 590/570/550 family was just great with AMD processors, their Intel variants left a lot to be desired. NVIDIA is a company that proclaims loudly its support of the enthusiast, and it had to be embarrassing that the NVIDIA chipsets for Intel were also the worst overclocking chipsets in the market. NVIDIA needs credibility as a provider of enthusiast chipsets in order to sell their top-end SLI to Intel buyers, since Intel has supported the competitor's ATI CrossFire as their multi-GPU standard. Features of nForce 590 looked great, but the overclocking performance, or rather the lack of it, kept enthusiasts away from the 500 series for Intel.

In addition, in the past 6 months AMD bought ATI, NVIDIA's major competitor in graphics. NVIDIA had become the leading supplier of chipsets for motherboards supporting the AMD processor, and with ATI moving to AMD that market position was now in jeopardy. ATI also had competent chipsets for AM2, and everyone expected AMD to make good use of those capabilities in the future.

What had been the minor annoyance of not having a good enthusiast chipset for Intel's Core 2 Duo quickly became a major problem for NVIDIA. The enthusiast was now buying Intel processors instead of AMD, their major competitor was now part of their largest customer in the chipset market, and the world's largest supplier of chipsets for Core 2 Duo - Intel themselves - was supporting the ATI CrossFire multi-GPU solution. NVIDIA needed a new product for the Intel Socket 775 that would excite the enthusiast enough to buy NVIDIA for Intel, increase NVIDIA's market share in the Intel chipset market, and provide a superior platform for SLI on Intel.

That product launches today in the NVIDIA 600i chipset family. The "i" is for Intel, and for now the 600 family is only available for the Intel Socket 775. (Future NVIDIA chipsets for the AMD platform will be named with a small "a" following the number.) The family will include some value boards and a top-end 680i that claims incredible overclocking on Core 2 Duo and Core 2 Quad processors. The new chipset also delivers dual x16 SLI to the Intel platform in a board NVIDIA is confident enthusiasts will want to own.

NVIDIA cut their teeth in the AMD market, but the Intel chipset market is a much more ambitious target. In the past AMD was only a minor player in the AMD chipset market, but Intel is the largest supplier of chipsets for their own socket 775 processors. Intel also has a long and impressive history of innovations in the chipset market. Intel chipsets are widely regarded as the top performers in almost any category supporting Intel processors. This is a very different market than the AMD platform NVIDIA targeted and conquered. With ATI now part of AMD, with current AMD chipsets moving toward the value category, and with the enthusiast buying Intel processors, the desire to target the Intel market is logical. However, as NVIDIA quickly found out with the 500 family for Intel, they must have the goods to persuade buyers to choose NVIDIA instead of Intel.

The real question then is whether the 680i and the 600i chipset family are the best available in the Intel market. If we believe NVIDIA marketing the answer is a resounding yes. Does the 680i live up to all the advance hype? We hope to provide answers to that question.



Features: nForce 600i Platform

While our testing will concentrate on the EVGA 680i motherboard designed by NVIDIA, it is important to point out that 680i is a member of a whole new family of NVIDIA chipsets for the Intel Socket 775 platform.

NVIDIA Intel Chipsets
Market Segment	Chipset	Price
Hard-Core Enthusiast	nForce 680i SLI	$249-$299
Performance Gamers	nForce 650i SLI	$149-$199
Mainstream Gamers	nForce 650i Ultra	$99-$149

Across the line NVIDIA emphasizes that buyers can expect spectacular overclocking with their new 600i Series motherboards. Other features are basically a refinement of the feature set introduced with the NVIDIA 550/570/590 series this past summer. These include DualNet, Massive RAID 5, FirstPacket, and MediaShield which will be discussed in more detail in the features section. The 600i family also fully supports Intel Core 2 Extreme (dual & quad), Core 2 Quad, and Core 2 Duo processors.

NVIDIA nForce 600i Family Specifications




nForce 680i Platform

At the top of the product line, and targeted at the "hard-core overclocker", is the nForce 680 SLI. It consists of two discrete chips, the 680i MCP and the 680i SPP. As we saw in the 590 chipset, one x16 PCIe slot is controlled by each chip. This provides dual x16 SLI slots for the 680i, similar to what was provided by the 590 chipset.


There is a significant new addition with 680i, however, in a third PCIe x16 slot between the x16 PCIe slots for SLI. This third slot is x8 PCIe instead of x16, but it can be used for a GPU on the 680i, with a new twist. This x8 slot is suitable for a physics processor combined with SLI.


If you prefer multiple monitors instead, you can even drive up to six monitors - two per card - with video cards in all 3 PCIe slots.


The nForce 680i provides a total of 46 PCIe lanes, with 18 lanes available on the 680i SPP and 28 lanes provided on the 680i MCP.

New features of the 680i are the addition of Extreme Overclocking capabilities and basic refinements of the features first introduced with the 590 SLI chipset. Features include:

Extreme Overclocking

• FSB speeds of 1333 MHz can be achieved with a CPU that supports this specification.


• DDR2 speeds of DDR2-1200 and beyond can be used to keep pace with overclocked system components. NVIDIA nForce 680i SLI MCPs support high-speed SLI-Ready memory

NVIDIA nTune^TM Utility allows adjustment of CPU and memory speeds without rebooting. There is also access to most BIOS settings from inside Windows.

True 2 x16 PCI Express SLI Support provided by two full-bandwidth, 16-lane PCI Express links.

Third PCIe Slot for Graphics Expansion

NVIDIA MediaShield^TM Storage includes RAID and SATA drive support.

• Multiple Disk Setup Through a wizard-based interface, you can set up your drives for better data protection, faster disk access or maximum storage capacity. MediaShield automatically selects a RAID 0, 1, 0+1 or 5 configuration according to your needs. Advanced users can access RAID options directly.
• DiskAlert System In the event of a disk failure, MediaShield users see an image that highlights which disk has failed to make it easier to identify, replace, and recover.
• RAID Morphing MediaShield allows users to change their current RAID setup to another configuration in a one-step process called morphing. This eliminates the need to back up data and follow multiple steps in the process.
• Bootable Multidisk Array MediaShield storage fully supports the use of a multi-disk array for loading the operating system at power-up.
• Six SATA 3Gb/s Drives Combine up to six SATA drives into one volume for bigger, faster RAID. Drives can be configured as six RAID 0 (striped) drives for maximum throughput, or Single or Dual RAID 5 arrays. Hot plug and Native Command Queuing are supported.

Networking with Dual Gigabit Ethernet with Teaming allows the two connections to work together to provide up to twice the Ethernet bandwidth.

High Definition Audio (HDA) can deliver 192 kHz/32-bit quality for eight channels.

USB 2.0 plug-and-play interface that provides easy-to-use connectivity for up to 10 USB devices.



nForce 650i SLI & 650i Ultra

For buyers not interested in the top-end features or top end price of the 680i, NVIDIA will offer two additional models at lower price points.


The 650i SLI, which will sell in the $150 to $199 price range, will support a single x16 PCIe video card, or SLI with two video cards driven by two x8 PCIe slots. PCIe lanes are reduced to 18, SATA-II ports from 6 to 4, USB 2.0 from 10 to 8, and it comes with a single Gigabit Ethernet port instead of the two with teaming on the 680i.


The nForce 650i Ultra, designed to sell in the $99 to $149 price range, offers a single x16 PCIe slot in 18 total PCIe lanes. Features are otherwise the same as the 650i SLI.

While there is some overlap in market segments, the 680i is designed to compete with the Intel 975X premium motherboards while providing dual x16 SLI graphics compared to dual x8 CrossFire on the 975x. The 650i SLI should compete with premium Intel 965 boards, while the 650i Ultra will compete with low-end or entry-level P965 motherboards.



LinkBoost

One of the features unique to the nForce 590SLI and 680i SLI MCP is a system called LinkBoost. If a GeForce 7900 GTX or GeForce 8800 is detected on either MCP then LinkBoost will automatically increase the PCI Express and MCP HyperTransport (HT) bus speeds by 25%. This increases the bandwidth available to each PCI Express and HT bus link from 8GB/s to 10GB/s.

Since this technology increases the clock speed of the PCI Express bus by 25% to the x16 PCI Express graphics slots, NVIDIA requires certification of the video card for this program to work automatically. In this case, the 7900GTX and 8800 series are the only compatible cards offered, although you can manually set the bus speeds and achieve the same results depending upon your components. We feel this feature is worthwhile for those users who do not want to tune their BIOS and go through extensive test routines to find the best possible combination of settings.


In essence, NVIDIA is guaranteeing their chipset's PCI Express and HT interconnect links are qualified to perform up to 125% of their default speeds without issue. While LinkBoost is an interesting idea, the 25% increase in PCI Express x16 slots and HT bus speeds yielded virtually the same performance as our system without LinkBoost enabled in most cases.

Its actual implementation did not change our test scores in single video card testing but did provide a 1%~2% difference in SLI testing at resolutions under 1600x1200 in several game titles. The reason for the minimal increases at best is that the performance boost is being applied in areas that have minimal impact on system performance as the link to the CPU/Memory subsystem is left at stock speed thus negating the true benefits of this technology.

FirstPacket

As part of the overhaul of the networking features first introduced in the NVIDIA nForce 590SLI and now 680i SLI Series, FirstPacket is a packet prioritization technology that allows latency-sensitive applications and games to effectively share the upstream bandwidth of their broadband connection. Essentially this technology allows the user to set network data packets for applications and games that are more latency sensitive with a higher queue priority for outbound traffic only.

FirstPacket is embedded in the hardware and offers driver support that is specifically designed to reduce latency for networked games and other latency-sensitive traffic like Voice over IP (VoIP). When network traffic constrains a connection, latency is increased which in turn can result in dropped packets that would create a jitter and delay in VoIP connections or higher ping rates to the game server resulting in stutters and decreased game play abilities.


In the typical PC configuration, the operation system, network hardware, and driver software are unaware of latency issues and therefore are unable to reduce it. The standard interfaces that allow applications to send and receive data are basically identical to the OS in a typical system. This type of design results in latency-tolerant and large packet applications like FTP or Web browsers filling the outbound pipeline without regards to the needs of small packet and very latency-sensitive applications like games or VoIP applications.


FirstPacket operates by creating an additional transmit queue in the network driver. This queue is designed to provide expedited packet transmission for applications the user determines are latency-sensitive applications. The ability of the designated applications to get preferential access to the upstream bandwidth usually results in improved performance and lower ping rates. The FirstPacket setup and configuration is available through a revised Windows based driver control panel that is very easy to use.

In our LAN testing, we witnessed ping rate performance improvements of 22% to 36% during the streaming of video from our media server while playing Serious Sam II across three machines on our LAN. We noticed ping rate performance improvements of 14% to 33% while uploading files via BitTorrent and playing Battlefield 2 on varying servers.

The drawback at this time is that only outbound packets are prioritized so if you spend more time downloading than uploading the FirstPacket technology will have little impact on your computing experience. Worth mention is that nearly all broadband connections have a lot more downstream bandwidth than upstream bandwidth, so focusing on prioritizing outbound traffic does make sense. Also, the upload time for our test file increased by 41% with FirstPacket turned on but the overall gaming experience was significantly better. However, in NVIDIA's defense they cannot control the behavior or quality of service on other networked clients, so FirstPacket addresses the services NVIDIA can control - namely uploading.



DualNet

DualNet's suite of options actually brings a few enterprise type network technologies to the general desktop such as teaming, load balancing, and fail-over along with hardware based TCP/IP acceleration. Teaming will double the network link by combining the two integrated Gigabit Ethernet ports into a single 2-Gigabit Ethernet connection. This brings the user improved link speeds while providing fail-over redundancy. TCP/IP acceleration reduces CPU utilization rates by offloading CPU-intensive packet processing tasks to hardware using a dedicated processor for accelerating traffic processing combined with optimized driver support.

While all of this sounds impressive, the actual impact for the general computer user is minimal. On the other hand, a user setting up a game server/client for a LAN party or implementing a home gateway machine will find these options very valuable. Overall, features like DualNet are better suited for the server and workstation market. We believe these options are being provided (we are not complaining) since the NVIDIA professional workstation/server chipsets are based upon the same core logic.


NVIDIA now integrates dual Gigabit Ethernet MACs using the same physical chip. This allows the two Gigabit Ethernet ports to be used individually or combined depending on the needs of the user. The previous NF4 boards offered the single Gigabit Ethernet MAC interface with motherboard suppliers having the option to add an additional Gigabit port via an external controller chip. This too often resulted in two different driver sets, with various controller chips residing on either the PCI Express or PCI bus and typically worse performance than a well-implemented dual-PCIe Gigabit Ethernet solution. .

Teaming


Teaming allows both of the Gigabit Ethernet ports in NVIDIA DualNet configurations to be used in parallel to set up a 2-Gigabit Ethernet backbone. Multiple computers can to be connected simultaneously at full gigabit speeds while load balancing the resulting traffic. When Teaming is enabled, the gigabit links within the team maintain their own dedicated MAC address while the combined team shares a single IP address.

Transmit load balancing uses the destination (client) IP address to assign outbound traffic to a particular gigabit connection within a team. When data transmission is required, the network driver uses this assignment to determine which gigabit connection will act as the transmission medium. This ensures that all connections are balanced across all the gigabit links in the team. If at any point one of the links is not being utilized, the algorithm dynamically adjusts the connection to ensure an optimal connection. Receive load balancing uses a connection steering method to distribute inbound traffic between the two gigabit links in the team. When the gigabit ports are connected to different servers, the inbound traffic is distributed between the links in the team.


The integrated fail-over technology ensures that if one link goes down, traffic is instantly and automatically redirected to the remaining link. If a file is being downloaded as an example, the download will continue without loss of packet or corruption of data. Once the lost link has been restored, the grouping is re-established and traffic begins to transmit on the restored link.

NVIDIA quotes on average a 40% performance improvement in throughput can be realized when using teaming although this number can go higher. In their multi-client demonstration, NVIDIA was able to achieve a 70% improvement in throughput utilizing six client machines. In our own internal test we realized about a 36% improvement in throughput utilizing our video streaming benchmark while playing Serious Sam II across three client machines. For those without a Gigabit network, DualNet has the capability to team two 10/100 Fast Ethernet connections. Once again, this is a feature set that few desktop users will truly be able to exploit at the current time. However, we commend NVIDIA for forward thinking in this area as we see this type of technology being useful in the near future.

TCP/IP Acceleration

NVIDIA TCP/IP Acceleration is a networking solution that includes both a dedicated processor for accelerating networking traffic processing and optimized drivers. The current nForce 590SLI and nForce 680i SLI MCP chipsets have TCP/IP acceleration and hardware offload capability built in to both native Gigabit Ethernet Controllers. This capability will typically lower the CPU utilization rate when processing network data at gigabit speeds.


In software solutions, the CPU is responsible for processing all aspects of the TCP protocol: Checksumming, ACK processing, and connection lookup. Depending upon network traffic and the types of data packets being transmitted this can place a significant load upon the CPU. In the above example all packet data is processed and then checksummed inside the MCP instead of being moved to the CPU for software-based processing that improves overall throughout and CPU utilization.

NVIDIA dropped the ActiveArmor slogan for the nForce 500 release and it is no different for the nForce 600i series. Thankfully the ActiveArmor firewall application was jettisoned to deep space as NVIDIA pointed out that the basic features provided by ActiveArmor will be a part of Microsoft Vista. We also feel NVIDIA was influenced to drop ActiveArmor due to the reported data corruption issues with the nForce4 caused in part by overly aggressive CPU utilization settings, customer support headaches, issues with Microsoft, and quite possibly hardware "flaws" in the original nForce MCP design.

We have found a higher degree of stability with the new TCP/IP acceleration design but this stability comes at a price. If TCP/IP acceleration is enabled via the control panel, then certain network traffic will bypass third party firewall applications. We noticed CPU utilization rates near 14% with the TCP/IP offload engine enabled and rates near 26% without it.



MediaShield

While the networking side has undergone an extensive makeover, the storage side of the nForce 680i has been fine tuned but the features remain the same as the nForce 590SLI. The nForce 680i SLI series offers three separate SATA controllers each with integrated dual PHYs that are capable of operating at 1.5Gb/s or 3.0Gb/s speeds. This results in six Serial ATA devices being available for the user instead of four as in the nForce4, Intel ICH7/ICH8, or ATI SB600. These devices can be configured in RAID 0, 1, 0+1, and 5 arrays. There is no support for RAID 1+0 although the performance numbers with RAID 0+1 are very similar.


Considering the support for six drives, it is now possible to run a massive RAID 5 drive consisting of a pair of three-drive RAID 5 arrays, or running multiple combinations of RAID technology together. NVIDIA also supports the shared spare (or dedicated spare) technique in MediaShield. The spare disk feature, available with MediaShield RAID 5, offers protection with a dedicated spare drive that can take over for a failed disk until the repair is completed. However, the performance results during our RAID testing found no measurable differences between the nForce4, nForce 590SLI, and nForce 680i SLI storage systems. In fact, the "average" write performance of the nForce4 and nForce 590SLI in RAID 5 continues in the 680i SLI chipset although we still find its performance to be 5% better than the Intel ICH8R.

NVIDIA introduced a new twist to improve their SATA controller performance by offering performance profiles for specific hard drive models in the nForce 590SLI MCP. These performance profiles continue in the 680i SLI MCP. Since each hard drive has unique performance characteristics, NVIDIA matched the capabilities of their controller logic to each drive's particular strength. So far, Western Digital's 150GB Raptor has the only performance profile loaded, but there are plans to profile additional performance oriented drives that are popular in the market. In our testing with dual WD1500 Raptors we noticed benchmark results that were on average about 3% better in our IPEAK tests while the synthetic tests realized a 4% gain in some areas.

While NVIDIA has implemented six native SATA ports in their current MCP55 chipset, they reduced the available native PATA ports to one with support for two drives. This is an improvement over the PATA challenged Intel ICH8 series which requires a separate chipset for PATA support. We firmly believe the reduced number of PATA ports is still a mistake for all chipset manufacturers.

Considering the Optical drive manufacturers have been very slow to implement SATA technology in their drives, this decrease in drive support could affect those users who have multiple optical drives for audio/video content creation and manipulation. However, our discussions with the major optical drive manufacturers show an aggressive transition to SATA technology by the third quarter of 2007.

HD Audio

NVIDIA has finally decided that life after SoundStorm no longer means the continual punishment of users by only offering AC-97 based audio support in their chipsets. As with the nForce 500 lineup, the nForce 600 series will offer full support for the various "Azalia" based High Definition Audio codecs. While the choice of which HDA codec along with the associated circuitry can still greatly impact audio quality and performance, any of these options are better than the AC-97 solutions previously offered.



NVIDIA Control Panel & nTune 5

One of most interesting features at the nForce 500 launch was the revamped control panel and nTune 5.0 performance applications. The Control Panel and nTune have undergone some fine tuning the past few months with version 5.05 launching for the nForce 600 and GeForce 8800 series of products shortly. NVIDIA has combined all of their various program applets into an integrated control panel to provide a common user interface. We found that this common interface makes it easier for a user to control the various functions of the board and video (NVIDIA based) from a central access point.

Click to enlarge

The new control panel offers configuration sections for 3D Settings for NVIDIA GPUs, Display settings, Networking options, Performance, System Stability testing, Storage, and Video/Television settings. The information contained in each section will already be familiar to those with nForce4 or nForce 500 boards but will include a few extended configuration options for the nForce 600 and GeForce 8800 products.

Click to enlarge

One noteworthy change with the GeForce 8800 series is that Coolbits no longer works for overclocking the GPU. The overclocking panel will still be visible in the legacy control panel but it will tell you to download nTune in order to overclock your GPU. The new GPU overclocking utility works in the same manner but is now integrated into the revised Control Panel.

Click to enlarge

One of the more interesting features for us is the Adjust Motherboard Settings application. We found the ability to dynamically write a significant number of performance settings directly to the BIOS without requiring a reboot a very quick and easy way to test or set overclocking options without exiting Windows. There are several third party applications that perform this same function but we believe the official support within the BIOS and nTune software gives it a significant advantage over the other solutions. The settings can be saved in a profile that can be loaded within Windows without having to reset the BIOS for individualized situations where settings for games might differ from those for audio/video playback. The system also allows for automatic or direct fan control on supported fan headers.

Click to enlarge

The Dynamic BIOS Access section offers the ability to change BIOS setting in five different categories that will take effect on reboot. The two sections not displayed are for power management and peripheral settings.

Click to enlarge

If you are not the type to enter the BIOS and tune your system or set it up for overclocking then NVIDIA offers an application that can do it for you based on your combination of components. We found the Coarse Tuning option would generally set up our system for a 6%~9% overclock of the FSB and memory along with a 3%~5% overclock on the GPU. The Fine Tuning option resulted in an 8%~15% overclock of the FSB and memory while the GPU usually increased 5%~8%.

Click to enlarge

NVIDIA also includes a System Information utility along with a Windows based monitoring application. Overall, the polished look, feel, and operation of the new control panel along with the performance improvements offered in nTune 5.05 has raised the bar once again for user enhancements offered by the core logic suppliers.



Test Setup

NVIDIA designed the 680i chipset motherboard that is being released as the EVGA 680i SLI.

Click to enlarge

The 6-phase, 6-layer motherboard is passively cooled for normal operation. An accessory fan for the chipset is included for extreme overclocking. Since this is a review of the new 680i chipset there will not be in-depth comments on the board layout. However, readers should be aware of the horrible location of the front panel connectors in the middle right edge of the board. They are stacked in line on-top of the auxiliary 12V Molex and the IDE connector, with the memory slots on the other side.

This busy location makes it impossible to do much of anything in setup without dislodging the front panel LEDs and switches. The color code for the front panel connector is also wrong, and does not match any case setup we have tested. Color coding is a good idea but colors should match common setups. NVIDIA is aware of the issues with the location of the front panel connector and the color-coding and they have told us both issues will be fixed in a future revision of the motherboard.

The EVGA 680i SLI was used for all testing of the 680i chipset.

Performance Test Configuration
Processor:	Intel Core 2 Duo E6700 (X2, 2.67GHz, 4MB Unified Cache)
RAM:	2 x 1GB Corsair TCM2X1024-9136C5D Tested at DDR2-800 3-3-3 2.2V
Hard Drive(s):	Hitachi 250GB SATA2 enabled (16MB Buffer)
System Platform Drivers:	NVIDIA - 9.35
Video Cards:	1 x EVGA 7900GTX - All Standard Tests 2 x EVGA 7900GTX - SLI on NVIDIA 1 x ATI X1900XTX - ATI Standard Tests on Intel 2 x ATI X1900XT (Master+Standard) - CrossFire on Intel
Video Drivers:	NVIDIA 93.71 ATI Catalyst 6.10
CPU Cooling:	Tuniq Tower 120
Power Supply:	OCZ GameXstream 700W
Motherboards:	EVGA 680i SLI ASUS P5W-DH Deluxe (Intel 975X) Intel 975XBX (Intel 975X) ASUS P5B Deluxe (Intel P965) ASUS P5N32-SLI (nF4 SLIX16 Intel) Biostar TForce P965 Deluxe (Intel P965) Gigabyte GA-965P-DQ6 (Intel P965) DFI Infinity 975X/G (Intel 975X) ASRock 775Dual-VSTA (VIA PT880 PRO)
Operating System:	Windows XP Professional SP2

The AnandTech launch article for the NVIDIA 8800 GPUs provides test results with the 680i, 8800, and Core 2 Duo and Quad processors. This chipset review in contrast concentrates on comparing performance with our standard setup of the E6700, 2GB of DDR2 running DDR2-800 3-3-3 timings, and the NVIDIA 7900GTX to other tested Socket 775 Core 2 motherboards.



Memory Performance

The AMD AM2 has essentially removed concern about the Memory performance of chipsets by providing a low-latency memory controller directly on the processor. Intel continues to integrate the memory controller in their chipsets, and the Intel Core 2 chipsets achieve competitive memory performance and low latency with caching schemes and read-ahead algorithms. Because Intel has done such an excellent job of providing stellar memory performance in the 975x/P95 chipsets, any chipset that hopes to compete with Intel would have to perform similarly in memory controller performance.

To assess the effectiveness of the NVIDIA 680i chipset we measured memory performance with Everest 3.50 from Lavalys and compared results to the top-end Intel 975X chipset.

Everest 3.5 Memory Performance
Chipset	Read	Write	Copy	Latency
NVIDIA 680i	8045	4865	5506	54.8ns
Intel 975x	7751	4868	5512	52.3ns

The results of memory performance on the NVIDIA 680i are truly impressive compared to the 975X. NVIDIA has clearly produced a competitive memory controller for Intel Socket 775, a task which has eluded other chip makers who have tried to compete on this platform. Results can only be called equivalent in any of the Everest benchmarks.

Another widely used measure of latency and memory bandwidth is ScienceMark 2.0.

ScienceMark 2.0 Memory Performance
Chipset	Latency (512 byte stride)	Memory Bandwidth
NVIDIA 680i	37.12ns	5449.37 MB/s
Intel 975x	37.81ns	5430.69 MB/s

ScienceMark confirms the results with Everest 3.5. NVIDIA has built a memory controller in the new 680i chipset that is every bit as good as the outstanding Intel memory controller in the 975X. This is not a minor achievement.

The final confirmation of memory controller performance with the 680i chipset comes with SiSoft Sandra 2007. The most common measurement is buffered or Standard Memory Bandwidth.

Sandra 2007 Standard Memory Performance
Chipset	INT (Integer)	FLT (Float)
NVIDIA 680i	5969	5944
Intel 975x	5949	5953

General System Performance

The PCMark05 benchmark developed and provided by Futuremark was designed for determining overall system performance for the typical home computing user. This tool provides both system and component level benchmarking results utilizing subsets of real world applications or programs. The test is useful for providing comparative results across a broad array of graphics subsystems, CPU, hard disk, and memory configurations along with multithreading results. In this sense we consider the PCMark benchmark to be both synthetic and real world in nature while providing consistency in our benchmark results.


The 680i leads the PCMark 2005 results, but scores are close at the top. While the 975X/P965 and 680i are close in performance, the 680i provides a slight edge in General Performance - about 1% which can be considered negligible.

General Graphics Performance

The 3DMark series of benchmarks developed and provided by Futuremark are among the most widely used tools for benchmark reporting and comparisons. Although the benchmarks are very useful for providing apple to apple comparisons across a broad array of GPU and CPU configurations they are not a substitute for actual application and gaming benchmarks. In this sense we consider the 3DMark benchmarks to be purely synthetic in nature but still valuable for providing consistent measurements of performance.


Our nForce 680i SLI results are essentially the same as the performance of the 590 and Intel 975x/965 families. Since the most recent 3DMark06 is very tied to the GPU, this does not come as much of a surprise.

Rendering Performance

The Cinebench 9.5 benchmark heavily stresses the CPU subsystem while performing graphics modeling and rendering. We utilize the standard benchmark demos along with the default settings. Cinebench 9.5 features two different benchmarks with one test utilizing a single core and the second test showcasing the power of multiple cores in rendering the benchmark image.



The nForce 600i desktop platform shows competitive performance in these benchmarks. The range of results are not performance differences you will see looking at the daily performance of 975X and 680i systems.



Gaming Performance






The overall gaming performance of the 680i platform is excellent and is competitive with or faster than any Intel Socket 775 motherboard tested at AT. These results confirm the 680i is an excellent gaming platform for Core 2 processors.

Dual GPU Performance

With the Intel 975X chipset supporting CrossFire only and the review/benchmarks of the NVIDIA 8800 GPU also posting today, there was little point in running benchmarks with the 7900 GTX SLI with nothing to really compare to in performance. For more details on performance of the NVIDIA 680i with the NVIDIA 7900 GTX and 8800 series GPUs, please refer to the 8800 launch review that was also posted today.



nForce 680i Overclocking

The defining area for the 680i chipset has to be overclocking. The 500 family chipsets for Intel performed fine in many areas, but they were one of the worst chipsets available for Intel in overclocking, topping out somewhere below 350 MHz on the bus speeds. This area kept many Enthusiasts from embracing the NVIDIA SLI solution for their Core 2 processors. It appeared NVIDIA had designed the 500 chipset family for spectacular overclocking on the 200 MHz HTT AMD AM2 chipset. The problem, of course, is that Core 2 processors were released at 1067 FSB (266 base bus speed), and the 500 family chipsets did not have anywhere close to the headroom of the competing Intel chipsets.

All of that has changed, NVIDIA tells us, with a 680i that can do 1333 FSB and more and DDR2 memory speeds of 1200 and beyond. To test NVIDIA's claims we tested the EVGA 680i SLI using the following setup:

EVGA 680i SLI Overclocking Testbed
Processor:	Intel Core 2 Extreme X6800 Dual Core, 2.93GHz, 4MB Unified Cache 1066FSB, 11x Multiplier
CPU Voltage:	Default (Auto) to 1.575V
Cooling:	Tuniq Tower 120 Air Cooling
Power Supply:	OCZ GameXstream 700W
Memory:	Corsair Dominator 1142MHz (2x1GB) (Micron Memory Chips)
Hard Drive	Hitachi 250GB 7200RPM SATA2 16MB Cache
Maximum OC: (Standard Ratio)	363 (1452 FSB) x 11 - 3.993GHz at 1.575V (+36.3%) 334 (1336 FSB) x 11 - 3.675GHz at Default Voltage
Maximum OC: (Reduced Multiplier)	525 (2100 FSB) x 7 - 3.675GHz at Default Voltage

The overclock results are not a mistake. We managed to reach right at 4.0GHz with an X6800 at 1.575V. This represented the standard 11x ratio at a 1452 FSB. Even more spectacular was the reduced multiplier overclocking, where we reached a 2100 FSB (525 base) at a 7X multiplier AT DEFAULT VOLTAGE.



These results were with air cooling using a Tuniq 120 air cooler, which sandwiches a 120mm fan in a finned heatpipe core as pictured above.

Click to enlarge

One of our X6800 processors reached 2100 FSB, confirming NVIDIA's claims of overclocking an X6800 in their labs to 2070 FSB. Not only did we reach 2100 FSB (525) at a 7X multiplier (3.675GHz), we managed to reach that speed at default voltage. The system has run at those settings for several days without incident and has handled every test and benchmark we have thrown at the system. However, it should also be pointed out that a second X6800 CPU would not overclock 1 MHz higher than 1900 FSB (450) on this same motherboard, even though that X6800 reached a similar 4GHz maximum overclock and similar "default voltage" overclocks.

We have yet to find anything obvious that would explain these differences in maximum FSB in the X6800 chips we tested. Revision and Stepping explained nothing, and we also had four retail E6600 we tested that would not reach above 1800 FSB. We are looking into these findings further and hope to find some explanations to share with you that will explain these maximum FSB differences among 4MB Cache Core 2 chips.

Click to enlarge

Interestingly, all of the 2MB cache Core 2 processors we tested reached at least 2000 FSB, with the two tested E6300 reaching 2100 FSB (522x7 and 6x525). We were prepared to call the amount of cache the defining difference in FSB performance until the late X6800 reached 2100 in our labs and NVIDIA confirmed their own 2070 results with an X6800 Extreme. We are much more confident that 2MB Cache chips can top 2000 FSB and we are very anxious to find something that will help you identify 4MB cache chips that will reach the "magic" 2000, 2070, and 2100 FSB overclocks.

Overall overclocking performance with the 680i was spectacular, and the results should excite any of you who want to do some serious overclocking with a Core 2 processor. At the very least the nForce 600i chipset family gives shoppers new choices in motherboards that will make the most of their Core 2 processor.



Final Words

Russian novelists used to be compensated by the weight of their manuscripts, which is one reason why Tolstoy's War and Peace is so incredibly long. If we weigh NVIDIA's launch by the reams of marketing information and public relations materials from the company, the nForce 600i product family is an impressive product release. We said something similar in our review of the launch of the nForce 500 chipset family just 6 short months ago, and then concluded that the release was more evolutionary than revolutionary. Is this a similar case of much ado about less than you first think?

There are many innovations, reintroductions, and refinements with the new NVIDIA 600i family chipsets, and this review would have been hundreds of pages long had we explored all of the features in detail. No matter the marketing spin or the positive light on innovation, the bottom line is the answer to the question we asked in the beginning of this article. Is the nForce 680i the best chipset for the Intel Core 2?

This time around our answer has to be a resounding YES. NVIDIA has finally got it right on the Intel platform and the 680i is definitely worthy of being called the Hard-Core Enthusiast chipset. We still have questions about why some Core 2 Duo processors overclock to 525 FSB and others can barely reach 450 FSB on this chipset, but we have seen similar behavior with the same CPUs on Intel P965 platforms. Since we were able to personally test an X6800 Extreme 4MB Cache chip at 2100 FSB (525 quad pumped), we lean toward NVIDIA's explanation that different processors are capable of different maximum FSB, regardless of their base overclocking capabilities.

We found one X6800 which maxed out at 1900 FSB, and another that reached 2100 FSB, even though both overclocked with the right combination of multiplier and FSB to around 4GHz. There was no easy answer of the Revision or Stepping to explain these differences. It was also interesting that every 2MB Cache chip we tested reached over 2000 FSB, with our two E6300 reaching 2100 FSB.

The point is we can't tell you, as much as we would like to, why some Core 2 processors reach above 2000 FSB and why others won't go beyond 1800 at any multiplier - no matter how low. We wish we could, and when we do find answers to the question of which processors will do the magic 2000+ we will share them with you.

While we can't answer the processor issue, we are confident that the seriously reworked 680i chipset can take you wherever you want to go as an Intel Core 2 enthusiast. If you want a fast system that does most of the thinking for you, you have only to buy a 680i and SLI certified components. You can let the hardware make all your decisions for you and overclock the video card and memory. This can happen even at stock speed, or you can also overclock the CPU. For those "wannabee" enthusiasts this will be very appealing.

Hard Core enthusiasts often want to do it all themselves, and you can also do this with the 680i. In our own benchmarking we were able to turn off all the automatic overclocks generated by the 680i and dial in our own choices. Using this approach, performance was the most satisfying we have yet found with a Core 2 processor.

NVIDIA has made dramatic improvements in overclocking abilities with 680i. They have effectively moved the new chipset from the embarrassment of mediocre overclocking that characterized the 590 chipset to the 680i evolving into the best overclocking platform you can buy for Intel. When you add to this the proven features like dual x16 SLI video, FirstPacket, dual Gigabit Ethernet with Teaming, HD audio, and the full MediaShield storage array with dual RAID 5, you have a solid product that stands out from the crowd. The added option to run a physics video card with SLI is just delicious icing on the cake.

NVIDIA should also be congratulated for keeping one very significant feature, namely IDE. Intel may have been trying to drive the market to SATA in the 965+ICH8 chipset by leaving out IDE. The problem, of course, is you simply can't find the SATA optical drives you are supposed to buy to use on the 965 boards. Instead manufacturers are forced to add a chipset to support IDE devices on their 965 boards. Thank you, NVIDIA, for avoiding this same pitfall.

The NVIDIA 680i is the chipset we would choose for our own purchase right now. It is the best chipset for the Core 2 platform that we have tested. If you don't need or want all the high end features, the 600i family also offers the lower priced 650i SLI and 650i Ultra aimed at those looking for a more economical motherboard. We still expect ATI to release RD600 in the next few weeks - at least in limited release. It will feature dual x16 CrossFire for Intel Core 2. We are looking forward to reviewing those boards, but it will take some truly revolutionary performance to top the NVIDIA 680i motherboards. If you got the clear message we like the 680i, you would be reading us correctly.