Nvidia’s nForce 780a SLI chipset

AMD launched its Phenom processor nearly six months ago as a part of an ambitious Spider platform that also included a new 790FX chipset and Radeon HD 3800 series graphics. As the first core logic chipset to bring PCI Express 2.0 to AMD processors, and the only one capable of taking advantage of Phenom’s HyperTransport 3 link and dual power planes, the 790FX was a particularly important part of the Spider platform. Unfortunately, it was saddled with an old and flaky south bridge chip, and motherboard makers were slow to adopt it.

Only a handful of boards based on the 790FX are currently on the market, which for AMD’s flagship processor platform is frankly a little lackluster—not that we’re surprised. Nvidia chipsets powered the bulk of the enthusiast-oriented motherboards for the Athlon X2, and the Phenom has been sorely in need of an nForce to call its own—an nForce 780a SLI, to be exact.

Nvidia gave us a first peek at the 780a at CES in January, revealing plans dripping with second-gen PCI Express, HyperTransport 3.0, and all the trappings you’d expect from a high-end Phenom chipset. The 780a also packs what Nvidia has taken to calling a motherboard GPU, enabling Hybrid SLI technology that promises significant power savings, HD video decoding capabilities, and additional monitor outputs.

We’ve been eagerly anticipating the 780a’s arrival for months, particularly since erratum-free B3 Phenoms are now out in the wild, the 780a has finally arrived astride Asus’ M3N-HT Deluxe motherboard. Keep reading for the goods on Hybrid SLI and an in-depth look at how the 780a stacks up against AMD’s 790FX.

Introducing the mGPU

By far the most novel element to the nForce 780a SLI is its graphics processor, which Nvidia refers to as the motherboard GPU, or mGPU. Putting a graphics processor inside a core logic chipset is hardly a new idea—low-end chipsets have been doing it for years. Nvidia’s approach with the 780a’s mGPU is quite different, though. Traditional integrated graphics chipsets are designed to provide basic functionality for budget systems without the need for a discrete graphics card. The 780a’s mGPU, however, is designed to work in tandem with discrete graphics products in Hybrid SLI configurations.

Before you get too excited about the potential for asymmetrical SLI, we should note that Hybrid SLI isn’t quite what it sounds like. Hybrid SLI is an umbrella term that covers two very different technologies: GeForce Boost and HybridPower. The former allows for, er, hybrid SLI configurations that team a motherboard GPU with a discrete graphics card to improve 3D performance. In order for GeForce Boost to improve performance, the motherboard GPU and discrete graphics card must be closely matched in terms of horsepower. Since integrated graphics processors are generally pretty weak, GeForce Boost only works with low-end graphics cards, making it a poor fit for a high-end platform like the nForce 780a SLI.

Hybrid SLI’s HybridPower component is far more relevant to the 780a because it’s designed specifically to reduce the power consumption of high-end graphics configurations. With HybridPower, the mGPU acts as the primary display adapter, and you plug your primary display right into the motherboard. If the system is idling or you’re just messing around on the desktop, not using applications that demand significant graphics resources, the motherboard GPU runs the show while the discrete graphics card lies dormant to conserve power. This power saving mode doesn’t rely on fancy clock throttling or special low-power states—it literally turns off the discrete graphics processor.

When additional graphics horsepower is required, HybridPower switches the system into performance mode, awakening the discrete graphics processor from its slumber. The discrete GPU then takes over graphics processing duties. Since it’s not actually hooked up to the system’s primary display, the discrete GPU still needs to cooperate with the motherboard GPU to put pixels on the screen. This task is achieved by copying the contents of the discrete GPU’s frame buffer over to the motherboard GPU’s frame buffer, which then outputs them to the display.

Nvidia says that PCI Express 2.0 provides more than sufficient bandwidth and low enough latency for seamless frame buffer replication, particularly because the discrete graphics card is typically rendering a few frames ahead of what’s actually being displayed on the screen. The company acknowledges that performance can suffer in certain situations, but insists that “aggregate performance loss is expected to be less than 5%.” Such a performance loss might be worth the potentially substantial power savings that HybridPower can provide, and we’ll test both a little later in this review.

In addition to the potential for reduced performance, HybridPower comes with some other caveats. The first, and perhaps most important, is that it requires a HybridPower-aware graphics card. The commands used to turn the discrete graphics processor on and off are sent over an SMBus that’s part of the PCI Express specification, but the graphics card has to know how to interpret those commands. To date, only Nvidia’s high-end GeForce 9800 GTX and GX2 graphics cards are HybridPower-aware. We can expect Nvidia’s high-end graphics products to be compatible with HybridPower moving forward, although the company seems less keen on moving the technology down into low-end parts that just don’t consume much power. HybridPower is also a Vista-only feature at the moment, and it requires that the motherboard GPU occupy at least 256MB of system memory.

Interestingly, the nForce 780a SLI’s motherboard GPU is much more than just a frame buffer middleman—it’s a fully functional DirectX 10-class graphics processor. The GPU core has the same architecture as the GeForce 8400 GS discrete GPU and features 16 stream processors. Clock speeds are actually a little higher than the GS, with the 780a’s mGPU core running at 500MHz and its shaders clocked at 1.2GHz. The mGPU even includes a PureVideo HD decode engine capable of handling “full” MPEG2, VC-1, and AVC high-definition video decoding duties for the Blu-ray and the now defunct HD DVD standards.

The inclusion of an HD video decode engine seems a little indulgent considering that Nvidia’s high-end graphics cards boast similar capabilities, but it makes perfect sense in a HybridPower context. Why spin up a discrete graphics card that’s only going to draw power and generate fan noise when movie playback can be handled by the mGPU? There are more limitations here, however. PureVideo HD requires a processor with a HyperTransport 3 interface because of the bandwidth requirements of HD video decoding, and its noise reduction filters won’t work if you’re running Vista’s Aero interface.

A whole new set of chips

The nForce 780a SLI’s motherboard GPU sits at the heart of a brand new MCP chip fabricated using 65nm process technology. Of course, this just wouldn’t be a new nForce without a little love from Nvidia’s silicon back-catalog, so the chipset also includes an nForce 200 chip first introduced with the nForce 780i SLI. The nForce 200 provides the bulk of 780a’s PCI Express 2.0 connectivity, serving up a total of 32 lanes split between one x16 link and a pair of x8s for three-way SLI. These lanes could also be split evenly between four x8 links, but Nvidia doesn’t yet offer an SLI configuration that requires four slots.

The nForce 200 packs a couple of features designed to improve performance with SLI configurations. The chip supports a posted-write shortcut that allows commands to be passed between graphics cards without hitting the processor or system memory, and a broadcast function enables a single command sent from the CPU to be replicated across all graphics cards, saving the CPU from having to generate and send multiple commands.

Both of these capabilities should reduce the number of commands needed to manage multiple graphics cards, decreasing congestion on the chipset’s link to the CPU. They should also trim traffic on the 16-lane PCI Express 2.0 link that connects the nForce 200 with the 780a SLI MCP. This link delivers a full 16GB/s of bi-directional bandwidth, which Nvidia says is plenty even for three-way SLI. However, it is worth noting that 16GB/s of interconnect bandwidth is only half of what would be needed to saturate all 32 of the nForce 200’s PCIe lanes.

In addition to the 16 lanes it has running to the nForce 200, the nForce 780a SLI MCP has three lanes of PCIe reserved for x1 slots and peripherals. That still leaves the 780a a few lanes short of the 790FX’s effective 38 lanes of PCIe 2.0 connectivity, but the gap closes to only two lanes when you consider that the AMD chipset will lose at least one lane to a Gigabit Ethernet controller. The 780a SLI has an integrated GigE MAC, so auxiliary networking controllers aren’t required. This single GigE controller is a bit of a step back for Nvidia, though; previous high-end nForce chipsets have included dual Gigabit Ethernet controllers.

To make chipset specs a little easier to digest, we’ve distilled them into a handy chart comparing the 780a SLI with AMD’s 790FX.

AMD 790FX Nvidia nForce 780a SLI
Processor interface 16-bit/2GHz HyperTransport 16-bit/2GHz HyperTransport

PCI Express 2.0 lanes
38 35

Multi-GPU support
CrossFire SLI

Chipset interconnect
PCIe 1.1 x4 PCIe 2.0 x16
Interconnect bandwidth 2GB/s 16GB/s
Serial ATA ports 4 6
AHCI Y Y
Native Command Queuing Y Y
RAID 0/1 Y Y
RAID 0+1/10 Y Y
RAID 5 N Y
ATA channels 2 1
Max audio channels 8 8
Audio standard AC’97/HDA HDA
Ethernet N 10/100/1000
USB ports 10 12

As you can see, both have a HyperTransport 3.0 processor interface and support for their own brand of multi-GPU graphics, be it CrossFire or SLI. Don’t pay too much attention to differences in chipset interconnect bandwidth, though. The 790FX uses a traditional north/south bridge layout, while the 780a’s MCP and nForce 200 chips are arranged in almost an opposite configuration. Nvidia needs all the interconnect bandwidth it can get to keep the nForce 200’s PCI Express lanes fed, but the peripherals integrated into the 790FX’s SB600 south bridge component aren’t nearly as bandwidth-hungry.

The 790FX’s older SB600 south bridge definitely looks overmatched in the storage department, where the nForce offers two more Serial ATA ports and support for RAID 5 arrays. Nvidia has added AHCI support to its latest storage controller—a feature that AMD still hasn’t managed to get working right with the 790FX.

12 USB ports round out the nForce MCP, giving the 780a another edge over the 790FX. The 780a also has full support for the Enthusiast System Architecture specification that Nvidia has proposed to the USB-if governing body.

For all you core logic geeks, here are a couple of close-up nudies of the 780a’s MCP and nForce 200 components.

The nForce 780a SLI MCP…

And its nForce 200 sidekick

Asus leads with the M3N-HT Deluxe

Unlike on the Intel side, where Nvidia’s recent chipset launches have been accompanied by retail motherboards of its own design, our first look at the nForce 780a SLI comes from Asus. The M3N-HT Deluxe is a decidedly high-end offering with a suggested retail price of $249. That’s certainly well within the range we’d expect from a premium mobo, but it’s a tricky proposition given the lack of a truly high-end Phenom processor to ride shotgun.

At least the Deluxe lives up to its luxury status. Asus uses solid-state conductive polymer capacitors throughout, which bodes well for the board’s longevity. Also, the six-layer PCB design features a “Stack Cool 2” layer of copper plate to aid heat dissipation.

On the layout front, the M3N-HT looks good overall, with power connectors located along the edges of the board right where cables won’t impede airflow around the CPU socket. There are no major clearance problems to report elsewhere, either.

Asus lines up a total of ten power phases next to the Deluxe’s AM2+ socket—eight phases for the processor and two for its memory and HyperTransport controller. The voltage regulation circuitry gets passive cooling, too, although with not nearly as much copper as we’re used to seeing on high-end motherboards.

A pair of heatpipes link the voltage circuitry cooler to a heatsink that sits atop the nForce 200 chip. From there, another heatpipe snakes down to the MCP. All of the heatsinks used are low-profile designs. They leave plenty of room around the CPU socket for larger coolers, but there isn’t quite enough clearance for the Scythe Ninja we use to test compatibility with gargantuan CPU coolers.

Asus also throws in a nifty memory heatsink that links right into the chipset cooler. This optional add-on is fully adjustable and easy to install, but it can only be used with two-DIMM configurations.

Moving south, Asus leaves plenty of room for longer graphics cards with a low-profile MCP cooler and edge-mounted IDE and Serial ATA ports. Heck, there’s even enough room between the DIMM slot retention tabs and the top PCI Express x16 slot to allow memory modules to be swapped with a graphics card installed.

We should note that edge-mounted SATA ports can conflict with extremely tight ATX enclosures that leave little room around the motherboard tray, but it’s a trade-off we’re willing to accept for additional graphics card clearance. Besides, if your case is too small to accommodate edge-mounted SATA ports, you probably won’t be able to squeeze in longer graphics cards like the GeForce 9800 GTX, which protrudes past the edge of the M3N-HT Deluxe by about an inch.

With Nvidia pushing three-way SLI, it’s only fitting that the M3N-HT features a trio of PCI Express x16 slots. The M3N-HT Deluxe also includes a single PCIe x1 slot and a couple of standard PCI slots. SLI threesomes, which are currently only possible with double-wide cards, will block access to all of those other expansion slots, though.

Just below the top PCI slot you can see a little riser card that contains a flash memory chip hooked into an onboard USB header. This chip contains a Linux-based Express Gate embedded operating system that provides web browsing, chat, and Skype applications that can be accessed when the system boots. Express Gate is an interesting little extra, but we’re not crazy about it cannibalizing one of the board’s USB ports. You can’t disable Express Gate and reclaim the lost USB port, either.

The M3N-HT Deluxe only has four USB ports on its outer edge, with internal headers for an additional six. Onboard headers are also available for two Firewire ports powered by an LSI chip, but there’s no 1394a connectivity in the port cluster. You do get an external Serial ATA port fed by a Marvell storage controller, though.

As one might expect, the Deluxe sports a full complement of analog and digital audio ports. Asus uses an Analog Devices codec to run the show, with digital S/PDIF outputs in both coaxial and TOS-Link flavors. The board lacks a digital audio input port, though.

Of course, the stars of the Deluxe’s port cluster are the VGA and HDMI outputs connected to the nForce 780a SLI’s motherboard GPU. The mGPU is capable of dual monitor output split between one analog and one digital display, and Asus provides an adapter to convert the HDMI output to DVI. However, DVI is limited to single link, whose maximum supported resolution is only 1900×1200—well short of the 2560×1600 native resolution of 30″ displays.

In addition to the HDMI-to-DVI adapter, the M3N-HT Deluxe comes with an optional chipset fan and a collection of front-panel connector blocks that make case wiring a snap.

If the M3N-HT Deluxe isn’t quite flashy enough for your tastes, Asus’ stable of nForce 780a SLI boards also includes the Crosshair II Formula. This mobo is even more upscale than the Deluxe, featuring fancy lighting effects, an external POST code display, an audio riser card, onboard power and reset buttons, and an external CMOS reset button in the port cluster.

The Crosshair II has a nice array of additional features and even comes bundled with a copy of Company of Heroes, but it suffers even more than the Deluxe from the lack of high-end Phenom processors. A motherboard may tie all of a system’s components together, but performance and core functionality don’t vary much between mid-range, high-end, and even ultra-premium products, making it hard to justify spending more on a mobo than a processor or graphics card. With a suggested retail price of $289, the Crosshair II Formula costs more than any Phenom that AMD makes—and more than plenty of potent graphics cards, too.

BIOS options galore

Asus is one of the best in the business at providing oodles of BIOS options, so it’s no surprise the M3N-HT Deluxe is packed to the gills with tweaking potential.


Bus speeds
CPU base clock:
200-600MHz in
1MHz increments

PCIe: 100-200MHz in 1MHz increments

DRAM: 400, 533, 667, 800, 1066MHz
HTNB: 200-2000MHz in 200MHz
increments


Bus multipliers
CPU: 5x-25x in 0.5x increments (Phenom
X4 9850)
CPU
NB: 5x-16x in 1x increments (Phenom X4 9850)
Voltages CPU:
0.775-1.55V in 0.0125V increments
CPU NB:
0.775-1.55V in 0.0125V increments

VDDA:
2.52, 2.622, 2.728, 2.83V

DRAM: 1.8-2.5V in 0.02V increments

NF200: 1.2-1.56V in 0.02V increments
HT: 1.2-1.5V in
0.02V increments
NB:
1.1-1.7V in 0.02V increments


Monitoring
Voltage, fan
status, and temperature monitoring

Fan speed control
CPU, chassis

Most Phenom overclockers will probably grab a Black Edition with an unlocked multiplier and take the easy route. The Deluxe has them covered with CPU multiplier control in 0.5X increments and a separate multiplier for the processor’s north bridge. If you want to kick it old-school, CPU base clock options are available between 200 and 600MHz in 1MHz increments. It’s even possible to adjust the speed and bus width of the HyperTransport link between the CPU and chipset, should you be so inclined.

Voltage adjustments are a key component to any overclocking attempt, and the M3N-HT’s BIOS has plenty on offer. CPU voltages range between 0.775 and 1.55V in 0.0125V increments, with separate settings for the CPU core and the north bridge. DRAM voltages extend up to 2.5V, alongside a full complement of chipset voltage options.

Speaking of memory, the Deluxe’s BIOS has all the basic timing settings you’d expect. Asus supplements these with a couple of pages worth of advanced timing control for those who wish to dig deeper into the Phenom’s integrated memory controller. Separating the basics from more advanced memory timing options makes perfect sense, since I suspect the vast majority of even hard-core enthusiasts probably won’t go beyond the fiddling with the big four memory timings and the DRAM command rate.

The BIOS exposes some Hybrid SLI controls, too. One may specify how much system memory is allocated to the motherboard GPU and whether it’s enabled at all. Keep in mind that you’ll need to dedicate at least 256MB of system memory to the mGPU for HybridPower to work.

We think fan speed control is just as important as overclocking, so we’re pleased to see the M3N-HT’s BIOS include temperature-based fan speed control for its processor and system fan headers. However, we’d like to see Asus provide more granular control over fan speeds and temperature targets; the Deluxe’s BIOS only lets users select from a couple of pre-defined profiles.

If you’re not comfortable poking around in the BIOS, you’ll be pleased to know that the M3N-HT Deluxe comes with Asus’ AI Suite and PC Probe tweaking, overclocking, and hardware monitoring software for Windows. Thanks to the nForce 780a SLI’s ESA support, the board is also compatible with Nvidia’s latest System Utility software. Well, sort of compatible, anyway.

We covered the new Nvidia system utility in great detail in our first look at ESA in action, so I won’t dwell too much on it here. Suffice to say that it’s by far the slickest system monitoring utility we’ve seen in the last few years. Unfortunately, the M3N-HT Deluxe doesn’t have the necessary BIOS hooks to take full advantage of the utility’s capabilities. The Nvidia control panel’s clock speed controls are intact, but its voltage and fan speed options don’t work. The Deluxe doesn’t report temperatures or voltages to the utility’s hardware monitoring component, either.

Sure, Asus replicates some of the missing Nvidia system utility functionality with its own software, but not the app’s ability to interface with other ESA-compliant components. Asus would do well to bring the Deluxe’s BIOS up to speed, but given its spotty support for Nvidia’s previous nTune apps, we’re not holding our breath.

Specifics on specifications

Should you prefer your motherboard specifications carefully compiled into a single chart, the goods on the M3N-HT Deluxe are provided below.


CPU support
Socket AM2/AM2+ Phenom,
Athlon X2 processors

North bridge
Nvidia nForce 200

South bridge
Nvidia nForce 780a SLI
MCP

Interconnect
PCI Express 2.0 (16MB/s)

Expansion slots
3 PCI Express x16

1 PCI Express x1
2 32-bit/33MHz PCI


Memory
4 240-pin DIMM
sockets

Maximum of 8GB of DDR2-533/667/800/1067 SDRAM


Storage I/o
Floppy disk

1 channel ATA/133

6 channels 300MB/s Serial ATA with RAID 0, 1, 0+1, 5 support

Audio 8-channel HD audio via Analog
Devices AD1988B codec
Ports 1 PS/2 keyboard

4 USB
2.0 with headers for 6 more

1 RJ45 10/100/1000
1
eSATA via Marvell 6111
Headers for 2 1394a Firewire via LSI FW322
1 VGA out
1 HDMI out

1 front out
1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in
1 digital TOS-Link S/PDIF out
1 digital coaxial S/PDIF
out

Our testing methods

We had numerous problems getting the 790FX chipset’s SB600 south bridge component working correctly in AHCI mode. Not only do you need an auxiliary storage controller (or a slipstreamed SP1 disc) to install Vista, but we’ve found that you also have to choose between drivers that offer strong performance with poor CPU utilization or those that exhibit low CPU utilization with weak performance. Given these issues, you’re better off running the SB600 in native IDE mode, which we did for our testing. The nForce 780a SLI has no problems running in AHCI mode, which is what we used for that platform.

This first wave of tests were run with HybridPower disabled. We’ll get to the mGPU in a moment.

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

Processor

Phenom X4 9850 Black Edition
System bus 2GHz HyperTransport

Motherboard


MSI K9A2 Platinum
Asus M3N-HT Deluxe
Bios revision 1.3 T0425A

North bridge
AMD 790FX Nvidia nForce 200

South bridge
AMD SB600 Nvidia nForce 780a SLI
MCP
Chipset drivers Catalyst 8.4 ForceWare 18.11
Memory size 2GB (2 DIMMs 2GB (2 DIMMs)

Memory type


Corsair CM2X2048-8500C5 DDR2 SDRAM
at 714MHz
CAS latency
(CL)
4 4
RAS to CAS
delay (tRCD)
4 4
RAS precharge
(tRP)
4 4
Cycle time
(tRAS)
12 12
Command rate 2T 2T

Audio codec
ALC888
with 1.91 drivers
AD1988B with 6.10.1.6320
drivers
Graphics

GeForce 8800 GT 1GB PCIe
with ForceWare 169.25 drivers
Hard drive
Western Raptor X 150GB
OS

Windows Vista Ultimate x86
with Service Pack 1

Thanks to Corsair for providing us with memory for our 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.

We used the following versions of our test applications:

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

Even sharing the same integrated Phenom memory controller and identical memory timings, there’s a slight difference in memory bandwidth between our 790FX and 780a boards. The nForce pulls up a little short, but since the memory controller is located on the processor die, that’s more a knock on Asus’ memory controller tuning than on the 780a chipset.

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.

Our 790FX board extends its bandwidth lead over the 780a with four DIMMs installed, and it steals a win in the latency test. MSI has had more time to tweak the memory performance of its 790FX board, though, and I suspect Asus’ 780a will catch up in a BIOS revision or two.

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. Yellow represents L1 cache, light orange is L2, red is L3, and dark orange is main memory.

Same integrated Phenom memory controller, same access latency picture. Moving along…

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.

No doubt thanks to its slight advantage in memory bandwidth, our 790FX system pulls out ahead of the 780a in Euler3d. The two are pretty close, though.

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.

The tables turn in WorldBench, with the 780a edging out its AMD competition by one point. Let’s break down WorldBench’s individual application test scores to see if we can find out why.

The nForce is quicker in WorldBench’s multimedia editing and encoding tests, although not by huge margins.

AMD bounces back in the multitasking test and ekes out a slim victory with Office 2003, but the nForce takes the Firefox test. Again, scores are close.

3ds max favors the nForce, but not by much.

Nero’s a wash, but the nForce comes out ahead in WinZip, where the 780a is 9% quicker than the 790FX—its biggest WorldBench margin of victory.

Gaming

Frame rates are pretty close across our quartet of gaming tests, with only Crysis teasing out much of a difference between the 780a and 790FX. In that test, the nForce trails by a couple of frames per second.

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.

The 790FX’s utterly broken AHCI mode essentially strips the platform of its support for Native Command Queuing—a shortcoming that IOMeter nicely highlights. As soon as we go beyond a single I/O request, the nForce’s transaction rate scales much better than that of the 790FX with both the workstation and database test patterns.

IOMeter response times are quicker with the 780a, too.

And quicker disk performance doesn’t come with a CPU utilization penalty. Both chipsets are well below 0.5% CPU utilization across all load levels, with the nForce just that little bit more frugal.

HD Tach

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

The nForce’s burst speed is a little quicker than that of the 790FX, but the big difference between the two comes in HD Tach’s write speed test. This test tends to run much faster on platforms that support Native Command Queuing, which would explain why the 780a has such a big lead.

A 0.1-millisecond difference in random access time doesn’t amount to much.

HD Tach’s CPU utilization results are well within the app’s +/- 2% margin of error for 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.

The 790FX’s aging SB600 south bridge has long-standing USB performance issues that comparison with 780a makes all too clear. Not only does the nForce achieve higher burst, read, and write speeds, but it does so with lower CPU utilization, 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.

There may only be one of them, but the nForce 780a SLI’s integrated Gigabit Ethernet controller is pretty spiffy. Throughput is solid, and CPU utilization is very low. That’s more than can be said for our 790FX board, whose Realtek GigE controller sucks up nearly three times the CPU cycles of the nForce 780a.

To be fair, it’s up to motherboard makers to choose which GigE controller is paired with the 790FX. We’ve seen better Ethernet performance from other 790FX boards, but that kind of variance shouldn’t affect the nForce 780a as long as motherboard makers correctly implement its embedded networking component.

PCI Express performance

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

Throughput is comparable here, but the 780a consumes slightly more CPU cycles than the 790FX.

PCI performance

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

Here the nForce has a slight throughput advantage and a bit of an edge in CPU utilization. Scores are very close, though.

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.

We’ll get to HybridPower in a moment, but this first look at 780a power consumption isn’t terribly encouraging. Our 780a board consumes quite a bit more juice than one based on the 790FX, regardless of whether Cool’n’Quiet is enabled. The difference in power consumption shrinks under load, but even then, the nForce platform pulls an additional 13 watts.

Nvidia’s nForce chipsets have never been particularly power-efficient, so these results aren’t all that shocking. However, we’re not inclined to pass judgment on the chipset’s power consumption just yet. Remember that this is an Asus M3N-HT Deluxe motherboard we’re dealing with here, not an Nvidia reference design. We’ve seen higher-than-normal power consumption from Asus motherboards in the past, most recently with the company’s initial P35 offerings.

Overclocking

Phenom processors can be overclocked either by adjusting the CPU multiplier or tweaking the CPU reference clock. The latter should work for any Phenom CPU, since an unlocked multiplier that can be adjusted upwards isn’t needed. That’s where we started, first dropping the CPU and memory multipliers to take them out of the equation, and then turning the screws on the CPU reference clock. Stability was tested along the way with Prime95 crunching on all four cores.

Unfortunately, we couldn’t get the system stable with a CPU base clock higher than 215MHz. 220MHz produced Prime95 errors and even the occasional blue screen, and no amount of voltage fiddling or playing with the CPU north bridge multiplier seemed to help.

Of course, if you’re serious about Phenom overclocking, you should really be using a Black Edition CPU with an unlocked upper multiplier.

Just by adjusting the CPU multiplier, we were able to get our Phenom stable at up to 2.8GHz with only a slight bump to the CPU voltage. Additional voltage wouldn’t allow our processor to boot with a 14.5X multiplier, though.

We haven’t been particularly impressed with Phenom overclocking thus far, and given the affordability of the unlocked Black Edition, your success isn’t likely to be dependent on the chipset or motherboard. As is always the case with overclocking, though, your mileage may vary.

HybridPower in action

Our initial nForce 780a SLI testing was conducted with a GeForce 8800 GT graphics card that isn’t compatible with Hybrid SLI. To test Hybrid SLI—more specifically, its HybridPower component—we swapped in a GeForce 9800 GTX with HybridPower-aware ForceWare 174.14 drivers.

Before jumping into some benchmark results, I should spend a moment explaining how HybridPower works from a user perspective. Or at least how it’s supposed to work. HybridPower actually consists of three modes: save power, boost performance, and additional displays. The first two are self-explanatory, and Windows is supposed to switch between them automatically based on the system load. Automatic mode switching didn’t work right on our system, though, regardless of which Vista power plan we tried. I suspect the combination of beta graphics drivers and the fact that we previously had a GeForce 8800 GT with an older graphics driver revision installed might have gummed things up, but who knows.

Fortunately, HybridPower has a provision for manual mode switching via a handy taskbar icon, and that worked just fine. Switching between modes is flicker-free and only takes a couple of seconds with an empty desktop. However, the more windows you have open, the longer it will take to switch modes. Also, you can’t switch modes if an open application is using the 3D subsystem. That might not seem like much of a hindrance—you aren’t likely to switch modes while playing games—but many desktop apps make use of a graphics card’s 3D horsepower. Paint Shop Pro Photo X2, for example, can’t be open if you want to switch HybridPower modes. Vista’s Aero interface isn’t a problem, though.

Another minor annoyance related to mode switching is its propensity to center the mouse pointer. This is a minor problem at best, and one that Nvidia should be able to rectify.

If you’re not interested in HybridPower’s energy-saving potential, an “additional displays” option allows the motherboard’s display outputs to be used in conjunction with those of a discrete graphics card to power multiple displays. This isn’t a particularly new capability for integrated graphics, but it is a separate option that effectively disables power saving mode.

So what kind of impact does HybridPower have on system performance and power consumption? We whipped up a quick batch of benchmarks to find out. Performance tests were run with HybridPower in boost performance mode to gauge the impact of frame buffer replication to the motherboard GPU.

HybridPower slows our system down by a hair, but that’s about it. All in all, not bad for a first attempt, especially when you consider the potential for power savings.

Our idle power consumption tests were run in power-saving mode, and the results speak for themselves. With a GeForce 9800 GTX, HybridPower saves a whopping 57W. Impressive.

Under load, where we tested with HybridPower in boost performance mode, power consumption is actually higher. I suspect at least some of the extra power draw can be traced to the additional processing required for frame buffer replication. Also, the 780a’s motherboard GPU slips into a low-power state when HybridPower is disabled, reducing its power consumption.

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. To provide a closer look at the peripheral performance you can expect from the motherboards we’ve tested today, we’ve complied Firewire and audio performance results below.

HD Tach
Firewire performance

Read burst

speed (MB/s)


Average read

speed (MB/s)


Average write

speed (MB/s)


CPU utilization

(%)


Asus M3N-HT Deluxe
42.3 37.6 28.6 1.0

MSI K9A2 Platinum
41.2 37.2 25.5 0.7

The Asus board’s Firewire chip has faster write speeds than the one used on MSI’s K9A2 Platinum.

RightMark Audio
Analyzer audio quality

Overall score

Frequency response

Noise level

Dynamic range

THD

THD + Noise

IMD + Noise

Stereo Crosstalk

IMD at 10kHz

Asus M3N-HT Deluxe

4

5

3

3

3

1

3
5
3

MSI K9A2 Platinum
4 5 3 3 3 1 3 4 3

There’s very little difference between these boards on the audio front.

Conclusions

Hybrid SLI is an interesting new direction for Nvidia, one the company intends to pursue aggressively by integrating motherboard GPUs into all its new chipsets. We really like the idea. Although its initial implementation in the nForce 780a SLI has a few rough edges, HybridPower’s reduction in energy use is dramatic enough we’re willing to forgive a few flaws. The problems we experienced with HybridPower are more annoyances than show-stoppers, and they don’t take away from what is an otherwise impressive technical achievement.

In fact, we’re so keen on HybridPower that we’d like to see it extended to all of Nvidia’s graphics products. High-end cards like the GeForce 9800 GTX and GX2 may have the most to gain from being powered down, but there’s no reason why budget cards like the GeForce 9600 GT shouldn’t benefit, as well. We are living in an age of soaring energy prices and increasing environmental awareness, after all.

While we’re griping, addressing the single-link DVI limitation of the 780a’s motherboard GPU should be high on Nvidia’s list of priorities. 1900×1200 might be a reasonable resolution ceiling for most folks, but since HybridPower is currently restricted to high-end graphics configurations, it really needs to be able to handle 30″ displays at their native 2560×1600 resolution. Perhaps the most compelling case for HybridPower can be made for SLI systems with multiple high-end graphics cards, and with that kind of pixel pushing horsepower, you really want to be driving a 30″ panel at full resolution.

Of course, there’s more to the nForce 780a SLI than Hybrid SLI. Even if you discount its mGPU, the 780a is still the best Phenom chipset on the market. With more USB and SATA ports, Native Command Queuing that actually works, support for RAID 5 arrays, and an integrated Gigabit Ethernet controller that should prevent motherboard makers from using sub-par networking chips, we heartily recommend the nForce over AMD’s existing 790FX chipset—that is, of course, if you can find a board at a decent price with reasonable power consumption.

Power consumption has never been a strong suit of nForce chipsets, but the wattage pulled by Asus’ M3N-HT Deluxe is a little alarming. We’re also not too crazy about the board’s $249 price tag, which makes it more expensive than the fastest Phenom you can buy. AMD simply doesn’t have a CPU that’s fast enough to put in a high-end rig. That’s hardly the fault of Asus or Nvidia, but it does blunt our enthusiasm for the M3N-HT Deluxe and the nForce 780a SLI.

Fortunately, Nvidia has an ace up its sleeve with a mid-range nForce 750a SLI chipset. The 750a loses three-way SLI compatibility and only has 19 lanes of PCI Express 2.0. But it supports Hybrid SLI, and we have a feeling it uses the exact same MCP chip as the 780a. What’s more, the 750a should arrive on affordable motherboards that are a better fit for the Phenoms AMD currently has on the market. That should tide us over until Hybrid SLI makes its way to nForce chipsets for Intel processors, which is where we’re really dying to see the technology in action.

Comments closed

Pin It on Pinterest

Share This

Share this post with your friends!