Nvidia’s nForce 780i SLI chipset

A new reference motherboard design
The nForce 780i SLI comes with a new reference motherboard design that will be brought to market as-is by some of Nvidia’s partners. Larger motherboard makers like Asus, Gigabyte, and MSI tend to spin their own board layouts. However, Nvidia’s reference designs have been popular with XFX, EVGA, BFG Tech, Foxconn, and even ECS. We used an XFX 780i SLI board for testing and have a second board from EVGA that’s essentially identical.

Overall, the board layout hasn’t changed much since the nForce 680i SLI, which isn’t necessarily a bad thing. Nvidia has managed a layout that is packed, but not crowded, with most of the connectors in just the right place.

Take the board’s power plugs, for example. Both the primary 24-pin connector and the 8-pin auxiliary 12V plug are located along or near the top edges of the board to allow for cleaner cable routing within traditional ATX enclosures.

Things have changed a little around the socket, where a brand new cooling array covers the chipset’s SPP and MCP components, the nForce 200, and the board’s voltage regulation circuitry. The chunky cooler is actually less elaborate than the twisted heatpipe networks we’ve seen on some motherboards, despite the fact that it has an additional chip to cool in the nForce 200. There’s plenty of surface area, though, and the heatsinks are carefully shaped to provide clearance for larger aftermarket CPU coolers. Even Scythe’s massive Ninja heatsink will fit on this board.

We had no problems running the 780i SLI reference board at stock speeds with passive chipset cooling, but for overclocking, Nvidia provides an auxiliary fan that snaps onto the north bridge cooler. This fan plugs into the motherboard, and its speed can be controlled through the BIOS. The fan is pretty loud, though, so you’ll want to avoid using it if you can.

On an open test bench with a single graphics card installed, the fan was only necessary once front-side bus speeds started to approach 450MHz (or 1800MHz, quad-pumped). However, we’ve also observed the need for additional chipset cooling with SLI configurations, so you might have to bust out the fan for multi-GPU systems, as well.

Moving south, the 780i’s chipset cooler covers the south bridge with a low-profile heatsink that won’t interfere with longer graphics cards. Nvidia has also done well to ensure that none of the board’s Serial ATA ports will be blocked by longer or double-wide graphics cards. Even with a trio of massive cards installed, you’ll still have access to all six of the board’s SATA ports and even its IDE port, too.

One minor change for the reference design is the location of the board’s front-panel connectors. On the 680i SLI reference board, these connectors were oddly located up by the DIMM slots. The 780i, they’ve been moved to a more traditional location in the bottom right-hand corner of the board (bottom left in the picture above). There, the front-panel connectors are joined by onboard power and reset buttons and a handy two-digit POST code display.

Three-way SLI compatibility calls for a trio of PCI Express x16 graphics slots. Only the top two slots are PCIe 2.0, with the third hanging off the south bridge’s first-generation PCI Express implementation. You also get a couple of standard PCI slots and one PCIe x1 slot.

With three double-wide graphics cards installed, there won’t be room for any other expansion cards. Fortunately, such configurations are likely to be uncommon. Even with a two-way SLI config, you’ll still have access to standard PCI and PCIe x16 slots. And because PCIe slots are backwards compatible, an x16 slot can also handle x8, x4, and x1 cards.

Little has changed around the 780i SLI’s port cluster, where you’ll find a pretty standard array of expansion ports. A full suite of analog audio ports make an appearance fed by Realtek’s ALC888S codec chip. You get a TOS-Link digital audio output, too, but no provision for digital audio input or coaxial output.

Six USB ports make an appearance here, with the remaining four accessible via onboard headers. An additional Firewire port is available via an onboard header, as well, with both fed by a Texas Instruments TSB43AB22A chip that hangs off the PCI bus and supports 1394a “Firewire 400” speeds.

The only glaring omission from the port cluster—apart from serial and parallel ports that perhaps only half a dozen folks will actually miss—is external Serial ATA connectivity. eSATA enclosures aren’t overwhelmingly common just yet, but with eSATA ports available on just about every high-end board, they should really be included here.

BIOS options galore
Nvidia has long done a good job of providing plenty of tweaking and overclocking options in its reference BIOSes, and the one that comes with the 780i SLI is no exception.

There’s a little something for everyone, including incredibly fine-grained control over the system’s front-side and memory bus clocks. The 780i SLI may not officially support processors built to run on a 1600MHz front-side bus, but the BIOS will let you overclock the board to that speed and beyond. Memory bus speed options go all the way up to 1400MHz, as well, although Nvidia only guarantees up to 1200MHz. Slackers.

If you’re going to dabble in extreme overclocking, you’ll be pleased to note that the BIOS offers CPU multiplier control. A generous array of voltage options is available, as well. CPU voltage options go all the way up to 1.8V, and memory can be cranked up to 2.5V. Combine that with control over front-side bus, interconnect, and chipset voltages, and the 780i should offer overclockers all the flexibility they desire.

As one might expect, the BIOS serves up all the usual memory timing controls, including the ability to manipulate the DRAM command rate. Fan speed control makes an appearance, too, and in style no less. Users can set temperature-based fan speeds for the processor and chassis fan, and manual fan speed control is offered for three additional onboard fan headers.

If you don’t feel like poking around in the BIOS, the 780i SLI reference design offers full support for Nvidia’s nTune tweaking and system monitoring utility. Nvidia actually has a new, beta version of the app that’s ESA-enabled and should be available for download today. Look for a more detailed report on this new utility, and all things ESA, soon.

Specifics on specifications
For those who prefer motherboard specifications distilled into chart form, here’s the skinny on the nForce 780i SLI reference motherboard design. Note the lack of auxiliary Serial ATA and Gigabit Ethernet chips; Nvidia likes to handle as much as possible within the chipset, so it doesn’t need to outsource much functionality to third-party silicon.

Our testing methods
We’ve assembled a small collection of boards for testing to represent the 780i SLI’s most direct competitors. From the Intel camp, we have P35 and X38 Express-based motherboards from Asus and Gigabyte. The Asus boards use DDR3 memory, while the Gigabytes are equipped with DDR2. We’ve also thrown in an nForce 680i SLI reference board from EVGA to illustrate how the 780i compares with its predecessor. The nForce 780i SLI comes with a slightly newer ForceWare driver package, which is why that board is using different drivers than the 680i.

All application tests were run on each board. However, since our Intel-based DDR2 and DDR3 configs all share the same ICH9R south bridge, we’ve limited our Serial ATA, USB, PCI, and PCIe testing to the DDR2 systems.

All tests were run three times, and their results were averaged.

Thanks to Corsair for providing us with memory for our testing. 2GB of RAM seems to be the new standard for most folks, and Corsair hooked us up with some of its 1GB DIMMs for testing.

All of our test systems were powered by OCZ GameXStream 700W power supply units. Thanks to OCZ for providing these units for our use in testing.

Finally, we’d like to thank Western Digital for sending Raptor WD1500ADFD hard drives for our test rigs. The Raptor’s still the fastest all-around drive on the market, and the only 10K-RPM Serial ATA drive you can buy.

We used the following versions of our test applications:

• SiSoft Sandra Standard 2008 12.34
• WorldBench 6.0 Beta 2
• TCD Labs HD Tach 3.01
• Futuremark 3DMark06 Build 110
• RightMark Audio Analyzer 6.05
• NTttcp
• CASE Lab Euler3d CFD benchmark multithreaded edition
• Intel IOMeter v2004.07.30
• Crysis
• Enemy Territory: Quake Wars Demo
• Bioshock
• CPU-Z 1.41

The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Memory performance
Memory subsystem performance doesn’t always track with real-world applications, but it’s a good place to start.

And the nForce 780i SLI starts off pretty well, trailing only the 680i in memory bandwidth and latency. The gap in performance is more notable with latency than it is with bandwidth, where the two nForce boards essentially tie. Note that the nForce boards have a particularly large advantage over the Intel chipsets when the latter are confined DDR2 memory, as well.

Memory controllers don’t always handle four DIMMs gracefully, so we popped an additional two memory modules into each system for another round of tests. In these tests, we had to back off to a 2T command rate for the nForce and DDR3-equipped X38 systems. This is common adjustment for four-DIMM configurations.

Popping additional DIMMs into our 780i board improves its standing in our memory subsystem tests. It essentially ties the DDR3-equipped Intel chipsets on the bandwidth front while delivering much lower memory access latencies. In the latency test, the 780i is just a fraction of a nanosecond from turning in the fastest overall performance, too.

The following latency graphs are a little indulgent, so I won’t be offended if you skip them. They show access latencies across multiple block and step sizes, painting a fuller picture of memory controller performance with each chipset. I’ve arranged the graphs in order of highest latency to lowest. Yellow represents L1 cache, light orange is L2, and dark orange is main memory.

Nvidia has one heck of a DDR2 memory controller in its nForce 680i and 780i SLI chipsets.

STARS Euler3d computational fluid dynamics
Few folks run fluid dynamics simulations on their desktops, but we’ve found this multi-threaded test to be particularly demanding of memory subsystems, making it a good link between our memory and application performance tests.

For whatever reason, both nForce chipsets struggle in our STARS fluid dynamics simulation test. The 780i SLI does no worse than the 680i, but that still puts it off the pace set by Intel’s best.

Worldbench
WorldBench uses scripting to step through a series of tasks in common Windows applications. It then produces an overall score. WorldBench also spits out individual results for its component application tests, allowing us to compare performance in each. We’ll look at the overall score, and then we’ll show individual application results alongside the results from some of our own application tests.

We ran into a curious issue with the latest version of WorldBench when testing our P35 system with DDR2 memory. In the first run, the system posted a score in the WinZip test that matched our expectations given the scores of the other systems. However, subsequent test runs registered WinZip test times that were 20% faster than the first run. This behavior wasn’t consistent with that of the other systems, despite the fact that each starts testing with a fresh hard drive image. We suspect Windows Vista’s intelligent disk caching may be responsible for the faster scores, so we’ve reported the results of the first WorldBench test run for the P35 DDR2 config below.

The 780i scores one point off the pace set by its predecessor in WorldBench, putting it ahead of most of the Intel chipsets. Let’s break things down to see if we can find a gem or two within WorldBench’s individual test results.

Little separates the 780i SLI from its competition through WorldBench’s multimedia editing and encoding tests.

The nForce chipsets are a little slower in Firefox, though, and that also affects performance in WorldBench’s multitasking test.

Performance doesn’t vary much from board to board in the 3ds max tests.

However, both Nero and WinZip show advantages for the nForce line. The 680i and 780i trade the lead back and forth, but they both manage to outpace the closest Intel competition.

Gaming

As one might expect, there’s little difference between platforms in our game tests. Even the system-crippling Crysis doesn’t have an overwhelming preference for one platform over another at the resolution and detail level we used for testing.

Serial ATA performance
The Serial ATA disk controller is one of the most important components of a modern core logic chipset, so we threw each platform a selection of I/O-intensive storage tests using a Western Digital Raptor WD1500ADFD.

IOMeter
We’ll begin our storage tests with IOMeter, which subjects our systems to increasing multi-user loads. Testing was restricted to IOMeter’s workstation and database test patterns, since those are more appropriate for desktop systems than the file or web server test patterns.

IOMeter throughput is a little lower on the 780i that it is with the 680i, perhaps due to the slightly different driver revisions used for each board. The ICH9R south bridge paired with the P35 and X38 Express fares better than both nForce chipsets overall, ramping transactions more quickly in response to load increases and scaling a little bit higher under the heaviest loads.

Differences in IOMeter response times are really only apparent under extremely heavy loads, where the nForce chipsets yield a little ground to Intel.

CPU utilization is low across the board, rarely breaking 1%.

HD Tach
We used HD Tach 3.01’s 8MB zone test to measure basic SATA throughput and latency.

The 780i SLI’s transfer rates look a little low here. Write speeds are higher than they should be for all the chipsets, though. The combination of HD Tach’s write speed test, our Raptor hard drive, and chipsets that implement Native Command Queuing tends to produce artificially inflated scores.

Access times are pretty close, although the 780i SLI is half a millisecond off the 680i.

CPU utilization results are close considering HD Tach’s +/- 2% margin of error in this test.

USB performance
Our USB transfer speed tests were conducted with a USB 2.0/Firewire external hard drive enclosure connected to a 7200RPM Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.

Recent Nvidia chipsets have had a knack for USB, and that trend continues with the 780i SLI. The chipset not only delivers the fastest burst speeds of the lot, but its write speeds are a couple of MB/s quicker than Intel’s fastest chipset. CPU utilization is nice and low, as well.

Ethernet performance
We evaluated Ethernet performance using the NTttcp tool from Microsoft’s Windows DDK. The docs say this program “provides the customer with a multi-threaded, asynchronous performance benchmark for measuring achievable data transfer rate.”

We used the following command line options on the server machine:

ntttcps -m 4,0,192.168.1.25 -a

..and the same basic thing on each of our test systems acting as clients:

ntttcpr -m 4,0,192.168.1.25 -a

Our server was a Windows XP Pro system based on Asus’ P5WD2 Premium motherboard with a Pentium 4 3.4GHz Extreme Edition (800MHz front-side bus, Hyper-Threading enabled) and PCI Express-attached Gigabit Ethernet. A crossover CAT6 cable was used to connect the server to each system.

The boards were tested with jumbo frames disabled.

You’ll see that we’ve changed the names here and included a lot more results. All of our P35 and X38 Express-based motherboards use third party chips to handle Gigabit Ethernet, this despite the fact that the ICH9R has an integrated GigE MAC. The nForce 780i, however, integrates a pair of Gigabit controllers directly into the chipset, just like the 680i did before it. That leaves us with little choice but to compare networking performance on a motherboard rather than chipset level, so you’ll find boards listed by name followed by their Ethernet source in brackets.

The 780i SLI’s GigE performance shadows that of the 680i, and it’s really not that impressive when compared with what third-party controllers offer. Only one of the 780i’s GigE connections delivers full throughput, with the second stuck around 816Mbps. CPU utilization is higher for the nForce chipsets, too.

PCI Express performance
We used ntttcp to test PCI Express Ethernet throughput using a Marvell 88E8052-based PCI Express x1 Gigabit Ethernet card.

Apart from the P35’s lower CPU utilization, there isn’t much variance in PCI Express throughput with a standard Gigabit Ethernet card.

PCI performance
To test PCI performance, we used the same ntttcp test methods and a PCI VIA Velocity GigE NIC.

nForce PCI performance pulls up about 100Mbps short, though. And to make matters worse, it does so with higher CPU utilization—not a good combination.

Power consumption
We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Power consumption was measured at idle and under a load consisting of a multi-threaded Cinebench 10 render running in parallel with the “rthdribl” high dynamic range lighting demo.

In testing, we discovered that the BIOS for the GA-P35-DQ6 motherboard we used for our P35/DDR2 system doesn’t correctly throttle clock speeds with either C1E or SpeedStep. Older BIOS revisions used to work, but Gigabyte broke something with its last release. That could explain the lower CPU utilization results we saw for the P35 configuration in some of our peripheral testing.

Recent nForce chipsets have been quite power-hungry, and the 780i SLI falls right in line. With an extra nForce 200 chip under the hood, the 780i draws about 7W more power at idle and under load than the 680i SLI. That puts its power consumption much higher than of Intel’s P35 and X38 Express chipsets, at least when they’re paired with DDR2 memory.

Overclocking
For our overclocking tests, we dropped our CPU multiplier to 6X—its lowest possible value. The memory bus was also maintained at 800MHz to keep our DIMMs running well within their limits at overclocked front-side bus speeds. Next, we turned our attention to the front-side bus, cranking it up and using a combined load of Prime95 and the rthdribl HDR lighting demo to test stability along the way.

When we hit a 450MHz front-side bus, it became necessary to add the 780i’s optional chipset fan to keep the system stable under load. But we didn’t stop there. The board easily sailed up to 480MHz without the need for extra voltage or additional tweaking. It would go no further, though, and no amount of voltage fiddling or human sacrifice could coax 490MHz from the front-side bus.

480MHz is still quite an overclock and more than enough to run a QX9770 on a 1600MHz quad-pumped bus. However, it’s worth noting that we’ve seen 500 and 510MHz front-side bus speeds with identical hardware on X38 Express boards.

As is always the case with overclocking, your mileage may vary.

Motherboard peripheral performance
Core logic chipsets integrate a wealth of peripherals, but they don’t handle everything. Firewire and audio are farmed out to auxiliary chips, for example. Intel chipsets also rely on third-party silicon for networking, and many motherboards feature additional SATA controllers to complement south bridge Serial ATA offerings.

To provide a closer look at the peripheral performance you can expect from the motherboards we’ve tested today, we’ve complied USB, Firewire, Serial ATA, and Audio performance results below. You’ll notice that there isn’t much variance from one board to another, but there are a few things worth pointing out. Our 780i-based motherboard is highlighted and in bold to make it easier to pick out from the crowd.

Nothing to see here, move along.

The 780i reference board’s Texas Instruments Firewire chip is an improvement over what was used with the 680i SLI. However, the board’s average write speeds are still lower than the competition’s.

Without auxiliary storage controllers, the 780i SLI’s only SATA option is the chipset itself.

RMAA scores are pretty close. Note that 780i SLI board posts a higher overall score than the 680i-based EVGA board, though.