Nvidia nForce 680i SLI Chipset Review

by Ilya Gavrichenkov
11/08/2006 | 11:00 AM

Nvidia has been present in the chipset market for just a few years but they have managed to become pretty successful in this field already. Starting 2001, when they introduced their first Athlon XP core logic set, Nvidia became the chipset leader for AMD platform. I believe I have every right to state that the majority of today’s enthusiast platforms built around AMD Athlon 64 processors are based on the chipsets from this particular manufacturer. In other words, every new nForce product for AMD processors is better than the previous one and wins more fans. And there are multiple reasons for that. It is not only the technological innovativeness of each new nForce chipset, but also very smart marketing campaign. Besides, Nvidia’s competition in the Athlon 64 chipsets market has never been serious enough.

As for the situation in the Intel chipsets market, things are turning completely different here. Since they got the license for Intel Quad Pumped Bus in early 2005 and until recently they haven’t managed to introduce a core logic solution that would outperform Intel’s own developments from the consumer features standpoint. nForce Intel Edition chipsets lost to Intel’s own products from the performance prospective as well as from the overclocking potential standpoint. As a result, neither nForce4 (Intel Edition), nor nForce 5XX Intel Edition managed to win a significant market share in the chipset field. Most mainboard manufacturers have turned their back to these chipsets and wouldn’t use them in their products at all.

In fact, it hasn’t been a dramatic misfortune for Nvidia until recently. They were targeting their solutions for gamers and enthusiasts and hence could do without the focus on Intel platform, because it was not that popular in the above mentioned market segment. Now, however, the situation has changed to just the opposite. The launch of extremely successful processors on Core microarchitecture changed the common state of things: CPUs from Core 2 Duo family in no time became the most popular enthusiast solutions. Therefore, Nvidia has to introduce new chipsets for Core 2 Duo in order to retain the respect and popularity among the computer elite.

The growing popularity of multi-GPU configurations has become an additional stimulus to work harder on new Intel solutions. High-performance SLI systems that consisted of two GeForce graphics cards are only supported by nForce chipsets. Therefore, Nvidia has to provide high-quality chipsets for LGA775 platform, if they want to increase the sales of their GPUs.

Until today they have already introduced two modifications of Intel core logic sets. However, the only difference between the nForce4 SLI Intel Edition and the newer nForce 590 SLI Intel Edition was their South Bridges (Media and Communications Processor – MCP), which have moved to these solutions from the chipsets for Athlon 64 processors. In reality, the currently existing chipsets for Intel platform were based on the same chipset North Bridge (System Performance Processor – SPP) and hence suffered from identical problems, the major being extremely scarce CPU overclocking friendly options.

Today Nvidia has finally decided to give up the past heritage – the not very successful SPP chip. New chipset family for Intel processors they are announcing today, Nvidia nForce 600i features different SPP chip also known as C55. The major improvements of the new SPP chip are higher performance of the integrated memory controller and 1333MHz processor bus support that gives us hope for better overclocking results. So, looks like Nvidia nForce 600i should be free from all major drawbacks of the previous generation products and may become a good basis for new Core 2 Duo based systems. However, practice is not always following the theory that is why we should definitely run some benchmarks before we make up our mind about the new Nvidia chipset solution. For our today’s performance analysis we managed to get our hands on a mainboard based on the top solution from the nForce 600i family – Nvidia nForce 680i SLI. So, our today’ article will be devoted to this particular product.

Nvidia nForce 600i Chipset: Half-New

Nvidia usually releases more than one chipset modification. The same is true for the 6^th nForce series: they used the same MCP and SPP chips with slightly different sets of supported features to build three products targeted for different price groups. The top-of-the-line solution is nForce 680i SLI, the mainstream chipset is called nForce 650i SLI, and the value chipset in the family will be called nForce 650i Ultra. By artificially disabling some of the nForce 680i SLI features and reducing the number of supported PCI Express lanes they will create better value chipset modifications from the top nForce 680i SLI solution, therefore we will devote our today’s article to this particular product in the first place, because this way we will get a good idea of all three newcomers.

Although Nvidia nForce 680i SLI is a pretty advanced product, which features look very impressive even against the background of competitor solutions, our today’s story will not be too long. Like Nvidia nForce 590 SLI for Athlon 64 processor family, the new nForce 680i SLI chipset consists of two chips: SPP and MCP connected with HyperTransport bus. The MCP of this chipset is am MCP55PXE chip that you should know very well from the previous materials devoted to Nvidia’s chipset solutions. In fact, this MCP serves as single-chip nForce 570 SLI core logic for AMD platforms. The new Nvidia nForce 680i SLI has an additional SPP chip connected to it via the HyperTransport bus, and not the processor. It features a number of integrated controllers, supports Quad Pumped Bus and additional PCI Express lanes.

MCP analogs from the Nvidia nForce 680i SLI chipset have already been used in other Nvidia products, such as nForce 590 SLI for AMD and Intel systems, for instance. So, we have already discussed most of the features implemented in the new Nvidia chipsets before, so for more information please see our article called Chipsets for Socket AM2 Platform: ATI CrossFire Xpress 3200 and Nvidia nForce 590 SLI.

The most interesting part of the new Nvidia nForce 680i SLI chipset is certainly the chipset North Bridge (SPP). Nvidia hope that this microchip will help them compete successfully with the Intel’s own chipsets from the performance as well as overclocking prospective.

As you know, the previous Nvidia solutions for Intel platform (based on the old C19 SPP) couldn’t boast any significant overclocking potential. The increase in FSB frequency above the nominal rate would very soon lead to loss of system stability. The new C55 SPP officially supports 1333MHz bus that will be used by upcoming dual-core and quad-core processors on Core microarchitecture. And it means that Nvidia nForce 680i SLI boasts higher frequency potential.

The memory controller has also undergone pretty significant changes. Nvidia nForce 680i SLI has officially acquired DDR2-800 SDRAM support that the previous chipset solutions for Intel platform didn’t have. However, in reality, the new SPP features a greater choice of dividers forming the memory bus frequency from the FSB speed, so that you will be able to increase the memory speed up to 1200MHz in the nominal mode. Moreover, the new SPP boasts very well implemented mechanism for asynchronous memory bus clocking. As a result, Nvidia nForce 680i SLI may have the DDR2 SDRAM working at frequencies that are not connected with the FSB speed and can vary in a very broad range. Some overclocking fans may also be happy about the asynchronous memory clocking. Although the performance of the DDR2 memory controller gets slightly lower in case of asynchronous memory clocking, it allows setting the memory frequency lower than the front side bus frequency. As a result, you can use low-cost memory modules with limited overclocking potential in overclocked platforms on Intel Core 2 Duo and Core 2 Quad processors.

Note that Nvidia engineers paid special attention not only to the flexibility of the integrated memory controller of their nForce 680i SLI, but also to its performance. The SPP of the new chipset features enhanced DASP 4.0 unit (Dynamic Adaptive Speculative Preprocessor) responsible for pre-opening of those memory pages that might be later requested by the CPU. Together with the data prefetch algorithms implemented in contemporary Core based processors this technology may ensure additional performance improvement in applications working with large amounts of data.

The second important advantage of the Nvidia nForce 680i SLI memory controller is its ability work at 1T Command Rate. Mainboards built around this chipset will offer a corresponding option in the BIOS Setup. Note that competitor’s solutions, the Intel chipsets in the first place, work only with 2T Command Rate and do not allow adjusting this parameter. This gives us another reason to believe that the new chipsets will demonstrate higher performance.

Also the memory controller of Nvidia nForce 680i SLI supports overclocker’s DDR2 SDRAM modules with EPP (Enhanced Performance Profile), which allows setting automatically optimal memory work modes for memory kits with EPP support, that are getting more and more popular these days.

Besides the processor bus controller and the memory controller, the SPP of the new Nvidia nForce 680i SLI features another two important units. First of all, it is a PCI Express x16 bus for connection with the graphics card. Since SPP and MCP of the new Nvidia nForce 680i SLI support the total of 46 PCI Express lanes, the mainboard based on this chipset can be equipped with up to three PCI Express x16 slots (one of them is logically connected via the PCI Express x8 bus). As a result, the new Nvidia chipset not only supports SLI configurations but allows installing three graphics cards simultaneously, where one of the cards might work as a physics accelerator. The second important unit worth mentioning is the HyperTransport bus that connects the SPP and MCP together. The bandwidth of this bus is 8GB/s, which should be enough for two graphics cards in SLI mode if both of them are connected via the PCI Express x16 interface of the two different chipset bridges.

Nvidia nForce 680i SLI chipset South Bridge is identical to the Nvidia nForce 570 SLI MCP, so its features look as follows:

• Dual-port gigabit network controller supporting FirstPack technology for managing the priority of processed network packets and DualNet technology for consolidated use of network ports.
• Integrated ATA RAID controller supporting six Serial ATA ports with up to 3Gbit/s speed and a Parallel ATA-133 port. The controller can work with a single HDD as well as with RAID arrays 0, 1, 0+1 and 5. Serial ATA ports support NCQ, “Hot Swap” HDD connection and eSATA.
• 10 USB 2.0 prots.
• Built-in High Definition audio controller (Azalia).

So, Nvidia nForce 680i SLI looks like a pretty feature-rich solution that looks as good as any available competition out there. And in order to back up this statement once again, let’s take a look at a comparative feature chart for the solutions designed for Intel platforms:

From the formal specification standpoint, Intel chipsets cannot outpace the new Nvidia nForce 680i SLI. However, Intel developers are not pursuing this goal in the first place. This company is targeting the mainstream and budget market segments having left the high-end chipset field for Nvidia and ATI that are also pushing forward their multi-GPU SLI and CrossFire technologies in the same performance segment.

Nvidia didn’t stop on the high-performance Nvidia nForce 680i SLI solution, however. They have also launched two more chipset models based on the same component base but boasting cut-down features list. They are Nvidia nForce 650i SLI and Nvidia nForce 650i Ultra. Both these solutions have only one PCI Express x16 graphics bus and two PCI Express x1 lanes for external devices. Nvidia nForce 650i SLI differs from the Ultra by the possibility to configure the graphics bus as 2 x PCI Express x8, so that the system would be able to accommodate two graphics cards. Other than that, Nvidia nForce 650i SLI and Nvidia nForce 650i Ultra are identical, although they boast fewer features than the high-end Nvidia nForce 680i SLI: they have only one network port instead of two, four SATA and 8 USB 2.0 ports.

Note that thanks to these changes that may seem minor at first glance, Nvidia nForce 650i SLI and Ultra fall into a completely different price category, as intended. While the approximate price of Nvidia nForce 680i SLI based mainboards should be around $250-$300, the 650i based mainboards should be priced starting from $100. So, the new Nvidia chipsets can be used not only in high-end solutions for gamers and hardware enthusiasts but also in inexpensive products.

However, the mainboards on low-cost nForce 600i chipset modifications will start selling a little later. Only a few selected manufacturers are ready to start shipping Nvidia nForce 680i SLI based mainboards at this time, however, even they cannot yet offer any diverse choice. In fact, it looks like Nvidia’s closest partners will only be able to offer solutions based on the reference design at this time. So, today we are going to discuss a mainboard like that, also.

Reference Mainboard: Now Available in Stores

We are usually very skeptical about reviewing reference mainboards coming from the chipset developers. The thing is that these solutions rarely have anything in common with the actual retail products that the users can buy in stores later on, therefore testing reference solutions doesn’t make too much practical sense. However, in case of Nvidia nForce 680i SLI the situation turned out completely different.

The thing is that Nvidia tried to use their graphics card distribution strategy in the mainboard market. In reality it will look as follows. There will be one large contract manufacturer responsible for mass production of reference Nvidia nForce 680i SLI based mainboards. Then these mainboards will be distributed among the mainboard vendors who will sell them under their own brand names and will only be able to change the package design and accessories bundle. There are five companies that will start shipping Nvidia nForce 680i SLI based mainboards in the nearest future: eVGA, BFG, XFX, ECS and Biostar.

There will be another five mainboard manufacturers who will try to introduce their own Nvidia nForce 680i SLI based mainboard designs. They are ASUS, DFI, MSI, Gigabyte and ABIT. However, their own mainboard designs are not ready yet, so we have no other choice but review a reference Nvidia nForce 680i SLI based product at this time, and if you like it you will be able to buy it in stores.

At first let’s take a closer look at the specification:

As you can see from the specification, maximum of the available chipset features have been implemented on this reference mainboard. Note that there are no additional controllers onboard except the Texas Instruments chip responsible for IEEE1394 bus implementation. However, despite the absence of additional onboard controllers, the solution is quite feature-rich – all thanks to the Nvidia nForce 680i SLI chipset.

The diagnostic CPU-Z utility reports the following about this mainboard:

We included this screenshot into our article because it clearly indicates that the South Bridges of Nvidia nForce 680i SLI and Nvidia nForce 570 SLI are identical. Other than that there is nothing else interesting about it, especially since the current utility version cannot detect the chipset North Bridge correctly.

Our mainboard supports all contemporary Intel processors for LGA775 form-factor and is claimed to support the upcoming processors on Core microarchitecture that will use Quad Pumped Bus with 1333MHz frequency.

There are four DDR2 DIMM slots on the board, two slots per channel. They are color-coded advising the user the best way to install memory modules in pair for dual-channel mode. There are two ways to set the memory frequency in the BIOS Setup: by selecting the divider that will derive the memory frequency from the FSB speed, or by clocking the memory asynchronously. The first way is certainly preferable, because it ensures lower latencies when addressing the memory subsystem, however there are very few dividers to choose from. There are only the following three memory-to-FSB frequency ratios available: 1:1, 5:4 and 3:2.

However, the asynchronous mode is extremely flexible, and you can set the memory frequency at pretty much any rate between 400MHz and 1200MHz. in asynchronous mode you can also set the memory frequency lower than the FSB, which might be very helpful during overclocking.

Speaking about the memory configuration possibilities offered by this mainboard we should also check out the BIOS Setup options for latency adjustment. The mainboard allows adjusting all major timings alongside with Command Rate parameter that is unavailable on Intel based mainboards.

Among the features that catch your eye right away we would like to mention the availability of three PCI Express x16 slots. The middle slot (blue) is logically connected to the PCI Express x8, although you can still install graphics cards into it. In fact, this slot is supposed to accommodate a physics accelerator, which doesn’t require high bandwidth.

So, you should configure the SLI system using black PCI Express x16 slots. But look how far apart they are! Nvidia engineers have given some serious thought to proper graphics card cooling and placed the graphics card slots as far apart as possible, which is definitely a great advantage of this design.

Actually, I have to stress that the reference Nvidia nForce 680i SLI based mainboard design is exceptionally smart. All connectors are very conveniently located, the HDD connectors are at the very front of the PCB, and those that couldn’t fit in front of the DIMM slots are turned parallel to the PCB. There are two power supply connectors in front of the DIMM slots: a 24-pin connector and an additional MOLEX for SLI configurations. The 8-pin 12V ATX connector has been moved to the very right side of the board behind the processor socket. Even the Clear CMOS jumper is placed in a very reasonable spot. It is placed at the left edge of the PCB, so that you can easily reach it even when the mainboard is installed into the PC case.

The processor voltage regulator features six-phase design and uses only high-quality components. The MOSFET of the regulator are covered with matt black-anodized aluminum heatsinks matching the PCB color of the mainboard.

The mainboard I/O panel is pretty standard. It features 6 USB 2.0 ports, PS/2 ports for keyboard and mouse, two RJ-45 network connectors with diagnostic LEDs, a FireWire port, six audio-jacks and an optical SPDIF out.

The remarkable thing about positioning of this product for hardware enthusiasts is not the availability of numerous additional controllers, but a few very pleasant trifles that come in handy during overclocking experiments. Among these trifles most certainly are the POST-codes indicator and Power On and Reset micro-buttons soldered directly onto the PCB.

Also Nvidia made sure that a lot of additional fans can be connected to their mainboard: there are seven corresponding connectors on the mainboard PCB. Moreover, the mainboard allows monitoring and adjusting the rotation speed of five fans out of seven from the BIOS Setup.

However, it is very hard to judge whether these small nice things justify the price of the solution which exceeds $250. But in the end, there are no other acceptable platforms for SLI configurations built around Intel Core 2 Duo CPUs. So, it would be not quite correct to regard this solution as a direct competitor to mainboards on Intel chipsets.

I would definitely like to say a few words about the chipset cooling solution. Although Nvidia uses 90nm process to manufacture the SPP of its nForce 680i SLI chipset, it heats up quite noticeably and hence requires efficient cooling system. Nvidia used a very unique heatpipe cooler on its mainboard.

This heatpipe transfers the heat from MCP to a pretty large SPP heatsink with large cooling surface provided by numerous thin copper ribs.

This cooler design allows Nvidia nForce 680i SLI based reference mainboard to do without active chipset cooling. However, only until you decide to perform some overclocking. Once the frequencies of the major busses increase, the chipset dissipates much more heat, so Nvidia recommends using a special 5000rpm fan to protect the chipset against fatal overheating. This fan comes with the mainboard and can be fastened with special clips and two tiny screws that can be tightened with a watch screwdriver. So, you will have to sacrifice the quiet operation for the sake of proper SPP and MCP cooling on Nvidia nForce 680i SLI based reference mainboard.

In conclusion I would like to say that although we admitted the smart design of the Nvidia reference mainboard, we still came across some design issues during our test session. And these issues were hiding on the reverse side of the PCB. Unfortunately, there are two relatively big case-free capacitors right beneath the processor socket, so you may experience some problems with the installation of the cooling solutions requiring backplate at the bottom of the PCB.

Overclocking: Nvidia’s Success

Previous Nvidia chipsets for Intel processors didn’t boast any overclocking records. It was really hard to increase their FSB frequency above the nominal value and the system would lose stability very quickly. With the new Nvidia nForce 680i SLI chipset Nvidia claims they are offering great overclocking experience. The new SPP is claimed to be stable at the FSB speeds far beyond the nominal values and should also work fine with the CPUs supporting 333MHz nominal FSB frequency.

Of course, we couldn’t miss the opportunity to check out these brave claims, so we devoted a special section of our review to investigating the mainboard performance at higher processor bus speeds.

The BIOS Setup of the reference Nvidia nForce 680i SLI based mainboard offers pretty rich set of overclocking friendly adjustable parameters. In fact, this mainboard can be compared with the best overclocking friendly mainboards out there. Here is a list of parameters that you can play with in the BIOS Setup during overclocking:

• You can change the FSB frequency from 100 to 625MHz with 1MHz stepping;
• The PCI Express bus frequency can be adjusted independently for all busses;
• You can independently adjust the frequency of the HyperTransport bus connecting the SPP and MCP;
• The CPU voltage can be adjusted from the nominal value to 1.8V with 0.0125V stepping;
• The FSB Termination voltage can be adjusted from 1.2V to 1.5V with 0.1V increment;
• The DIMM voltage can be adjusted from 1.8 to 2.5V with 0.05V stepping;
• SPP voltage can be adjusted from 1.2V to 1.55V with 0.05V increment;
• The MCP voltage can be adjusted from 1.5V to 1.755V with 0.025V increment;
• The HyperTransport bus voltage can be adjusted from 1.2V to 1.55V with 0.05V increment.

At first glance this list is pretty extensive. These functions should be more than enough even for extreme CPU overclocking experiments.

By the way, among the options offered in the BIOS Setup there is a setting that allows disabling processor cores independently, which may also be helpful for setting overclocking records.

For our practical tests we assembled a system around the Nvidia nForce 680i SLI based reference mainboard and Intel Core 2 Duo E6300 CPU working at 1.86GHz nominal speed. The system was also equipped with 2GB of Corsair TWIN2X2048-8888C4DF (Dominator) memory. The CPU was cooled down with Zalman CNPS9500 LED cooler and the PowerColor X1900 XTX 512MB represented the graphics subsystem.

Our first task was to determine the maximum FSB frequency when the Nvidia nForce 680i SLI based platform would remain stable. In order to eliminate possible issues right from the start we clocked the memory synchronously with the FSB and the timings were set to 4-4-4-12-2T.

To our great satisfaction the reference mainboard easily surpassed 333MHz FSB, as well as 400MHz FSB. However, after that overclocking didn’t go any further, and the board wouldn’t boot at 450MHz FSB. We were pleased with the fact that the after freezing on the POST stage the mainboard would restart with “safe” parameters so that we could get into the BIOS Setup and adjust the settings. This extremely useful function functions saves you a lot of time during testing, because you don’t have to reach for Clear CMOS jumper.

As we continue our overclocking discussion, we would like to remind you that Core 2 Duo E6300 processor used to work stably at 500MHz FSB on the best overclocker’s mainboards for LGA775 form-factor based on Intel P965 Express chipset. Therefore, it is evident that something new has hindered further overclocking on our Nvidia nForce 680i SLI based mainboard. Unfortunately, we couldn’t figure out what the problem was in this case. We couldn’t get beyond that point even after raising the processor Vcore, chipset voltage and bus voltages. All we could do is find that maximum FSB speed when the mainboard could boot safely. As we have found out almost immediately, the system boots and works absolutely stable at 449MHz FSB frequency. However, if we added at least 1MHz in the BIOS Setup, the board couldn’t pass the POST stage.

I think that it could be the not completely finalized BIOS that is responsible for this situation, things could improve with the future BIOS versions.

However, as of today we have to admit that Nvidia nForce 680i SLI based mainboards are a less efficient for processor overclocking than the solutions on Intel 975X Express and Intel P965 Express.

The second thing we studied during our overclocking experiments was results of memory overclocking on Nvidia nForce 680i SLI based reference platform. We decided to perform this test because contemporary mainboards on Intel chipsets available in the today’s market demonstrate absolutely different results when the memory bus frequency is increased. In particular, Intel P965 Express based mainboards support much higher memory bus frequency than Intel 975X Express based ones, which we have already mentioned in our article called Choosing the Right Memory for Core 2 Duo Platform - Part 2: DDR2-1000, DDR2-1067 SDRAM.

Within this series of experiments we decided to overclock our Corsair TWIN2X2048-8888C4DF (Dominator) kit that could hit 1152GHz in ASUS P5B Deluxe mainboard with 4-4-4-12 timings and 2.4V voltage (see our article called DDR2 SDRAM to Hit 1.1GHz: Corsair Dominator vs. OCZ PC2-8800 Gold Edition).

We didn’t clock the memory bus on our Nvidia nForce 680i SLI based mainboard asynchronously. We tried to hit higher speeds by adjusting the dividers and FSB frequency.

During our test session we managed to make a few very interesting conclusions. Firstly, when the memory frequency exceeds 800MHz, 1T Command Rate causes system instability. In other words, if you raise the frequency, then you should raise the Command Rate setting, although in this case the system may lose about 2-3% of its performance.

The second conclusion also turned out not very optimistic. Our mainboard failed to let the memory work at the same frequency as on Intel P965 Express mainboard we used before. Nvidia nForce 680i SLI reference solution allowed the memory to work at 1098MHz maximum, no higher than that. Although the system would still start at even higher DDR2 SDRAM memory frequency, the system would still fail the stability tests.

As a result, we can only conclude that Nvidia nForce 680i SLI cannot yet compete with Intel P965 Express, especially its new C2 revision, in terms of overclocking potential. At the same time, I have to stress that Nvidia nForce 680i SLI has considerably improved its overclocking friendliness since the times of nForce 590 Intel Edition. Moreover, we still hope that future BIOS versions as well as the mainboards based on manufacturers’ own design may be free from the problems we encountered today during our overclocking session.

Testbed Configuration

The final part of our Nvidia nForce 680i SLI based reference mainboard discussion will be devoted to the performance analysis. The memory controller in Intel systems is located in the chipset North Bridge therefore the chipset can exert serious influence on the system performance unlike the systems built for Athlon 64.

So, we will compare the performance of the system built around the new Nvidia nForce 680i SLI reference board with the performance of similar platforms built on Intel i975X and iP965 solutions. Our testbeds were configured as follows:

• CPU: Intel Core 2 Extreme X6800 (LGA775, 2.93GHz, 4MB L2).
• Mainboards:
  • ASUS P5W DH Deluxe (LGA775, Intel 975X Express);
  • ASUS P5B Deluxe (LGA775, Intel P965 Express);
  • NVIDIA nForce 680i SLI reference board.
• Memory: 2 GB DDR2-800 SDRAM (Corsair TWIN2X2048-8888C4DF).
• Graphics card: PowerColor X1900 XTX 512MB (PCI-E x16).
• HDD: Western Digital WD1500AHFD (SATA150).
• OS: Microsoft Windows XP SP2 with DirectX 9.0c.

Here I would like to say that during our test session we experienced some problems with the memory subsystem configuring. At first we wanted to test systems with DDR2-800 SDRAM with the minimal timings settings. Corsair TWIN2X2048-8888C4DF memory modules can work at 3-3-3-10 timings in Intel based mainboards, which we have already revealed in our article called DDR2 SDRAM to Hit 1.1GHz: Corsair Dominator vs. OCZ PC2-8800 Gold Edition. In this case the Command Rate was set at 2T, because Intel chipsets do not support 1T Command Rate.

We expected Corsair TWIN2X2048-8888C4DF memory to run with the same 3-3-3-10 timings in Nvidia nForce 680i SLI reference board, with 1T Command Rate setting. However, the system was running very unstably in this case. As a result, we had to raise the timings to 3-4-3-10 at 1T Command Rate.

So, the memory subsystems of the tested platforms were configured differently, however, these were the most aggressive settings possible. Intel chipsets worked with 3-3-3-10-2T timings, and Nvidia chipset worked at 3-4-3-10-1T timings. All other BIOS Setup parameters were adjusted to deliver maximum performance possible.

Memory Subsystem Performance: Synthetic Benchmarks

First of all we used some diagnostic utilities that help measure the practical latency and memory subsystem bandwidth.

According to Sandra 2007, the memory subsystem on Nvidia nForce 680i SLI based platform works faster than that on iP965 based one, but slower than on i975X based one. So, it turns out that the DASP 4.0 and 1T Command Rate do not guarantee high performance of the memory subsystem. ASUS engineers managed to squeeze much higher speeds out of the Intel 975X core logic.

However, we shouldn’t forget that the algorithms used in the chipset memory controllers depends on the data usage model selected by the running application. For example, the results in Everest 2006 differ from the results of Sandra 2007. According to this test, even though i975X provides the best latency, Nvidia nForce 680i SLI chipset is not any slower from the practical bandwidth standpoint.

Performance

Now let’s take a look at the performance of Nvidia nForce 680i SLI in complex benchmarks and real applications.

The results in popular synthetic benchmarks are pretty similar for Nvidia nForce 680i SLI based reference board and Intel 975X Express and Intel P965 Express. nForce 680i SLI is evidently falling behind the rivals only in SuperPi, where i975X proves 3.6% faster, and iP965 – 2.2% faster.

Intel chipsets are a little bit ahead of the new Nvidia nForce 680i SLI solution in games. Only the results of F.E.A.R. do not follow the overall trend here, because Nvidia nForce 680i SLI chipset managed to outperform i975X by about 2%. This is a great illustration to the fact that the performance balance between the core logic sets from different manufacturers depends a lot on the type of application used for performance analysis. It could be explained by serious heterogeneity of the algorithms used to accelerate the work with the memory subsystems in different chipsets.

During video encoding Nvidia nForce 680i SLI cannot boast superior results. It is slightly behind Intel 975X Express and Intel P965 Express based mainboards.

Similar picture can be observed in Adobe Photoshop. However, the performance difference here doesn’t exceed 2-3%, which is not a dramatic failure at all.

In WinRAR i975X manages to show a much higher result than Nvidia nForce 680i SLI.

3ds max is a pretty heavy professional application, and the new Nvidia nForce 680i SLI chipset cannot really show its best here. The all-Intel systems ensure about 4-5% performance advantage during final rendering as well as in viewports.

Power Consumption

In our today’s review devoted to the reference mainboard on Nvidia nForce 680i SLI chipset we have already mentioned its massive North and South Bridge cooling system designed using copper and heatpipe technology. The coolers used on Intel chipsets look much less sophisticated. It gives us some concern that one of the drawbacks of the new Nvidia nForce 680i SLI core logic set may be its high heat dissipation.

Since heat dissipation is directly connected with the power consumption we decided to measure how much power the similar systems built around different chipsets would consume. We measured the total power consumption of the entire platform except the monitor in several modes.

First of all we measured the platforms power consumption in idle mode, when there is nothing running. Note that we didn’t enable Intel Enhanced SpeedStep or Enhanced Halt State technologies during these tests.

The first results immediately reveal the difference. In idle mode Nvidia nForce 680i SLI based platform consumes about 20-30W more power than the competitors. And these 20-30W can be assigned to the chipset only, because the system’s configuration differed only by the mainboard they were built on.

Now let’s take a look at the results of our second experiment, when the CPU was under 100% workload traditionally created by S&M utility.

We once again observe a 20-30W difference.

The third experiment was performed with the graphics subsystem being fully loaded by the 3DMark06 testing suite.

In this case the difference in power consumption turns even greater and makes 30W compared with the iP965 and 50W compared with the i975X.

So, we can state that Nvidia nForce 680i SLI is a less economical chipset than the solutions from Intel. And if you are looking to get a quiet economical system, then Nvidia solution will hardly be your choice.

Summing up the power consumption and heat dissipation analysis we can conclude that Nvidia’s new chipset loses quite noticeably to the Intel P965 Express and Intel 975X Express from the performance-per-watt prospective.

Conclusion

There has been no question about the best chipsets for Intel platform until recently: they were designed by Intel, of course. However, the market realities have changed now: there is another highly ambitious player in the chipset market – Nvidia Corporation. And although they haven’t been able to roll out high-quality solutions for LGA775 systems for quite some time, sooner or later they had to break through.

The new Nvidia nForce 680i SLI is a serious candidate for success. Unlike its predecessors, this chipset is a pretty mature solution, free from numerous “children’s diseases”. Therefore, those users who prefer Nvidia graphics and are going to use SLI configurations should definitely go for the newcomer. Especially, since there are hardly any real alternatives in the today’s market for an SLI system based around Core 2 Duo or Core 2 Quad processors.

If you are not interested in putting together an SLI system, then we wouldn’t recommend Nvidia nForce 680i SLI as a basis for your platform at this time. The thing is that the only mainboard available today is the reference mainboard and it didn’t prove the best in our performance, overclocking and power consumption tests. In other words, it couldn’t outperform the overclocker’s solutions in Intel 975X Express and Intel P965 Express chipsets in any of the practical tests, except the formal specifications, though.

So, if you are happy with the today’s Intel 975X Express and Intel P965 Express based mainboards, then there is no need for you to switch to Nvidia nForce 680i SLI. As for us, we will continue testing new Nvidia nForce 600i chipset family for Intel platform as soon as we get our hands on the mainboards based on manufacturer’s own design, not the reference. And hopefully most of the problems we have described in our today’s articles will go away with the new BIOS versions and new design layouts.