review nvidias nforce 790i sli ultra chipset
Reviews

Nvidia’s nForce 790i SLI Ultra chipset

Barely three months have passed since Nvidia launched its high-end nForce 780i SLI chipset, and already the green team has another round in the chamber. To be fair, the 780i wasn’t so much a new design as a second-coming of the nForce 680i decked out with PCI Express 2.0 support via a custom SLI bridge chip. The 780i wasn’t the most elegant answer to the 32 second-generation PCI Express lanes lurking in Intel’s X38 Express chipset, but it was the best Nvidia could do at the time to keep up with the Joneses. Today, however, Nvidia’s real next-generation core logic arrives in the nForce 790i SLI and its spoiler-equipped Ultra variant.

Unlike its predecessor, the nForce 790i SLI has an all-new north bridge chip bristling with 32 lanes of bandwidth-rich PCIe 2.0 goodness. Within it also lies Nvidia’s first DDR3 memory controller, which has been carefully tweaked with Penryn in mind. The 790i has native support for 1600MHz front-side bus speeds, too. And, of course, everything comes neatly packaged on a fresh motherboard design tuned for overclocking and packed with ESA functionality.

This is Nvidia’s answer to the X48 Express—the very flagship of Intel’s chipset aristocracy—leading us to a heavyweight bout for high-priced bragging rights and technical superiority. Keep reading to see if Nvidia has managed to unseat the incumbent.

Will the real 7-series nForce please stand up
Nvidia has made a habit of recycling older chipset components, so the nForce 780i SLI’s 680i underpinnings and bodged PCI Express 2.0 functionality weren’t much of a surprise. But the 790i is different. It’s fresh silicon, albeit fabbed on a 90nm process that’s starting to look a little antiquated next to the 55 and 65nm tech used to fab AMD’s latest chipsets. Gone is the 780i’s LinkBoosted PCI Express 1.1 connection to an auxiliary PCIe 2.0 bridge chip, and in its place, 32 lanes of second-generation PCI Express in all its 5.0GT/s glory.

This native PCI Express implementation inherits a couple of SLI-specific optimizations from the nForce 200 bridge chip, as well. Among those functions is a posted-write shortcut that allows the PCI Express controller to pass messages directly between graphics cards without having to hit main memory. A broadcast function also allows the PCIe controller to take a single command from the CPU and distribute it across multiple graphics cards, saving the CPU from having to generate multiple commands for multiple cards while reducing congestion on the front-side bus.

Such SLI optimizations are a welcome addition to the 790i’s PCIe controller, but they’ll only affect multi-GPU configurations whose graphics cards are connected directly to the north bridge. That’s fine if you’re only going to be running a two-way setup. However, more exotic SLI threesomes will be forced to hang a third graphics card off the 790i’s south bridge component. The 790i south bridge’s first-gen PCI Express controller is decidedly more pedestrian, not only losing out on posted-write shortcut and broadcast functionality, but also offering signaling rates of only 2.5GT/s per lane. Mind you, with the GeForce 9800 GX2 effectively enabling four-way SLI using only two graphics cards slots, there may not be much demand for three-way setups.

To keep up with Intel’s upcoming Core 2 Extreme QX9770 processor, the 790i SLI chipset also features native support for a 1600MHz front-side bus. This isn’t the most exciting development given that the 780i and even 680i easily overclocked to similar speeds, but it’s a must-have for any new high-end chipset’s resume.

Speaking of other high-end chipsets, the 790i SLI actually comes in standard and Ultra flavors. Both are based on the same north bridge silicon, with Nvidia binning particularly speedy samples for use as Ultras. This cherry-picking will probably ensure that Ultra chips offer the best overclocking potential, although the only official difference between the two variants is in their support for higher memory bus speeds. The standard 790i SLI can handle DDR3 memory up to 1333MHz, while the Ultra boasts support for memory bus speeds up to a whopping 2000MHz.

X48
Express
X38
Express
P35 Express nForce 790i SLI Ultra SPP nForce 780i SLI SPP

Front-side bus
1600/1333/1066MHz 1333/1066MHz 1333/1066MHz
1600/1333/1066MHz
1333/1066MHz

Memory controller
DDR2-800
DDR3-1600
DDR2-800
DDR3-1333
DDR2-800
DDR3-1333

DDR3-2000*
DDR2-800

PCI Express 1.1 lanes
0 0 16
2
0

PCI Express 2.0 lanes
32 32 0
32
32

Multi-GPU support
CrossFire CrossFire CrossFire*
SLI
SLI

Chipset interconnect
DMI DMI DMI
HyperTransport
HyperTransport

Peak interconnect bandwidth
2GB/s 2GB/s 2GB/s
8GB/s
8GB/s

DDR3-2000 support sounds impressive, and it is, but we should qualify it with a couple of limitations. First, the 790i SLI only supports memory speeds beyond 1800MHz with one module per channel, and then only in a motherboard’s second and fourth DIMM slots. This “0101” memory configuration is necessary at high speeds because of how Nvidia’s reference motherboard design for the Ultra configures slot addressing; it may not be a problem for custom board designs from other manufacturers.

The nForce 790i SLI marks Nvidia’s first attempt at a DDR3 memory controller, so hitting 2000MHz is impressive regardless of the associated limitations. Nvidia boasts that the 790i can handle 2000MHz with memory voltages as low as 1.9V—only a modest boost over DDR3’s default voltage of 1.5V. Even at more sedate speeds, the 790i’s memory controller is tuned for performance. New prefetch algorithms within the memory controller’s Dynamic Adaptive Speculative Preprocessor (DASP) are specifically optimized for the larger L2 caches found in Intel’s new Penryn processors. Nvidia also claims the 790i’s memory controller has the lowest latency data path for memory reads and writes.

As one might expect, Nvidia has brought its open Enhanced Performance Profile (EPP) spec to DDR3 memory. EPP 2.0 offers the same extensions to standard SPD settings as the original Enhanced Performance Profile spec, this time for DDR3 memory instead of DDR2.

Like so many Nvidia chipsets before it, the 790i SLI uses a HyperTransport interconnect to link its north and south bridge components. Although it’s been used for years, this link offers a healthy 8GB/s of interconnect bandwidth—four times what’s available with Intel’s desktop chipsets. The extra bandwidth is really only necessary for three-way SLI implementations, so for most systems, it simply provides a massive pipe down to the south bridge.

It’s at the south bridge that we find a familiar face in the form of the nForce 570 SLI MCP. Nvidia may call this an nForce 790i SLI Ultra MCP in its block diagrams, but the chip’s markings clearly reveal its true identity. The 570 SLI first made an appearance back in 2006 as the south bridge component of the nForce 590 SLI chipset for Socket AM2. In the years since it’s been passed around like a bong at a Grateful Dead concert, pulling duty in the nForce 570, 590, 590i, 680i, and 780i.


ICH9R

nForce 790i SLI MCP

nForce 780i SLI MCP

PCI Express 1.1 lanes
6
28
28

Serial ATA ports
6
6
6

Peak SATA data rate
300MB/s
300MB/s
300MB/s

AHCI
Y
N
N

Native Command Queuing
Y
Y
Y

RAID 0/1
Y
Y
Y

RAID 0+1/10
Y
Y
Y

RAID 5
Y
Y
Y

Matrix RAID
Y
N
N

ATA channels
0
1
1

Max audio channels
8
8
8

Audio standard
HDA
HDA
HDA

Ethernet
10/100/1000
2 x 10/100/1000
2 x 10/100/1000

USB ports
12
10
10

Even at almost two years after its first release, the nForce 790i 570 SLI MCP’s feature set stands up reasonably well against the best of Intel’s south bridge offerings. Nvidia’s MCP chips have always packed more functionality than rival designs from Intel, and with 28 PCI Express 1.1 lanes and dual Gigabit Ethernet controllers with packet prioritization and teaming capabilities, the 570 SLI is no exception. Things are more even on the storage front, where the 570’s six Serial ATA RAID ports neatly match those in the ICH9R.

The 570 begins to show its age in the USB department, offering a mere 10 ports to the ICH9R’s 12. However, the Nvidia MCP’s age ends up paying dividends in the form of an ATA controller that the ICH9R lacks. IDE ribbons may be rare these days, but they’re not quite so uncommon that motherboards have banished “parallel” ATA altogether. With the nForce 790i SLI, mobo makers won’t have to resort to third-party storage controllers for a fix of old-school ATA lovin’.

A new reference motherboard design
As has become customary for Nvidia chipset launches, the nForce 790i SLI Ultra rolls out on a new reference motherboard that will be sold as-is by the likes of XFX, EVGA, and others. The board is a six-layer design, and at least in Ultra garb, it’s expected to sell for $350 and up. Only those with deep pockets need apply.

Motherboards in this price range essentially need to be perfect, and the 790i reference design gets off to a good start. Dressed in blacks and greys with only a splash of green on its PCI Express x16 slots, the board looks appropriately menacing and definitely distinctive. More importantly, the layout puts everything in its right place, all but eliminating clearance issues and the potential for excessive cable clutter.

Around the socket, six power phases feed the CPU. Nvidia also says it has optimized traces to the processor to ensure that the actual data paths (and not just physical board traces) are all exactly the same length.

Not that you’re going to notice little traces when there’s a virtual mountain range of chipset and voltage regulation circuitry cooling ringing the socket. This generous array of nickel-plated heatsinks is the most expansive we’ve seen coupled with an nForce chipset. Despite cutting an imposing profile, it does a remarkably good job of staying out of the way. We were even able to squeeze a massive Ninja CPU heatsink onto the socket without interfering with the chipset cooler.

When running at stock speeds, the chipset heatsinks can easily get by with silent, passive cooling, even on an open test bench with little ambient airflow. Nvidia also includes a snap-on cooling fan that’s recommended for overclocking. The fan plugs right into the motherboard, and although you can adjust its speed through the BIOS, it doesn’t benefit from temperature-based automatic fan speed control.

nForce chipsets have a reputation for running a little hot, so it’s encouraging that the auxiliary fan is only necessary for more extreme overclocking. The metal bracket that holds the fan in place could use a little work, though; it snapped off when we installed the fan, and although that doesn’t compromise the integrity of the chipset cooler, no one wants little bits of their motherboard falling off.

Below the chipset cooler’s fan mount lies a vapor chamber that extends down to the south bridge. Designed for efficient heat transfer from a notoriously toasty MCP chip, this link is lined with low-profile cooling fins that provide plenty of clearance for longer graphics cards. You don’t have to worry about gargantuan expansion cards interfering with onboard ATA or Serial ATA ports, either—they’re all located along the edge of the board or otherwise tucked out of the way.

Nestled between the board’s edge-mounted storage ports we find a two-digit POST code display that takes the beep-code guesswork out of troubleshooting boot problems. Nvidia also throws in handy onboard reset and power buttons for those who build or test systems outside standard enclosures.

Those onboard buttons sit at the bottom of a generous slot stack that boasts three PCI Express x16 slots. The first and third slots have PCIe 2.0 lanes, while the second is a gen-one slot connected to the south bridge. This arrangement nicely spaces two-way SLI configurations, providing plenty of breathing room even with double-wide cards. Those double-wide SLI configurations will also leave users with free PCI and PCIe x1 slots in addition to the third x16 slot.

Atop the slot stack, you can just make out a red Serial ATA port. This is the board’s seventh internal SATA port, and it’s connected to an auxiliary JMicron JMB362 SATA controller. Nvidia amusingly suggests the placement of this port is ideal for optical drives, which seems odd, because the board has four other internal SATA ports that are much closer to where optical drives would be mounted in most enclosures.

A more likely rationale for the red port’s placement is the fact that the JMB362 controller is also responsible for External Serial ATA connectivity on the port cluster. There, we find a single eSATA port—the first on an Nvidia motherboard design. With six internal SATA connectors already fed by the chipset, Nvidia might have been better off making both of the JMicron controller’s SATA ports external.

eSATA connectivity nicely fills out a port cluster loaded with all the staples, including coaxial and TOS-Link flavors of digital S/PDIF audio output. Thanks to a Realtek ALC888S codec chip, it’s possible to output multi-channel DTS Connect audio through those digital outs.

In addition to the external ports, onboard headers are available for an extra Firewire and four more USB ports. These headers are particularly important for the Enthusiast System Architecture specification, which relies on USB connections to talk to compliant components.

BIOS options galore
Nvidia intends the nForce 790i SLI Ultra to be the ultimate Core 2 overclocking platform, so it’s endowed the BIOS with a full complement of tweaking options.


Bus speeds
FSB: 400-2800MHz in
1MHz increments
PCIe x16 1&2: 100-200MHz in 1MHz increments
PCIe x16 3: 100-200MHz in
1MHz increments
DRAM: 400-2500MHz in 1MHz increments
Interconnect ref: 200-500MHz in 0.5-2MHz increments

Bus multipliers
CPU: 6x-8x (Core
2 Duo E6750)

SPP->MCP: 1x-5x
MCP->SPP: 1x-5x
Voltages CPU: 0.5125-2.275V in 0.00625V increments
DRAM: 1.5-2V in 0.025V increments
CPU FSB: 1.2-1.55V in 0.0.5V increments
SPP: 1.3-1.55V in
0.05V increments
MCP:
1.5-1.75V in 0.05V increments
GTLVREF lane 0, 1, 2, 3: +/-0.155V in 0.005V increments

Monitoring
Voltage, fan
status, and temperature monitoring

Fan speed control
CPU, chassis

Effective front-side bus speeds are available up to a dizzying 2800MHz, nicely complementing a maximum effective memory speed of 2500MHz. The BIOS lets users lock down the PCI Express clock between 100 and 200MHz, complete with independent control over PCIe links stemming from the chipset’s north and south bridge components. One can set arbitrary interconnect reference clocks and multipliers, as well, preventing front-side bus or memory overclocking from messing with the HyperTransport link.

In the voltage department, the BIOS can scale the CPU core voltage between 0.5125V and 2V in ultra-fine 0.00625V increments. DRAM voltages are supported up to 2.275V, as well, and users are free to manipulate front-side bus and chipset voltages. The BIOS is even capable of automatically adjusting voltages in response to user-set system and memory bus speeds. In fact, Nvidia recommends that overclockers first give the BIOS a shot at adjusting voltages on its own before they get theirs hands dirty with manual settings.

A full suite of memory timing options make an appearance in BIOS, too. If you’re running EPP 2.0-compatible modules, the BIOS will automatically set the correct timings and memory bus speed. In fact, the BIOS was smart enough to detect and set the 1.9V memory voltage called for by our EPP 2.0 DDR3-2000 DIMMs.

The good times keep rolling when we move to fan control, where the 790i SLI BIOS offers automatic temperature-based fan speed control for the processor and chassis fan headers. Here, users can set not only high and low temperature thresholds, but also the fan speeds that correspond to each. Arbitrary fan speed control is also available for three additional onboard fan headers.

Old-school enthusiasts love having all of a motherboard’s overclocking and tweaking options consolidated in the BIOS, but for those who want to fiddle from within the comforting confines of Windows, Nvidia’s latest nForce System Tools software does the trick. This handy collection of applications provides BIOS and driver update functionality alongside a healthy selection of tweaking, overclocking, and hardware monitoring options. The latest System Tools revisions are also fully ESA-aware, allowing users to monitor and control components that comply with the Enthusiast System Architecture spec.

ESA is a whole other topic unto itself, and you can read more about it in our initial take on the specification and subsequent first look at components and software in action.

Specifics on specifications
If you prefer your motherboard specifications lovingly distilled into a single chart, we’ve whipped up one below. Note that apart from an auxiliary JMicron controller, Realtek codec, and Firewire chip, all the board’s integrated peripherals are fed by the chipset.


CPU support
LGA775-based
Celeron, Pentium 4/D, Core 2 processors

North bridge
NVIDIA nForce 790i SLI SPP

South bridge
NVIDIA nForce 790i SLI MCP

Interconnect
HyperTransport (8GB/s)

Expansion slots
3 PCI Express x16
2 PCI Express x1
2 32-bit/33MHz PCI

Memory
4 240-pin DIMM
sockets
Maximum of 8GB of DDR3-1066-2000* SDRAM

Storage I/o
Floppy disk
1 channel ATA/133
6 channels 300MB/s Serial ATA with RAID 0, 1, 10, 5 support
1 channel
300MB/s Serial ATA via JMicron JMB362
Audio 8-channel HD audio via Realtek ALC888S codec
Ports 1 PS/2 keyboard
1 PS/2 mouse
6
USB
2.0 with headers for 4 more

2 RJ45 10/100/1000
1 1394a Firewire via
Texas Instruments TSB43AB22A with header for 1 more
1 eSATA via JMicron JMB362

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
Since the 790i SLI Ultra is Nvidia’s first shot at DDR3, we thought it only fitting to round up a trio of DDR3-equipped Intel chipsets for comparison. We’ve also thrown in the 790i’s predecessor, the nForce 780i SLI, for a little DDR2 flavor. Keep in mind that all three Intel chipsets use the same ICH9R south bridge chip and that both nForces are equipped with what is essentially the same 570 SLI MCP. That should keep peripheral performance relatively consistent across Intel’s Express lineup and between our selections from the nForce family.

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

Processor

Core 2 Duo E6750 2.67GHz
System bus 1333MHz (333MHz
quad-pumped)

Motherboard


Asus P5K3 Deluxe


Asus P5E3 Deluxe
Gigabyte X48T-DQ6 XFX
nForce 790i SLI Ultra


XFX MB-N780-ISH9
Bios revision 0604 0201 F4D 811N1P01 2.053.B2

North bridge
Intel P35 Express Intel X38 Express Intel X48 Express nForce 790i
SLI Ultra SPP
nForce 780i SLI SPP

South bridge
Intel ICH9R Intel ICH9R Intel ICH9R nForce 790i
SLI MCP
nForce 780i SLI MCP
Chipset drivers Chipset 8.3.1.1009
AHCI 7.8.0.1012
Chipset 8.3.1.1009
AHCI 7.8.0.1012
Chipset 8.3.1.1009
AHCI 7.8.0.1012

ForceWare 9.64
ForceWare 9.46
Memory size 2GB (2 DIMMs) 2GB (2 DIMMs) 2GB (2 DIMMs) 2GB (2 DIMMs) 2GB (2 DIMMs)

Memory type


Corsair CM3X1024-1600C7DHX DDR3 SDRAM
at 1333MHz


Corsair CM2X2048-8500C5 DDR2 SDRAM
at 800MHz
CAS latency
(CL)
7 7 7 7 4
RAS to CAS
delay (tRCD)
7 7 7 7 4
RAS precharge
(tRP)
7 7 7 7 4
Cycle time
(tRAS)
21 21 21 21 12
Command rate 1T 1T 1T 1T 1T

Audio codec
Integrated

AD1988B with 5.10.1.6110 drivers

Integrated AD1988B with
7.0.0.0 drivers
Integrated ALC889A with 1.88
drivers
Integrated ALC888S
with 1.88 drivers
Integrated ALC888S
with 1.88 drivers
Graphics

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

Windows Vista Ultimate x86
with KB936710, KB938194, KB938979, KB940105
updates

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. 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:

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 the best arena for Nvidia’s to show off its new DDR3 memory controller.

And what a memory controller it is. Despite running the same memory speed and timings as our other DDR3 platforms, the 790i SLI offers more bandwidth and significantly lower latencies than Intel’s best. The Ultra’s 5.4-nanosecond access latency advantage might not sound like much, but within a modern PC where bits flip at gigahertz speeds, it’s an impressive margin.

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 780i SLI—a common adjustment for most four-DIMM configurations.

The 790i SLI handles additional DIMMs with aplomb, retaining the top spot in both our memory bandwidth and latency tests.

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’s claim that the nForce 790i SLI’s memory controller has the lowest latencies around certainly looks legit when we spread CPU-Z’s access latency results across three dimensions. Note how much quicker the Ultra is with higher step and block sizes than its rivals.

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.

Despite a huge latency advantage, the 790i SLI can’t quite match the X48 Express in this memory-intensive fluid dynamics simulation.

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.

WorldBench sees the Ultra return to the head of the class, although it’s joined there by our X38 Express platform. Let’s break down WorldBench’s individual application test results to see if the Ultra has any weaknesses.

Scores are tight through the suite’s multimedia editing and encoding tests, with the 790i essentially tying its competition.

Interestingly, the Ultra is a little slower than Intel’s Express onslaught in the Firefox tests, which typically favor faster memory subsystems.

3ds max performance doesn’t vary much between the chipsets.

Scores in Nero and WinZip do, although we should note that Vista’s disk caching mechanism appears to impact performance in both tests. The 790i SLI is the fastest chipset we’ve seen here, but I wouldn’t expect its rivals to be much slower in the real world with these applications.

Gaming

Even at relatively low resolutions, today’s games tend to be bound by the graphics card and CPU rather than the chipset. The 790i SLI doesn’t take top honors across the board here, but it does come out ahead in two of three games. Only Half-Life 2: Episode 2 proves problematic for the Ultra, and even then, it’s just a few frames per second off the pace set by Intel’s X48 Express.

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.

Our nForce 780i and 790i platforms are using slightly different driver revisions, which would explain the minor discrepancy in IOMeter performance between them. Otherwise, the nForce chipsets do a good job of keeping up with Intel. The ICH9R boasts slightly higher transaction rates with a couple of load levels, though.

Response times are even practically across the board, with the Ultra only falling a little behind under the most demanding load.

We don’t see much CPU utilization for any of the chipsets.

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

The 790i SLI’s SATA read speeds are quite competitive, but with our Raptor hard drive, write performance is a little off that of Intel’s best. HD Tach’s write speed test tends to produce odd results when chipsets that support Native Command Queuing are combined with the WD1500ADFD, though.

Barely a fraction of a millisecond separates random access times…

And CPU utilization is within HD Tach’s +/- 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.

Although its USB read performance isn’t any faster than what you get with Intel’s latest chipsets, the 790i’s USB writes are slightly quicker. This does come with a slightly higher CPU utilization penalty, although one that’s within HD Tach’s margin of error.

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.

We’ve always liked how the integrated Gigabit Ethernet controllers in Nvidia chipsets have largely kept motherboard makers from skimping on components and equipping boards with either slow PCI-based GigE chips or ones with brutally high CPU utilization. Of course, that was when Nvidia’s GigE implementation had comparable throughput and lower CPU utilization than the best auxiliary Gigabit chips on the market. That’s just not the case with the 790i SLI, whose Ethernet throughput hits a wall at around 830Mbps—a good 100Mbps short of the Marvell 88E8056 used on the our X38 and P35 boards. What’s more, the Marvell chip also offers lower CPU utilization than the 790i SLI.

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

Throughput isn’t a problem for the Ultra here, but its CPU utilization is slightly higher than that of the other chipsets.

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

We’ve seen much better PCI throughput from our X38 and P35 motherboards in previous reviews, so I’m hesitant to make too much of these results. The 790i clearly doesn’t have a problem here, but something about the system configuration we used for this latest round of testing is affecting PCI Gigabit Ethernet throughput with Intel chipsets.

Pay no attention to the higher CPU utilization of the nForce chipsets. They’re pushing a lot more data, so significantly higher CPU utilization is to be expected.

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.

Power consumption has perhaps been the single greatest weakness of nForce chipsets, and with the 790i SLI Ultra, it appears little has changed. At idle, the Ultra sucks about 15 more watts than a similarly-equipped X48 board. That margin shrinks to ten watts under load, though. The 790i is more power-efficient than its predecessor, thanks in part to the fact that it doesn’t have to power an additional PCI Express bridge chip.

Don’t put too much stock into these P35 power consumption results. The Asus P5K3 Deluxe motherboard we used for testing has much higher power consumption than other P35 boards.

Overclocking
For our overclocking tests, we dropped our CPU multiplier to 6X—its lowest possible value. The memory bus was also maintained at 1333MHz 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.

We managed to get our test system up to a 490MHz front-side bus with the BIOS left to adjust voltages on its own. We didn’t have to use the chipset’s optional cooling fan, either. However, getting the system stable with a 500MHz front-side bus proved a little more problematic. Posting wasn’t an issue, and neither was idling at the Vista desktop, but after about five minutes of crunching our stress test, we’d get a blue screen or system crash. Adding the chipset cooler didn’t help, and neither did manually messing with chipset and other voltages.

490MHz isn’t the highest front-side bus overclock we’ve seen from a Core 2 motherboard, but it’s up there, and certainly high enough for all but the most extreme overclocking endeavors. Then again, your mileage may vary.

One area where it won’t vary, provided that you spring for the Ultra variant of the 790i SLI, is with memory speeds up to 2000MHz. With a set of EPP 2.0-certified Crucial Ballistix DDR3-2000 modules, we were able to push the 790i’s memory bus to an effective two gigahertz.

DDR3 is really delivering on its potential for higher clock speeds, and the nForce 790i SLI Ultra is primed for the fastest modules on the market.

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 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 790i-based motherboard is highlighted and in bold to make it easier to pick out from the crowd.

HD Tach
Firewire performance

Read burst
speed (MB/s)

Average read
speed (MB/s)

Average write
speed (MB/s)

CPU utilization
(%)

Asus P5E3 Deluxe
42.3 37.5 28.8 1.7

Asus P5K3 Deluxe
42.3 37.4 28.7 0.7

Gigabyte X48T-DQ6
42.1 37.5 28.7 2.3

XFX MB-N780-ISH9
42.2 37.5 21.8 3.0


XFX nForce 790i SLI Ultra
42.2 37.6 21.8 1.3

The XFX-branded nForce 790i SLI Ultra motherboard is about as high-end as you can get, but Nvidia’s reference design still features a Texas Instruments Firewire controller with relatively slow write performance.

HD Tach Serial
ATA performance

Read
burst
speed
(MB/s)

Average
read
speed
(MB/s)

Average
write speed
(MB/s)

CPU
utilization
(%)

Random
access
time
(ms)

Asus P5E3 Deluxe
133.5 75.2 101.0 2.0 8.4

Asus P5K3 Deluxe
133.7 75.2 98.7 3.7 8.3

Gigabyte X48T-DQ6 (ICH9R)
133.2 75.2 94.5 2.7 8.3

Gigabyte X48T-DQ6 (JMB363)
129.5 75.2 47.4 5.3 8.1

XFX MB-N780-ISH9
131.9 75.2 87.0 4.0 8.4

XFX nForce 790i SLI (nForce MCP)
132.2 75.2 87.5 3.3 8.3

XFX nForce 780i
SLI (JMB362)
109.8 75.3 42.7 20.3 8.7

The auxiliary JMicron storage controller Nvidia has hooked into the board’s eSATA port isn’t all that hot, either. Not only does it offer slower burst and write speeds than the 790i’s own SATA ports (and those on the other boards), it does so while consuming 20% of our test system’s CPU time.

RightMark Audio
Analyzer audio quality

Overall score

Frequency response

Noise level

Dynamic range

THD

THD + Noise

IMD + Noise

Stereo Crosstalk

IMD at 10kHz

Asus P5E3 Deluxe
4 5 3 3 3 1 3 4 3

Asus P5K3 Deluxe
4 5 3 1 3 1 3 4 3

Gigabyte X48T-DQ6
4 5 3 3 3 1 3 4 3

XFX MB-N780-ISH9
4 5 3 3 3 1 3 4 3

XFX nForce 790i SLI Ultra
4 5 3 3 3 1 3 4 3

Nothing to see here. Move along.

Conclusions
After delivering a less-than-inspiring nForce 780i SLI chipset last year, Nvidia has finally followed up with its true next-generation core logic in the 790 SLI Ultra. An all-new north bridge lies at the heart of this latest nForce chipset, complete with a DDR3 memory controller that not only boasts significantly lower latencies than its rivals, but also supports memory speeds all the way up to 2000MHz. 32 lanes of second-gen PCI Express are packed into the 790i SLI SPP, as well, complete with SLI-specific optimizations and the ability to run a third graphics card for three-way multi-GPU configurations.

Nvidia built the 790i SLI reference platform with enthusiasts in mind, endowing its BIOS with arguably the best mix of overclocking, tweaking, and fan speed control options on the market. The vast majority of these options are exposed in Windows through Nvidia’s excellent System Tools software, too. This all comes neatly packaged on a new motherboard design that sports a solid array of integrated peripherals, including support for multi-channel digital audio output. The layout of the new board is nearly perfect, and it can even accommodate some of the largest aftermarket processor heatsinks around.

However, it’s in this reference motherboard design that we see the first cracks in the 790i SLI’s appeal. A CPU-hungry auxiliary storage controller tasked primarily with eSATA connectivity is the first blemish, and it’s joined by a Firewire controller with relatively slow write speeds. These peripherals aren’t technically part of the chipset, of course, but we also observed lackluster throughput from the 790i SLI’s integrated Gigabit Ethernet controllers. The power consumption of Nvidia’s nForce reference platforms continues to be higher than that of Intel-based boards with similar peripheral payloads, as well.

These little flaws are hardly show-stoppers, but for a premium product like the nForce 790i SLI Ultra, they’re a little tough to swallow. $350 and up for a motherboard is a big ask, even for a high-end platform like the Ultra, making it tough to forgive even a small collection of imperfections.

Yet the nForce 790 SLI Ultra still manages to unseat Intel’s X48 Express from its throne as the reigning uber-high-end chipset thanks to two key features. The first and most obvious is support for SLI—a feature that Nvidia continues to deny its chipset competitors. With the new GeForce 9800 GX2 ready for effective four-way SLI configurations, the 790i SLI will be able to play host to what should be the highest performance graphics configuration on the market. That’s a big deal for an ultra-exotic platform like the, er, Ultra. The 790i SLI also offers full support for the Enthusiast System Architecture, which is the best solution for open hardware monitoring and control that we’ve seen to date. ESA is more than just a specification; it’s now backed up by real hardware and phenomenal Nvidia System Tools software.

Obviously, top-of-the-line chipsets like the Ultra are less concerned with value and more about winning bragging rights in a technical arms race. In the arms race between Core 2 chipsets, the nForce 790i SLI Ultra definitely leads.