Sandy Bridge-E motherboards from Asus, Gigabyte, Intel, and MSI

Intel is in a unique position in the market for ultra-high-end desktop PCs: it essentially has no competition. AMD hasn’t priced a desktop processor far north of $300 in years, while Intel has been selling six-core Gulftown CPUs at up to a grand a pop. With no one nipping at its heels, Intel has had the luxury of updating its flagship platform at what could charitably be described as a conservative pace.

When Gulftown made its debut with the Core i7-980X Extreme in early 2010, it breathed new life into an X58 Express platform that had been around since 2008. Only 133MHz has been added to Gulftown’s top end since, and the X58 hasn’t been touched, save for motherboard makers rolling out new models that augment the chipset with next-gen peripherals. So, yeah, Sandy Bridge-E and its accompanying X79 Express chipset have been a long time coming.

The two are inseparable thanks to a new LGA2011 socket with a rather timely pin count. If you want to get in on a six-core Sandy Bridge CPU, you’re gonna need a new motherboard. But which one? That’s a daunting question given the complexity of modern mobo designs, which house ever-growing peripheral payloads in addition to rapidly evolving firmware. To help make sense of your options, we’ve rounded up four LGA2011 boards from the biggest players in the business. Join us as the Asus P8X79 PRO, Gigabyte X79-UD5, Intel DX79SI, and MSI X79A-GD65 (8D) square off to see which is the best foundation for Sandy Bridge-E.

E is for Enterprise Extreme

If you haven’t already read Scott’s in-depth look at the Core i7-3960X, I recommend it as a starting point. This round-up will focus primarily on how the different motherboards compare rather than looking too deeply into the intricacies of the processor. We should, however, pause for a moment to drink in the glorious excess that is Sandy Bridge-E. Intel’s new flagship has a six-pack of Sandy Bridge cores backed by up to 15MB of last-level cache—50% more cores and nearly double the cache of the Core i7-2700K.

Sandy Bridge-E doesn’t have more cores than Gulftown (if you don’t count the additional two cores that Intel has chosen to disable for desktop-bound versions of the silicon), but it does increase the number of memory channels from three to four. Intel officially supports 1600MHz memory with one DIMM per channel, an increase from 1333MHz in both Gulftown and the Sandy Bridge CPUs that plug into LGA1155 sockets. Motherboard makers say their boards can handle higher memory speeds when paired with suitably fancy DIMMs, of course.

While Intel’s last halo platform consolidated much of its PCI Express connectivity in a north-bridge chip, Sandy Bridge-E has PCIe lanes built into the CPU. This is the first implementation we’ve seen of PCI Express 3.0, although Intel is very precise about how it talks about such things. The platform’s official block diagram lists PCI Express 2.0 lanes attached to the CPU, but buried in the technical documentation is an assertion that the PCIe connectivity is “fully-compliant to the PCI Express* Base Specification, Revision 3.0.” Third-gen PCI Express devices simply don’t exist in the wild, and Intel won’t be confident about compatibility until it has hardware to verify. Presumably, hardware that is likewise compliant with the spec will work at PCIe 3.0 speeds.

Source: Intel

Extra speed isn’t the only perk to gen-three PCI Express, but it’s one of the biggest benefits. Each lane has an 8GT/s data rate, up from 5GT/s with PCIe 2.0. The data is processed differently with the new spec, which scrambles the bits using binary math before feeding them through a 128b/130b encoding scheme whose overhead is an order of magnitude smaller than with the 8b/10b encoding from the old standard. As a result, PCIe 3.0 squeezes 1GB/s of bidirectional bandwidth out of its 8GT/s, doubling the effective throughput of gen-two PCIe with only a 60% increase in the data rate.

There are 40 PCIe 3.0 lanes in CPU split between three links: two with 16 lanes each and one with eight. One of the x16s can be split into a pair of x8s for four-way graphics goodness in an x16/x8/x8/x8 config. If you don’t need quite as much bandwidth per lane, one of the x8s can be cleaved into a pair of x4s. It would take a very fast PCIe SSD to saturate a fraction of the bandwidth available in just one of those quad-lane links.

Even more PCI Express lanes are housed within the Sandy Bridge-E die, but they hail from the second generation and have been repurposed as a chipset interconnect. A four-lane DMI link connects the CPU to its X79 Express sidekick, providing 2GB/s of bandwidth in each direction.

The X79 was supposed to have a second interconnect dedicated to new storage-controller logic bristling with 6Gbps SATA and Serial Attached SCSI ports. Dubbed SCU, this controller would have been a nice bit of baggage from Sandy Bridge-E’s real mission in enterprise environments. Unfortunately, the SCU and its associated interconnect have been “de-featured” on the chip “as a result of silicon health.” Intel wanted to meet its November 2011 launch schedule for the platform (cosmic alignment, perhaps?), and it apparently couldn’t get the SCU working in time.

We started hearing rumors about issues with the X79’s new storage hotness late this summer, and that might explain why the platform didn’t launch in the August/September timeframe expected by all the motherboard makers we spoke with at Computex in June. Intel suffered an embarrassing SATA bug that forced the recall of its initial Sandy Bridge chipsets earlier this year; perhaps it was contagious.

The X79 Express platform hub

With the SCU and its accompanying interconnect fused shut, the X79 looks a lot like Intel’s existing P67 platform hub. Note that I didn’t say Z68, which offers QuickSync video transcoding and Smart Response SSD caching. Sandy Bridge-E lacks an integrated GPU, so it can’t play the QuickSync game, and Intel hasn’t finished porting Smart Response to the platform. We’re told SSD caching will be integrated into a driver update due midway through next year, which seems like an awfully long time to wait given the fact that the X79 has similar storage controller logic to the Z68.

Like the bulk of Intel’s 6-series lineup, the X79 has dual SATA controllers: one with two 6Gbps ports, and another with four 3Gbps ones. RAID is on the menu, of course, but the array of SATA ports isn’t particularly aspirational considering the fact that AMD serves up six 6Gbps ports in its chipsets. Then again, high-performance SSDs are migrating toward PCI Express, which Sandy Bridge-E has in spades.

Speaking of PCIe, there are eight second-generation lanes in the X79 reserved for expansion slots and peripherals—like extra 6Gbps SATA controllers. As with the SATA ports, the logic is similar to what’s in Intel’s other desktop chipsets. The same goes for the Gigabit Ethernet and USB 2.0 controllers. USB 3.0 would’ve been a nice inclusion, but given Intel’s enterprise focus for Sandy Bridge-E, I can understand why more SATA and SAS ports took priority. If only they had worked.

Cutting-edge 32-nm process technology is used to create Sandy Bridge-E silicon, while Intel uses its 65-nm fab capacity to crank out the X79. Interestingly, the chip’s 7.8W TDP is higher than the 6.1W rating for the P67, which has essentially the same feature set. I suspect the extra transistors associated with supporting the SCU are responsible for that discrepancy.

The X79 Express four ways

Every time a new platform launches, I tell myself to be less ambitious with the initial motherboard round-up. And every time, I fail. We’ve rounded up no fewer than four motherboards to see what Asus, Gigabyte, Intel, and MSI have concocted for Sandy Bridge-E.

Asus P9X79 PRO Intel DX79SI Gigabyte X79-UD5 MSI X79A-GD65 (8D)
DIMM slots 8 DDR3 8 DDR3 8 DDR3 8 DDR3
Expansion slots 2 PCIe 3.0 x16

2 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x1

2 PCIe 3.0 x16

1 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x1

1 PCI

2 PCIe 3.0 x16

1 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x1

1 PCI

2 PCIe 3.0 x16

1 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x16 (x1)

1 PCIe 2.0 x1

Gigabit Ethernet Intel 82579V Intel 82574L

Intel 82579LM

Intel 82579 Intel 82579
USB 3.0 2 x ASMedia ASM1042

VIA VL810

2 x NEC D720200 2 x Fresco FL1009 2 x NEC D720200
Auxiliary SATA ASMedia ASM1061

Marvell 88SE9128

NA 3 x Marvell 88SE9172 ASMedia ASM1061
Audio Realtek ALC898 Realtek ALC892 Realtek ALC898 Realtek ALC892
FireWire NA VIA VT6315N VIA VT6308 VIA VT6315N
Warranty Three years Three years Three years Three years
Price $329 $280-300 $350 $280-299

All of the boards have eight DIMM slots and at least three PCI Express x16 slots. Their expansion slot configurations are otherwise quite different, although there is agreement on the Ethernet front. Motherboard makers are finally using the Gigabit Ethernet MAC included in the chipset rather than farming out networking to discrete Realtek controllers.

Since there’s no USB 3.0 connectivity in the chipset, mobo makers must employ auxiliary controllers. All but the Intel board is equipped with additional SATA connectivity, as well. I guess slapping a couple of additional 6Gbps ports onto the DX79SI would’ve been an admission that the two in the chipset aren’t enough.

The Asus board is the only one among the four to cut the cord on FireWire. That’s somewhat disappointing in light of the fact that the P9X79 PRO is among the most expensive of the lot. At least the board features Realtek’s latest ALC898 audio codec, which also appears on Gigabyte’s X79-UD5. Distilling these boards down to their base specifications doesn’t do them justice, though. Let’s take a closer look at what each one brings to the table.

Asus’ P9X79 PRO motherboard

Asus has been on a bit of a tear lately. The company’s first Sandy Bridge motherboards were easily the most attractive examples of the breed thanks to intelligent features and the best firmware in the business. With a $329 expected price tag, the P9X79 PRO will have to deliver on both fronts; this is the second most expensive board of the bunch.

At first glance, the PRO certainly looks the part of a high-end motherboard. There’s a lot going on, including no fewer than four separate heatsinks to cover the chipset and power regulation circuitry. Heatpipes link the coolers in pairs, and the mass of finned, anodized metal looks rather attractive. The mix of black, white, and different shades of blue is at least a small deviation from the black-and-blue uniform worn by most enthusiast-oriented motherboards lately.

Asus has used digital power delivery circuitry for the CPU through multiple generations of motherboards. With its X79 family, digital VRMs have also been applied to the memory. The digital circuitry is purportedly more accurate and responsive than analog alternatives, which Asus claims improves overall stability. A total of eight power phases is dedicated to the CPU core, while another two feed the rest of the chip. An additional four power phases are split between the system’s two banks of DIMM slots.

The memory slots are only 20 mm from the edge of the CPU socket, so you’ll want to be careful about combining big aftermarket heatsinks with tall DIMMs. Two of the VRM heatsinks also crowd the socket, although they’re not much more than an inch tall. After swapping a Core i7-3960X back and forth between motherboards for a week, I’ve gotta say I dig the new socket’s screw-in retention bracket.

Asus has gone all-PCIe with the PRO’s expansion slot layout, which includes a pair of x1 slots in addition to a quartet of x16s. The blue x16 slots have 16 lanes of PCIe 3.0 bandwidth each, and the bottom one shares lanes with the top white x16 slot in an x16/x8/x8 config. The second white slot has eight lanes of dedicated PCI Express 3.0 connectivity. As you can see, the slots are arranged such that dual-double-wide graphics configurations won’t block access to the white x16s, which enjoy a direct path to the CPU.

To the right of the expansion slots sits a cluster of edge-facing SATA ports. The white ones offer 6Gbps connectivity, while the blue ones are capped at 3Gbps. The 6Gbps ports with the little red sticker are tied to a Marvell controller and are a part of Asus’ new SSD caching scheme. Marvell has a HyperDuo caching system of its own, but Asus has whipped up custom software and algorithms for its own implementation. Asus’ approach works like Intel’s Smart Response Technology, allowing an SSD to cache both reads and writes associated with a mechanical hard drive attached to the same controller.

The port cluster provides plenty of goodies, including a Bluetooth transceiver and a BIOS Flashback button. The Flashback functionality allows the board’s firmware to be updated using only a thumb drive and a power supply—no CPU or memory required. This handy little feature should ensure older boards can easily be updated to support new CPUs.

With the exception of the missing FireWire port, the rest of the cluster is pretty standard fare. Kudos to Asus for equipping the eSATA ports, which are tied to an ASMedia 6Gbps controller, with USB power. Two more ASMedia controllers handle the USB 3.0 ports with a little help from a four-port Via hub chip. One of the controllers is tasked with supplying the front-panel ports, while the second combines with the Via chip to feed the four ports in the rear cluster. The hub chip has only one upstream port, so it’s possible that one of the rear ports is connected directly to the ASMedia controller, and thus not sharing bandwidth with the others. We’re still waiting for Asus to confirm whether that’s the case.

The PRO’s SuperSpeed ports are enhanced by a feature called USB Boost, which uses a USB Attached SCSI Protocol (USAP) that’s supposed to deliver faster transfer rates when paired with USAP-compatible devices. USAP-compatible devices appear to be rather rare, but there’s also a Turbo mode designed to work with USB 3.0 gear that relies on the old Bulk-Only Transport protocol that’s been around since version 1.1 of the standard. In the short amount of time we’ve been able to devote to USB Boost, we’ve encountered compatibility problems with at least one solid-state drive—in both an older USB 3.0 docking station and a USAP-aware enclosure that Asus sent over. Asus is looking into the matter, and we may have to spend more time with USB Boost once the kinks are ironed out.

Although Realtek networking chips have been banished from the X79 boards we’ve collected today, the company’s new ALC898 codec appears on the P9X79 PRO. Asus augments the chip with DTS software that offers surround-sound virtualization. Another DTS component encodes surround-sound bitstreams in real time, enabling multi-channel game audio to be piped to compatible receivers and speakers over a digital output that bypasses the onboard DAC. This is as robust an integrated audio solution as we’ve seen.

Asus’ UEFI is the standard by which all others are judged, not only for its visual flair, but also for the responsiveness of its interface and the wealth of options that lie within. The P9X79 PRO doesn’t bring big changes, but Asus has made a few tweaks that are worth noting. One recent addition is the ability to take screenshots, which are saved to an attached thumb drive and make the GUI look much nicer when shown off in reviews like this one. Asus has also added a new shortcut menu that’s invoked with the F3 key and provides quick access to commonly used options like fan controls, DRAM timings, and CPU performance settings.

The interface is as snappy as ever, and the overclocking and tweaking options are generally excellent. However, the firmware takes liberties with Turbo multipliers that it really shouldn’t. If you set the memory clock manually, Asus replaces CPU’s default Turbo behavior, which is to use a mix of 39, 37, and 36X multipliers depending on the number of active cores, with a blanket 39X multiplier for all loads. There’s no message to the user indicating that this change has taken place, and none of the other X79 boards we’ve tested mess with CPU multipliers in the same way. Simply put, this behavior is unacceptable. Changing one system setting shouldn’t secretly alter another, especially when they’re unrelated. Thankfully, it’s possible to define the correct Turbo behavior manually through the firmware.

Asus doesn’t mess around with the fan options, which offer temperature-based speed control for all the onboard headers. Users can set temperature limits and duty cycles in 1°C and 1% increments, respectively, and there’s even more fan control goodness available in Windows.

Rather than restricting users to minimum and maximum values, Asus’ FanXpert+ software provides control over a third point on each fan profile. The drag-and-drop interface makes tweaking fan profiles a joy, and as in the firmware, it’s possible to define different behavior for each onboard fan header. These fan controls are designed for four-pin PWM fans, but they’ll also work with three-pin DC models plugged into the CPU and first chassis headers.

FanXpert+ is neatly integrated into the AI Suite Windows utility, which also has built-in overclocking controls. Using AI Suite’s TurboV EVO application, it’s possible to change clock speeds, multipliers, and voltages. A separate AI Suite component provides control over the board’s power delivery circuitry for the CPU and memory. Stay tuned for a closer look at that auto-tuning button in the overclocking section of the round-up.

Intel’s DX79SI motherboard

Although much better known for its CPUs, Intel also makes motherboards—enthusiast-oriented models, too. The DX79SI is one such board, as its rather ominous skull logo attests.

When we’ve included Intel in previous motherboard roundups, the company’s boards have come out looking a little light on features given their premium price points. This time around, however, the DX79SI is one of the cheapest of the bunch, with an expected street price in the $280-300 range. Intel has also endowed the board with a few perks you might not expect.

One of those perks is fancier electrical components, which Intel hasn’t talked much about in the past. As is customary for high-end motherboards these days, the DX79SI is outfitted with solid-state capacitors, driver MOSFETs, digital voltage regulation for the CPU, and other exotic components aimed at improving power delivery and overall system stability.

Intel can lay claim to being the first motherboard maker to adopt a black-and-blue color scheme, so I can’t fault the company for sticking to its guns as everyone else has embraced the same palette. The DX79SI looks like it means business, in part thanks to the chunky heatsinks that sit atop the power regulation circuitry.

Alas, the lower heatsink is too close to the top PCI Express x16 slot, obscuring access to the retention clip. With a double-wide graphics card installed, my stubby fingers had a difficult time releasing the clip. This wasn’t a problem with a similarly placed heatsink on the Asus board, which leaves more room around a different retention clip design.

The VRM heatsinks cool eight power phases dedicated to the CPU: seven for the CPU cores and one for the rest of the chip. The coolers are tightly packed around the socket, leaving as little as 8.5 mm of airspace between their metal fins and the edge of the LGA2011 bracket.

Intel apparently prefers a tight fit, because the heatsinks and DIMM slots are closer to the socket on the DX79SI than on any of the other X79 boards we’ve measured. The heatsinks are relatively short, at 32.5 mm, but folks with aftermarket air-cooling towers will want to be careful about using DIMMs with tall heatspreaders.

Moving south brings us face to face with the skull-infused chipset cooler, whose low-profile design won’t interfere with longer expansion cards. There are three PCI Express x16 slots linked to the third-gen controller in the CPU. The first two slots have 16 lanes of bandwidth each, while the third has only eight lanes of electrical connectivity. Those eight lanes actually come from the second x16 slot, limiting the board to an x16/x8/x8 config when all three x16 slots are in use. Intel confirmed this configuration, and I’m still a little confused by it. After all, there are enough PCIe 3.0 lanes in the CPU to power a pair of full-bandwidth x16 slots in addition to a separate x8.

The DX79SI is one of only two boards in the bunch to sport a PCI slot, and perhaps it’s time to let the legacy standard go. As much as we love Asus’ PCI-only Xonar DG sound card, I’m finding it difficult to think of another PCI peripheral I’d recommend for a modern PC. You will want to plug a sound card of some kind into the DX79SI, though. The board’s integrated audio has neither surround-sound virtualization nor real-time multichannel encoding.

On the right edge, one can clearly see that the DX79SI was designed to serve up additional SATA ports. With the X79’s SCU disabled, Intel could have turned to third-party controllers to provide additional SATA ports—but it didn’t. Instead, the board has to make do with six SATA ports in total, only two of which offer 6Gbps connectivity.

Intel doesn’t even offer any external Serial ATA ports, which are admittedly less desirable now that USB 3.0 devices are widely available. The DX79SI has four SuperSpeed ports in total, two in the rear cluster and another two tied to a front-panel connector. Each pair of ports is fed by a dedicated NEC controller.

While the DX79SI scores points for offering a second Gigabit Ethernet connection based on a separate Intel controller chip, the rear cluster looks pretty barren otherwise. I wouldn’t mind seeing a few more USB 2.0 ports make their way from internal headers into the rear cluster. The white Back to BIOS button over to the left is a nice addition, though. Press it, and the board will boot with its firmware defaults without forgetting the last settings defined by the user.

Although our press kit didn’t come with one, the DX79SI boards sold to consumers will include an external wireless module that plugs into one of the internal USB headers. This module offers Bluetooth 2.1 and 802.11n Wi-Fi connectivity, making it a useful addition to the overall package. Intel also throws in a remote temperature probe that can be used in conjunction with the firmware’s excellent fan speed controls.

While the firmware lacks mouse support and a graphical interface, its fan controls are quite powerful. Rather than setting points on a curve, users are asked to define a target temperature for the fan to maintain. There are additional damping and responsiveness options to dictate how aggressively the fan speed changes with the temperature, plus a threshold that kicks fans up to their maximum RPM. In a separate screen, users have control over the minimum and maximum fan speeds. All these settings can be applied independently to all of the onboard fan headers, and Intel says the speed controls work with both three- and four-pin fans. The only thing I’d change is the interface, which awkwardly splits the temperature targets and speed controls between different menus.

The whole firmware could use a next-generation GUI, but at least the old-school approach is snappy and easy to navigate. There are nice little touches, too, like options changing color when they’ve been modified during the current session. However, there’s also some inconsistency in how changes can be made; some variables can be keyed in directly, while others can only be tuned by tapping the + and – keys. At least there are plenty of overclocking options on offer.

If Intel decides to drag its firmware interface out of the BIOS era, it should model the new GUI after its Extreme Tuning Utility for Windows. The app’s interface is slick, easily customized, and loaded with overclocking and tweaking options. Unfortunately, fan controls aren’t a part of the mix, which is a real shame considering the utility’s integrated monitoring panel.

Gigabyte’s X79-UD5 motherboard

With a $350 price tag, the X79-UD5 is the priciest option we’re looking at today. Of course, it also has the highest number of auxiliary peripheral controllers and associated ports. This is the big daddy of the bunch—literally. While the other boards measure up to the ATX standard of 12″ x 9.6″, the UD5 is a full 10.4″ wide. The extra inch shouldn’t complicate compatibility with too many enclosures, but it might be a tight fit for smaller mid-tower cases.

Gigabyte combines the UD5’s black circuit board with a smattering of monochrome slots and ports. The only splashes of color come on the heatsinks, whose designs have an almost pixellated vibe. I like the overall look, especially since the blue accents aren’t applied up and down the line. Green and orange highlights can be found on other members of Gigabyte’s X79 family.

3D Power is the buzzword of the day for the X79-UD5, and the 3D moniker isn’t as ridiculous as one might expect. The board has three clusters of digital power delivery circuitry: one for the CPU and one for each of the banks of four DIMMs flanking it. Like Asus, Gigabyte touts the superior power-delivery characteristics of these digital circuits over their analog counterparts.

Gigabyte says the UD5 uses a 12-phase power solution, although it wasn’t more specific about how those phases are distributed. Interestingly, this is the only member of the company’s X79 lineup with eight DIMM slots. The cheaper UD3 is a four-DIMM board, as are the pricier UD7 and Assassin models.

With extra board real estate at its disposal, the UD5 has the widest clearances we’ve seen surrounding an LGA2011 socket. The DIMM slots are a full 22 mm away, and the lone VRM heatsink steers well clear. To be fair, the socket is closer to the top PCI Express x16 slot than it is on any of the other boards.

That primary x16 slot has 16 lanes of PCIe 3.0 linked directly to the CPU. Another 16 lanes are piped to the bottom slot (top in the picture above), while the middle x16 slot makes do with eight lanes of electrical connectivity. From this angle, you can clearly see the missing traces in the middle x16 slot.

The chipset heatsink looks even better from this vantage point, and the board seems to have more unused surface area than its rivals. You wouldn’t guess that there are no fewer than three Marvell 6Gbps SATA controllers onboard. Two are tasked with supplying the gray internal ports, while the remaining controller is linked to eSATA ports at the rear.

Only one of those eSATA ports has USB power, but there are dual USB 3.0 ports plus an internal header for two more. Unlike the other motherboard makers, Gigabyte provides a 3.5″ drive bay insert for the front-panel USB 3.0 ports. The SuperSpeed ports are powered by a pair of Fresco Logic controllers, the first we’ve seen of that company’s offerings on enthusiast-oriented motherboards.

Let’s go back to the cluster, because there’s a lot going on. Over to the left lies a trio of buttons charged with clearing the CMOS, switching between the primary and backup firmware, and handling automatic overclocking duties. The integrated audio is also worth mentioning because Gigabyte has paired Realtek’s latest high-end audio codec with Dolby software that provides not only surround-sound virtualization, but also real-time Dolby Digital Live! encoding. At least on paper, the integrated audio ticks the same boxes as the Asus board, although Dolby Digital Live! isn’t quite as robust as DTS Connect.

In addition to all its onboard peripherals, the UD5 comes bundled with a wireless expansion card that plugs into a PCIe x1 slot and also consumes two internal USB ports. The card offers Bluetooth 4.0 and 802.11n Wi-Fi connectivity, and it’s joined in the box by a pair of Wi-Fi antennas.

At long last, Gigabyte has joined the ranks of motherboard makers with next-generation firmware interfaces. Dubbed 3D BIOS, this firmware takes full advantage of the graphical goodness supported by the UEFI standard. Users can choose between the 3D interface pictured above and an old-style menu system, so there’s a little something for everyone.

The 3D interface is particularly slick, allowing users to bring up tweaking options by clicking on relevant sections of the board. The tuning panels sport mouse-friendly sliders and come with a monitoring overlay, but there’s too much overlap with the latter.

Although we didn’t encounter any mouse flickering, the tracking feels sluggish. Our mouse wheel wouldn’t work, either, and the transition between the 3D and menu-style interfaces is painfully chunky compared to the much smoother transitions elsewhere. There’s also some inconsistency in how individual options respond to mouse clicks, making navigation more difficult than it needs to be.

Overall, the firmware has plenty of tweaking options. Gigabyte has even expanded its fan speed controls to cover two system fans in addition to the CPU fan. However, the only option available for each fan is a PWM/°C value that’s far from intuitive. Gigabyte tells us the speed control for the CPU fan header is compatible with both three- and four-pin fans. The speed control for the first system fan is designed for three-pin fans, while the logic behind the second system fan is geared toward four-pin fans.

The fan speed controls available through Gigabyte’s EasyTune6 Windows software are much easier to decipher, although they’re not as granular as what’s offered by Asus’ FanXpert+ app. Only two points can be set on each fan curve.

EasyTune6 also includes basic tweaking and overclocking controls, and it’s one of a handful of Gigabyte apps that offer similar functionality. A separate 3D Power app provides control over the board’s power circuitry with an awfully sluggish interface. Then there’s Touch BIOS, which provides a finger-friendly means to manipulate firmware settings from the comfort of Windows.

While the tweaking options provided by these various applications are valuable, there’s no consistency in their interfaces, which are generally cumbersome. Instead of trying to cater to a community of touchscreen-using overclockers that doesn’t exist, Gigabyte should focus its software efforts on coming up with a single, coherent tuning utility that observes sensible interface conventions. EasyTune was a good start, but instead of building on that, Gigabyte seems more interested in creating separate applications from scratch.

MSI’s X79A-GD65 (8D) motherboard

With a $280-299 expected price tag, the X79A-GD65 (8D) is cheaper than the Asus and Gigabyte boards and about the same price as the Intel. Note the 8D, which distinguishes this eight-DIMM model from the vanilla GD65. The standard board has only four DIMM slots, as do all the other entries in MSI’s current X79 lineup. A eight-DIMM Big Bang motherboard is in the works, but it’s not ready for prime time just yet.

While the Big Bang board will be almost completely blacked out, the GD65 blends in with the rest of the black-and-blue crowd. The color scheme certainly isn’t what I’d expect from the Military Class branding, although I suppose the counter-terrorists in Counter-Strike wear matching hues.

Here, Military Class III refers not to AWP-fueled headshots, but to MSI’s use of electrical components that meet the MIL-STD-810G standard for temperature, vibration, and shock tolerance, among other tests. MSI has been a vocal proponent of higher-grade electrical components, and it says the GD65 is loaded with ’em. The board is peppered with driver MOSFETs, solid-state capacitors, tantalum-core caps around the CPU, and super-ferrite-core chokes.

On a couple of its gen-three Z68 boards, MSI’s claims of superior component quality are backed by five-year warranty coverage. Like all the other X79 models in this round-up, though, the GD65 has a three-year warranty.

MSI uses a 12-phase solution to keep Sandy Bridge-E CPUs swimming in power. Ten phases are dedicated to the CPU cores, while two are reserved for the chip’s uncore component, otherwise known as the System Agent.

There’s a reasonable amount of breathing room around the GD65’s socket, in part thanks to a VRM heatsink that doesn’t reach its full height until 22 mm away. However, the DIMM slots are less than 20 mm from the edges of the retention bracket, making low-profile DIMMs look particularly attractive.

Before you get too excited about the GD65’s five PCI Express x16 slots, I should note that the second and fourth slots (the ones with the one-armed retention clips) have only one lane of PCIe 2.0 connectivity each. The larger slots can accommodate full-sized cards; they’ll just be starved for bandwidth. The other x16 slots split the processor’s 40 PCIe 3.0 lanes in the CPU in an x16/x16/x8 configuration.

Edge-mounted SATA ports are all the rage these days, and MSI complements the six fed by the X79 chipset with a pair of 6Gbps ports powered by an ASMedia controller. The 6Gbps ports are identified in white, with the ones on the left linked to the ASMedia chip.

That ASMedia controller is the only auxiliary SATA chip on the board, which means no eSATA connectivity for the port cluster. MSI does, however, include a PCI back plate with two eSATA ports and a four-pin Molex connector that can be linked to the motherboard. Also included is a similar back plate with two USB 3.0 ports for the front-panel connector.

There’s still plenty of action in the rear cluster, which includes a FireWire port and a convenient CMOS reset button. MSI loads up on USB 2.0 ports, offering more here than any of the other boards.

On the audio front, the GD65 sticks with Realtek’s ALC892 codec chip, which packs a dose of surround-sound virtualization. The board is also supposed to come with THX TrueStudio PRO software that provides another option for surround virtualization, among other audio effects. That software was missing from our sample, though.

MSI’s first attempt at a next-generation firmware interface was so bad that it’s been completely gutted and replaced with ClickBIOS II. The new interface sticks with a single menu system, and it’s much improved over the old design. There’s no mouse flickering and tracking feels smooth, although I’d prefer to be able to select items with a single click instead of double-tapping. The mouse wheel scrolls nicely, at least.

The firmware feels like a blend of the EZ and Advanced modes of Asus’ UEFI. There’s a drag-and-drop boot order up top, along with a collection of basic system details. Options are found in the middle column, whose contents are changed by clicking the entries in the columns on either side. Overall, the interface is intuitive and quick to navigate. A handy screenshot function makes capturing the view easy, as well.

While the firmware has ample overclocking and tweaking options, MSI’s fan controls have regressed. It’s still possible to set a minimum fan speed and a target temperature for the CPU. However, there’s no way to control the system fans. MSI used to offer manual speed control for system fans, but on the X79A-GD65, those are only available in Windows via the Control Center application.

The app isn’t the slickest tweaking utility around, but its interface has at least been designed for mouse-based input. If you don’t like the look of Control Center, a separate utility brings the ClickBIOS II interface to Windows.

The overall look translates nicely, and I’m curious to see whether MSI will keep updating Control Center or pursue this avenue instead. There’s certainly some benefit to maintaining a consistent tweaking interface between a motherboard’s firmware and Windows utilities.

Digging into the details

Congratulations for making it this far already. As a reward for not skipping ahead to the conclusion after the first page, I’m going to let you in on a little secret: this page isn’t all that interesting. Unless you like seeing firmware options, board specifications, and system configurations laid out in painstaking detail, you might want to skip ahead to the beginning of our performance results.

Asus P9X79 PRO Intel DX79SI Gigabyte X79-UD5 MSI X79A-GD65 (8D)
Clock speeds Base: 80-300MHz

DRAM: 800-2666MHz

Base: 89.93-146.98MHz

DRAM: 800-2400MHz

Base: 80-133MHz

DRAM: 800-2400MHz

Base: 90-200MHz

DRAM: 800-2400MHz

Multipliers CPU: 34-57X

Base: 1, 1.25, 1.67, 2.5X

CPU: 5-65X

Base: 1, 1.25, 1.67X

CPU: 12-59X

Base: 1, 1.25, 1.67, 2.5X

CPU: 12-60X

Base: 1, 1.25, 1.67X

Voltages CPU: 0.8-1.7V

VCCSA: 0.8-1.7V

DRAM: 1.2-1.99V

CPU PLL: 1.8-2.1V

CPU VTT: 1.05-1.7V

PCH 1.1: 1.1-1.7V

PCH 1.5: 1.5-1.8V

VTTDDR: 0.625-1.1V

CTRL: 0.392-0.63X

DATA: 0.392-0.63X

CPU: 1.0-1.92V

DRAM: 1.2-1.925V

System agent: 0.85-1.8V

CPU IO: 1.05-1.8V

CPU PLL: 1.8-2.0V

PCH: 1.1-1.5V

CPU: 0.8-1.735V

CP VTT: 0.87-2.08V

CPU PLL: 1.195-1.985V

IMC: 0.65-1.82V

DRAM: 1.1-2.1V

DRAM term: 0.574-1.953V

DRAM data: 0.616-1.501V

DRAM address: 0.616-1.501V

PCH: 0.825-1.51V

PCH 1.5: 0.925-2.175V

CPU: 0.8-1.8V

System agent: 0.85-1.8V

DRAM: 1.05-2.445V

CPU IO: 0.85-1.69V

CPU PLL: 1.4-2.5V

PCH 1.1: 0.9-1.9V

PCH 1.5: 1.2-1.9V

CTRL: 0.435-1.15X

DATA: 0.435-1.15X

Fan control CPU:

Min, max duty cycle

Min, max temp

Chassis 1-4:

Min, max duty cycle

Max temp

All 4 headers:

Min, max duty cycle

Target, max temp

Damping, responsiveness

All 3 headers:

Duty cycle/temp

CPU:

Target temp

Min duty cycle

Only extreme types with custom liquid cooling or vats of liquid nitrogen are likely to bump up against limits in the overclocking options available on these boards. For most users, the most meaningful differences between the firmware implementations are the interfaces and the level of built-in fan speed control.

Asus P9X79 PRO Intel DX79SI Gigabyte X79-UD5 MSI X79A-GD65 (8D)
DIMM slots 8 DDR3 8 DDR3 8 DDR3 8 DDR3
Expansion slots 2 PCIe 3.0 x16

2 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x1

2 PCIe 3.0 x16

1 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x1

1 PCI

2 PCIe 3.0 x16

1 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x1

1 PCI

2 PCIe 3.0 x16

1 PCIe 3.0 x16 (x8)

2 PCIe 2.0 x16 (x1)

1 PCIe 2.0 x1

Storage I/O 2 SATA RAID 6Gbps

4 SATA RAID 3Gbps

2 SATA 6Gbps

2 SATA RAID 6Gbps

4 SATA RAID 3Gbps

2 SATA RAID 6Gbps

4 SATA RAID 3Gbps

2 SATA 6Gbps

2 SATA 6Gbps

2 SATA RAID 6Gbps

4 SATA RAID 3Gbps

Audio 8-channel HD 8-channel HD 8-channel HD 8-channel HD
Wireless Bluetooth 2.1 802.11n Wi-Fi

Bluetooth 2.1

802.11n Wi-Fi

Bluetooth 4.0

NA
Ports 4 USB 3.0 w/ 2 headers

6 USB 2.0 w/ 6 headers

2 eSATA/USB 6Gbps

1 Gigabit Ethernet

1 analog front out

1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in

1 optical S/PDIF out

2 USB 3.0 w/ 2 headers

6 USB 2.0 w/ 8 headers

1 FireWire w/ 1 header

2 Gigabit Ethernet

1 analog front out

1 analog bass/center out

1 analog rear out

1 analog line in/surround out

1 analog mic in

1 optical S/PDIF out

1 PS/2 keyboard/mouse

2 USB 3.0 w/ 2 headers

7 USB 2.0 w/ 4 headers

1 eSATA 6Gbps

1 eSATA/USB 6Gbps

1 FireWire w/ 1 header

1 Gigabit Ethernet

1 analog front out

1 analog bass/center out

1 analog rear out

1 analog line in/surround out

1 analog mic in

1 optical S/PDIF out

1 PS/2 keyboard/mouse

2 USB 3.0 w/ 2 headers

8 USB 2.0 w/ 4 headers

1 FireWire w/ 1 header

1 Gigabit Ethernet

1 analog front out

1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in

1 optical S/PDIF out

1 coaxial S/PDIF out

If you’ve been paying careful attention over the preceding pages, you already know all the details in the table above.

Our testing methods

Sandy Bridge-E has no peers, so we’ll primarily focus on pitting the various X79 motherboards against each other. For comparative reference, we’ve included a couple of more affordable desktop platforms. AMD is represented by Asus’ Sabertooth 990FX motherboard, which we’ve loaded with a six-core Phenom 1090T. We’ve also put together a Z68 system based on a Gigabyte Z68X-UD7 motherboard and a Core i7-2600K. Neither of those two platforms is a direct rival for Sandy Bridge-E, so their results have been largely greyed out over the following pages.

The only exceptions to that color-coding rule are our chipset performance tests, which pit the X79 platform hub against the Z68 and 990FX chipsets. These chipset performance tests don’t lean on the CPU too heavily, so the different processors shouldn’t affect the results. Neither should the fact that the Sandy Bridge-E systems have twice the memory of the others. I don’t believe any of the workloads in our test suite will make use of more than 8GB of system memory.

We used the following system setups for testing. With few exceptions, all tests were run at least three times, and we reported the median of the scores produced.

Processor AMD Phenom II X6 1090T 3.2GHz Intel Core i7-2600K 3.4GHz Intel Core i7 3960X 3.3GHz
Motherboard Asus Sabertooth 990FX Gigabyte Z68X-UD7 Asus P9X79 PRO Intel DX79SI Gigabyte X79-UD5 MSI X79A-GD65 (8D)
Bios revision 0705 F9 0709 SI0280B F4 112
Platform hub AMD 990FX/SB850 Intel Z68 Express Intel X79 Express Intel X79 Express Intel X79 Express Intel X79 Express
Chipset drivers Catalyst 11.9 Chipset: 9.2.0.1030

RST: 10.6.0.1022

Chipset: 9.2.3.1030

RST: 3.0.0.9012

Chipset: 9.2.3.1030

RST: 3.0.0.9012

Chipset: 9.2.3.1030

RST: 3.0.0.9012

Chipset: 9.2.3.1030

RST: 3.0.0.9012

Memory size 8GB (2 DIMMs) 8GB (2 DIMMs) 16GB (4 DIMMs) 16GB (4 DIMMs) 16GB (4 DIMMs) 16GB (4 DIMMs)
Memory type Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1333MHz Corsair Vengeance DDR3 SDRAM at 1600MHz Corsair Vengeance DDR3 SDRAM at 1600MHz Corsair Vengeance DDR3 SDRAM at 1600MHz Corsair Vengeance DDR3 SDRAM at 1600MHz
Memory timings 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T 9-9-9-24-1T
Audio Realtek ALC892 with 2.66 drivers Realtek ALC892 with 2.66 drivers Realtek ALC898 with 2.66 drivers Realtek ALC892 with 2.66 drivers Realtek ALC898 with 2.66 drivers Realtek ALC892 with 2.66 drivers
Graphics Asus Radeon HD 5870 1GB with Catalyst 11.10 drivers
Hard drive Western Digital Caviar Black 1TB 6Gbps

Corsair Force GT 240GB with 1.3.3 firmware

Power Supply PC Power & Cooling Silencer 760W
OS Windows 7 Ultimate x64 with Service Pack 1

We’d like to thank Asus, Corsair, PC Power & Cooling, and Western Digital for helping to outfit our test rigs with some of the finest hardware available. Thanks to each of the motherboard makers for supplying their boards, too, and to Intel and AMD for providing the CPUs.

We used the following versions of our test applications:

The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at a 60Hz 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.

Serial ATA performance

Even with the lights turned out on its SCU component, the X79 still has a couple of wicked-fast Serial ATA ports under its belt. Intel’s 6-series chipsets have traditionally offered better 6Gbps SATA performance than those from AMD, and we were curious to see if that’s still the case. To find out, we pitted the DX789SI against the opposing Z68 and 990FX platforms in a battery of storage tests using a Corsair Force GT 240GB SSD. The SSD was secure-erased before each batch of tests to ensure accurate results.

HD Tune

We’ll begin with HD Tune, which lets us probe transfer rates and access times.

Intel has a clear lead over AMD in HD Tune’s burst speed tests, which can be influenced by how quickly a system’s CPU transitions out of low-power states. Interestingly, the Z68 edges out the X79 by 16MB/s with burst reads. The two are much closer with writes.

The tables turn when we look at average transfer rates. Writes prove problematic for our Z68 example, and we suspect that’s a motherboard issue rather than a chipset flaw. The X79 looks strong overall, and it has a notable lead over the 990FX with sustained reads.

Although we’re only talking about differences of a few microseconds, the Intel platforms have slightly slower access times than the 990FX and its SB850 south bridge. The SATA performance of this particular 990FX motherboard is definitely better than we’ve seen from earlier implementations of the SB850.

IOMeter

IOMeter subjects storage devices to an increasing number of concurrent I/O requests. We’ve limited the load to 32 simultaneous requests, which matches the depth of the Serial ATA spec’s Native Command Queue. Once again, we’re testing with a Force GT SSD.

The Intel chipsets enable slightly higher transaction rates in IOMeter—most of the time, anyway. In the database access pattern, the X79 and Z68 both experience a notable drop in performance at 32 concurrent I/O requests. The two share similar SATA controller logic, so it’s no surprise to see them so closely matched.

Comparing CPU utilization numbers is fraught with peril when dealing with platforms that have different core counts and clock-scaling behavior. Still, it’s interesting to see the Z68 and X79 on even footing considering the disparity in CPU power between them. The 990FX’s CPU utilization is quite a bit higher with both access patterns.

Peripheral performance

The X79, Z68, and 990FX chipsets all have built-in PCI Express 2.0 connectivity. To see which implementation offers the best performance, we attached a Syba SY-PEX40032 PCI Express x1 RAID card to all three. The card was linked to a RAID 0 array striped across two SSDS, a 120GB OCZ Vertex 3 and a 120GB Corsair Force GT.

Although we used the DX79SI for chipset-specific storage tests, we had to switch to the Asus board here. The Syba card worked in the DX79SI, but the card’s performance suggested that it was running at gen-one PCIe speeds. Poking around in the Intel firmware confirmed our suspicions. The firmware plainly states that it initializes some PCI Express devices at 2.5GT/s before kicking them over to 5GT/s in the OS. Windows drivers are supposed to request that transition, but it seems our RAID card does not.

Booting up in 2.5GT/s mode might not be an entirely bad idea. When we popped the Syba card into the MSI X79A-GD65, we got read errors every time we tried to benchmark the array. Those errors only occurred in the PCIe slots associated with the X79; the card was fine in the x16 slots tied to the CPU’s PCIe lanes.

We had no problems running the card in the Asus or Gigabyte boards. It also worked flawlessly in the Z68 and 990FX platforms. The X79 results below are from the P9X79 PRO.

Our solid-state RAID array turned in much faster burst speeds on the X79. However, its maximum and sustained transfer rates were only a little bit higher than on the other platforms.

For what it’s worth, we saw even higher transfer rates from the RAID card when it was installed in one of the MSI board’s PCIe 3.0 slots. The direct line to the CPU surely helped in that case.

The X79 Express lacks an integrated PCI interface, and only a couple of the boards even have PCI slots. Still, I wanted to quickly test PCI performance to ensure that the issues we saw with early P67 boards have been resolved. Since the X79 and most 6-series chipsets lack PCI connectivity, mobo makers have to provide it with third-party silicon. How does the DX79SI’s implementation fare?

Pretty well, at least with a Gigabit Ethernet card. Notice that the Z68 board has lower throughput than the others, though. I wouldn’t worry too much about the CPU utilization numbers based on the differences between the processors used on each platform.

Memory performance

That’s it for chipset testing. Now, it’s time to move onto the motherboards, starting with a look at memory bandwidth. The X79 posse has a distinct edge here because Intel officially endorses the use of 1600MHz memory with Sandy Bridge-E. The CPU also has twice the number of memory channels available on the Z68 and 990FX boards, so faster memory is only the beginning.

All the X79 boards were run with the same Corsair Vengeance DIMMs at 1600MHz with 9-9-9-24-1T timings. As one might expect, there isn’t much difference in the resulting memory bandwidth. The P9X79 PRO and X79-UD5 offer slightly more bandwidth than the Intel board, which is marginally ahead of the MSI. The gaps between those boards are nothing compared to the huge lead they all enjoy over the dual-channel stragglers.

Application performance

At least among the X79 boards, the results of our application performance tests are pretty close. The only one that shows any notable separation is the first pass of the x264 encoding test, where the X79-UD5 is fastest.

Power consumption

We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Readings were taken at idle and under a load consisting of a Cinebench 11.5 render alongside the rthdribl HDR lighting demo. We tested with Windows 7’s High Performance and Balanced power plans.

All the motherboards have additional power-saving features that can be controlled via their firmware or, in the case of the Z68X-UD7, a Windows utility. The names of these power-saving schemes are all different, of course, but they attempt to achieve the same thing: lowering power consumption without restricting performance. Each board was run with its power-saving features disabled and enabled, with the latter config detailed in parenthesis.

With fewer onboard peripherals than its rivals, the DX79SI’s lower power consumption isn’t entirely unexpected. Both at idle and under load, the board draws fewer watts than its X79 competition. In fact, its idle power draw is lower than that of our Z68 and 990FX systems.

The X79A-GD65 largely shadows the Intel board, at least when MSI’s power-saving features are enabled. That power-saving scheme is good for only a couple of watts saved at idle, but it cuts 14-19W under load.

Although Intel’s special power-saving switch is less effective, the board doesn’t have much excess wattage to lose. The same can’t be said for the P9X79 PRO, whose idle power draw is much higher than that of the other boards. Asus’ EPU power-saving scheme may not lower the board’s idle power draw, but it knocks off close to 20W under load.

Gigabyte’s 3D Power settings are less successful under load, shaving only a few watts off the UD5’s power draw. Still, being able to activate the power-saving measures through the firmware is much more convenient than relying on Gigabyte’s old Dynamic Energy Saver Windows software.

Overclocking

Overclocking Sandy Bridge-E presents new challenges and possibilities. With the Sandy Bridge chips we’ve had for nearly a year now, overclocking has been largely limited to changing multipliers, ideally with fully unlocked K-series CPUs. While it’s still possible to increase the CPU speed by ramping up the base clock, you’ll be lucky to push that clock more than a few ticks past its 100MHz default.

With Sandy Bridge-E, you can still expect only a few MHz of useful range with the base clock. However, 100MHz isn’t the only starting point. On the boards we’ve seen, users have the option of jumping up to 125, 167, and even 250MHz base clocks. Your CPU will have to be up to the task, but selecting these higher base clocks shouldn’t result in other elements of the system running out of spec.

Of course, it’s still possible to overclock by raising the CPU’s multiplier, either with a blanket approach that applies to all six cores or by tuning the per-core Turbo settings. Unfortunately, my time with these X79 boards was far too limited to deeply probe their overclocking potential. I’ve only had my Sandy Bridge-E CPU for ten days, and final boards, firmware, and drivers were showing up late into last week. Since sleep is for wussies, I did manage to spend some time with the auto-overclocking mechanisms built into each board. I also tried my hand at a little manual overclocking, but only by adjusting the CPU multiplier. We’ll have to save the base clock for another day.

I promised a closer look at Asus’ auto-tuning utility, and here it is. Unlike the other automatic overclocking schemes, which pick a higher clock speed based on what they think your CPU can handle, Asus’ TurboV tuning system slowly cranks the CPU speed while testing for stability along the way. This iterative process is slower, but it should deliver better results.

TurboV took our Core i7-3960X up to 4.6GHz without even touching the CPU multiplier. Instead, it increased the base clock speed up to nearly 127MHz. This speed was perfectly stable with a 12-way Prime95 load, which is what we used to verify the stability of all our overclocked configs.

After letting TurboV have its fun, I dropped the base clock back to 100MHz and started increasing the CPU multiplier. Despite using the Intel-branded water cooler, the CPU started throttling under load after I hit 45X, or 4.5GHz. The system was still stable, but it would scale back to a 33X multiplier before ramping back up. This behavior persisted as I pushed on to 4.9GHz. However, at 5GHz, the system crashed under load. The board had automatically raised the CPU voltage to nearly 1.5V by this point, so it didn’t seem wise to add more juice—at least not with other boards left to test.

Gigabyte’s built-in overclocking intelligence can be invoked by a button in the rear cluster or through the QuickBoost section of the EasyTune6 Windows utility. Either way, the board picks a target clock speed based on the results of internal testing that Gigabyte has done with a fleet of its own CPUs. For our core i7-3960X, that speed was 4.5GHz using the default 100MHz base clock.

The system was stable at this speed, making 4.6GHz the next stop. Hitting that speed required upping the CPU voltage to 1.375V, and that’s when the throttling began. It took a few minutes of Prime95 for the multiplier to start oscillating, and the additional voltage required to hit higher speeds seemed to hasten throttling’s arrival. The CPU hit 4.8GHz on 1.44V, but 4.9GHz locked the system under load, even with extra voltage. Increasing the CPU power limits didn’t help, either.

MSI’s OC Genie can be summoned with the push of a button, but he’s more conservative than a Fox News contributor. The auto-overclocking scheme pushed the CPU to just 4GHz with a 40X multiplier, a good 500MHz shy of the Asus and Gigabyte implementations. Stability wasn’t an issue at this speed, but we hoped the board would do better with a little personalized attention—and it did.

The GD65 had no problem running the CPU at 4.5GHz on 1.4V. Throttling didn’t rear its head until the voltage was cranked up to 1.47V to get the system stable at 4.7GHz. That proved to be the end of the road. 4.8GHz wasn’t stable under load, and increasing the voltage made no difference. Neither did a virgin sacrifice.

Intel doesn’t have an automatic overclocking option for the DX79SI, but the board fared pretty well when we took overclocking into our own hands. 4.5GHz was stable and throttle-free, and it took longer than on the other boards for throttling to kick in at 4.6GHz. Intel provided a few recommendations on the power setting to use when reaching for higher speeds, and using them, we managed to get up to 4.8GHz on 1.44V. Throttling came quickly at that speed, and 4.9GHz crashed the system under load, even with extra voltage applied.

All of this overclocking was done on an open test bench using the Intel-branded water cooler, which doesn’t generate any airflow around the VRMs and their associated heatsinks. With airflow directed at the VRMs, it’s possible we could have hit higher speeds on some of these boards. There also seems to be some correlation between higher voltages and throttling. The Asus board seemed to throttle quicker than the others, but its auto voltage scaling also picked higher CPU voltages than what we defined manually with some of the other boards.

Interestingly, the CPU temperatures we observed with each motherboard’s monitoring software were well short of what we would have expected to invoke throttling. The apps didn’t seem to be updating the CPU temperature as rapidly as may have been necessary to detect the spikes, though. Either way, pushing Sandy Bridge-E clock speeds much past their defaults requires some serious cooling.

Motherboard peripheral performance

We’re almost done, but before we get to the conclusion, I’m going to drown you in a mass of peripheral performance tables. Thanks for making it through a dozen pages! We’ll begin with USB performance using a Thermaltake BlacX docking station equipped with the same Corsair Force GT SSD from our earlier storage testing.

HD Tune USB 2.0 performance
Read Write
Burst (MB/s) Average (MB/s) Random 4KB (ms) Burst (MB/s) Average (MB/s) Random 4KB (ms)
Asus Sabertooth 990FX 29.4 33.1 0.88 31.2 30.3 0.78
Gigabyte Z68X-UD7 29.5 32.6 0.56 29.5 27.1 0.55
Asus P9X79 PRO 31.3 33.2 0.56 31.9 29.1 0.54
Gigabyte X79-UD5 31.3 33.2 0.55 31.3 29.1 0.56
Intel DX79SI 32.1 31.9 0.61 30.8 26.7 0.57
MSI X79A-GD65 (8D) 31.3 33.2 0.52 31.9 29.1 0.53

Yeah, it’s probably time to axe USB 2.0 performance from our suite of motherboard tests. Although there are some minor differences in read and write speeds between the boards, they’re all slow when compared to the throughput provided by USB 3.0.

HD Tune USB 3.0 performance
Read Write
Burst (MB/s) Average (MB/s) Random 4KB (ms) Burst (MB/s) Average (MB/s) Random 4KB (ms)
Asus Sabertooth 990FX 132 154 0.09 127 145 0.09
Gigabyte Z68X-UD7 176 191 0.11 169 142 0.10
Asus P9X79 PRO 177 189 0.11 168 164 0.10
Gigabyte X79-UD5 178 183 0.11 171 165 0.10
Intel DX79SI 182 189 0.10 172 163 0.09
MSI X79A-GD65 (8D) 180 186 0.11 169 130 0.11

Again, though, the deltas between the X79 boards are relatively narrow. The Asus, Gigabyte, and Intel models are closely matched even though they power their SuperSpeed ports with different controller chips. MSI keeps up with reads, but it’s more than 30MB/s slower with sustained writes despite using the same NEC controller as the Intel board. The GD65’s write burst speed is at least competitive with the rest of the field.

HD Tune Serial ATA performance
Read Write
Burst (MB/s) Average (MB/s) Random 4KB (ms) Burst (MB/s) Average (MB/s) Random 4KB (ms)
Asus Sabertooth 990FX 285 467 0.038 260 381 0.042
Asus Sabertooth 990FX (JMicron) 108 123 0.083 104 39 0.088
Gigabyte Z68X-UD7 388 490 0.041 342 279 0.046
Gigabyte Z68X-UD7 (Marvell) 226 353 0.089 212 116 0.101
Asus P9X79 PRO 380 506 0.030 351 376 0.029
Asus P9X79 PRO (Marvell) 2250 355 0.064 232 143 0.075
Gigabyte X79-UD5 372 497 0.046 339 373 0.046
Gigabyte X79-UD5 (1) 262 346 0.072 243 206 0.078
Gigabyte X79-UD5 (2) 260 345 0.074 244 205 0.079
Intel DX79SI 372 508 0.046 341 379 0.046
MSI X79A-GD65 (8D) 367 494 0.046 338 372 0.045
MSI X79A-GD65 (8D) (Marvell) 255 361 0.080 239 258 0.083

Asus and Intel have done the best job of implementing the X79’s 6Gbps SATA controller, which is slightly slower on equivalent offerings from Gigabyte and MSI. In comparison, the performance of the auxiliary storage controllers is uninspiring to say the least. With the exception of the Marvell chip on the P9X79 PRO offering a crazy-fast read burst speed (thanks, no doubt, to using a slice of system memory as a read cache), the secondary storage options are all much slower than the Intel chipset. With sustained reads and writes, the differences are well over 100MB/s. No wonder Intel didn’t bother with third-party controller chips on the DX79SI.

NTttcp Ethernet performance
Throughput (Mbps) CPU utilization (%)
Asus Sabertooth 990FX 943 11.0
Gigabyte Z68X-UD7 (1) 939 4.1
Gigabyte Z68X-UD7 (2) 942 3.0
Asus P9X79 PRO 942 2.2
Gigabyte X79-UD5 944 1.8
Intel DX79SI 939 3.3
Intel DX79SI (Intel) 943 1.8
MSI X79A-GD65 (8D) 944 1.8

There isn’t much difference in Gigabit Ethernet throughput or CPU utilization between the X79 boards.

RightMark Audio Analyzer audio quality: 24-bit/192kHz
Frequency response Noise level Dynamic range THD THD + Noise IMD + Noise Stereo Crosstalk IMD at 10kHz Overall score
Asus Sabertooth 990FX 5 4 4 5 3 4 5 5 4
Gigabyte Z68X-UD7 5 5 5 5 3 5 5 5 5
Asus P9X79 PRO 5 5 5 5 3 5 5 5 5
Gigabyte X79-UD5 3 6 6 5 3 1 5 5 4
Intel DX79SI 5 4 4 5 3 5 5 5 4
MSI X79A-GD65 (8D) 5 4 4 5 3 5 5 5 4

RightMark Audio Analyzer probes analog audio signal quality using the front-channel output and line input. The scores are close overall, but closer examination reveals a couple of low points for the Gigabyte board in the frequency response and intermodulation distortion tests. The X79-UD5 uses the same ALC898 codec chip as the P9X79 PRO, which has more consistent scores across the board and the best overall score of our four contenders.

Conclusions

Sandy Bridge-E is a beast, no doubt about it. In addition to six Sandy Bridge cores, she’s loaded with quad memory channels and 40 lanes of third-generation PCI Express. It’s a good thing, too, because her X79 sidekick appears relatively weak in comparison. Things weren’t supposed to go down this way; Intel intended the X79 to have gobs of additional SATA and SAS ports, but it couldn’t get the new storage controller working in time, leaving a neutered X79 that looks more like a P67 platform hub. Fortunately, the lack of additional 6Gbps SATA connectivity is somewhat balanced by the CPU’s massive PCI Express payload, which provides plenty of high-speed lanes for the PCIe-based SSDs. High-performance solid-state storage is heading in that direction, anyway.

That said, it seems likely the X79’s ill-fated storage controller may be responsible for the rushed feel of this particular platform launch. Intel didn’t have an AHCI and RAID driver ready until the middle of last week, and the one it released is tagged as a beta. Then there are the motherboards, which are a little rougher around the edges than the first batch of Sandy Bridge boards. I’ve given up on trying to keep up with all the firmware updates that have hit my inbox over the past few days. At least the interfaces are much improved over what we saw at the beginning of the year.

If I were to put together a Sandy Bridge-E system for myself today, I honestly don’t know which of the four motherboards I’d use. They all have admirable attributes, but each one also has flaws, some of which are more serious than others.

Perhaps the most striking shortcoming can be found on the Intel DX79SI, which fails to take advantage of all the PCIe 3.0 connectivity built into the CPU. Eight lanes are left on the table, and while you probably don’t actually need the extra bandwidth, it’s hard to justify paying the same price for a motherboard that offers less than its rivals. To be fair, the Intel board is the most power-efficient of the bunch, and the wireless module is a nice touch. The fan controls are excellent, too, even if the accompanying firmware interface is uninspired. This board grew on me the more I used it, but the PCIe config soured my crush.

The MSI X79A-GD65 (8D) shares a similar price range with the Intel board. Instead of wireless connectivity, it sports a couple of extra 6Gbps SATA ports and a slick firmware GUI. Unfortunately, the fan controls are the worst of the bunch, and the auto-overclocker is timid at best. The board’s sluggish USB 3.0 write speeds don’t help its case, either. MSI seems to be heading in the right direction, but it’s not there yet.

Gigabyte isn’t either, but it’s made notable improvements, too. The 3D BIOS interface has definite promise, even if it needs some tweaking and polish. The UD5 is also jam-packed with other goodies, including plenty of wireless options, real-time Dolby Digital Live! encoding, and actual USB 3.0 ports for your front panel. Gigabyte has delivered the kind of excess one might expect from a $350 motherboard, but it’s done so with relatively high power consumption and a mess of different tweaking utilities.

The P9X79 PRO sits in the middle of the pack we’ve put together, and it’s probably the most complete offering of the bunch. Asus continues to add thoughtful new features, and the PRO has the best integrated audio in addition to great tuning software, excellent fan controls, and a next-generation firmware interface that continues to evolve. The flashy interface doesn’t make up for Asus jacking up the Turbo multipliers behind the user’s back, though. The PRO also has much higher idle power consumption than the other boards.

While the Asus and Intel boards are my favorites among the four, both are difficult to fully endorse with our TR Recommended logo. I’d feel a lot better about the P9X79 PRO if Asus rectified its Turbo behavior, though. If only the DX79SI’s PCI Express lanes were as easy to address.

Comments closed
    • theampersand
    • 8 years ago

    This might have been addressed already, but doesn’t the AMD 990FX chipset use the SB950 southbridge? The article lists it as the SB850…

    [url<]http://en.wikipedia.org/wiki/AMD_900_chipset_series[/url<] [url<]http://www.newegg.com/Product/Product.aspx?Item=N82E16813131736[/url<]

    • Bensam123
    • 8 years ago

    Motherboard makers need to cut out the end on their PCI-E x1, x4, and x8 slots to fit bigger cards. I still don’t know why they don’t do this.

    • DarkUltra
    • 8 years ago

    Would you get better overclocking than 4.9GHz on the deluxe version of the Asus P9X79? It has 16 + 4 phase CPU power. Would the throttling be reduced too?

    I guess I would have no problem with throttling if normal CPU usage wouldn’t trigger it, like gaming, video encoding and rendering etc. Maybe you could look into this? With a 4.9GHz overclock it is tempting over a 2700K.

    • zorglub
    • 8 years ago

    I am disapointed. The number of expansion slots is lower than when using an X58: Asus P6T6 WS Pro X58 base gives more PCIe slots. I usually need 1 16x, 3 8x and 1 x4 lanes.

      • IlluminatiASUS
      • 8 years ago

      The P6T6 WS had an NF200 ,if we were to place a NF200 on the board you would limit the ability to support Gen3 mapping across the PCI-E lanes. Additionally with the increased bandwidth and more efficient encoding pipeline offered on Gen3 there is no need for an NF200.

      As for PCI-E the board supports Native 3 way or 3 way crossfire which covers almost the 99% of the actual market. Should you need more PCI-E expansion i would go for the P9X79 WS board.

      3 x PCIe 3.0/2.0 x16 (dual x16 or x16, x8, x8) *1
      1 x PCIe 3.0/2.0 x16 (x8 mode) *1
      2 x PCIe 2.0 x1

      Still looks pretty good to me in regards to expansion.

    • spuppy
    • 8 years ago

    The Intel board has an auto overclocking function in the BIOS called “Overclocking Assistant”. Simply choose your intended speed, and it sets everything for you. On the 3960X there are settings up to 4.6 GHz. It will also OC with the Host Clock multiplier if you want.

      • kamikaziechameleon
      • 8 years ago

      cool. There is allot of interesting in this processor/mobo launch but so many missteps I don’t want to one of the first ones in.

      • Dissonance
      • 8 years ago

      The overclocking assistant isn’t a whole lot different than how things work on the other boards. With the Intel, you choose the actual CPU speed rather than selecting a multiplier, which I suppose saves some math. However, when you change the multiplier on the other boards, variables like voltages are also tweaked to keep pace… as long as they haven’t been altered from their “auto” defaults.

      • IlluminatiASUS
      • 8 years ago

      Keep in mind that with this function there is a fixed voltage that is supplied at this preset and if the CPU has a variance in the voltage required to sustain this frequency then your system will not post/boot correctly or fail to achieve stability. That being said this is generally a really of any fixed OC profile/preset.

    • TechNut
    • 8 years ago

    Those Syba RAID cards are notoriously unstable! I have several of them, and they are all buggy. They all toss read errors for ESATA. They use a Silicon Image 3124 or similar chip. It’s supported but barely.

    • LoneWolf15
    • 8 years ago

    “Asus’ last generation of motherboards switched to digital power regulation circuitry for the CPU. With its X79 family, digital VRMs have also been applied to the memory. The digital circuitry is purportedly more accurate and responsive than analog alternatives, which Asus claims improves overall stability. A total of eight power phases is dedicated to the CPU core, while another two feed the rest of the chip. An additional four power phases are split between the system’s two banks of DIMM slots.”

    Geoff, this is the kind of information I like to know about. Thank you for your extensive attention to detail.

    Do you know if going digital changes heat output? I’ve noticed that ASUS’ P8P67 and P8Z68 mainboards seem to have less beefy heatsinks than competitors.

      • kamikaziechameleon
      • 8 years ago

      ^^^+1 this!

      • IlluminatiASUS
      • 8 years ago

      Technically the part of the board that produces the most heat is the driving element of the VRM ( Mosfet and Driver ( the PWM which is digital ) does not really produce much heat ( unless you are talking a legacy digital solution like a Volterra controller. As such the main benefit of going digital is the extensive control on how you tweak the regulation of the power delivery system. You can modify the VRM to be balance on temperature and limit current delivery or you can balance via current for max power ( but more heat ). Other options are to modify the loading policy at different load states for the phase switching. All this and more is configurable with a well executed digital design. In conclusion you achieve reduced temperatures but the configuration parameters need to be defined correctly.

    • kamikaziechameleon
    • 8 years ago

    This definitely stinks of a fumbled launch, between the 2 disabled cores on the processors themselves and these uneven Mobo offering with the premium price my excitement for the platform has kinda died.

    It feels extremely forced/rushed and with the lack of relevant competition there is no real reason to have such a weak mobo offering. The 2 missing cores stinks of intel basically stretching out the performance gains curve for the fab revision of this architecture so its less pointless of a point to jump in as a consumer.

      • swaaye
      • 8 years ago

      Yeah Intel is segmenting like crazy these days. Z68 being the same silicon as P67 and H67, etc, but with features turned on. 2 cores disabled here because they don’t need the speed to compete and it helps improve yields along with making it easier to fit in a power spec without lowering voltage or cherry picking less leaky silicon. And of course the wacky CPU / IGP features segmentation. All sorts of artificial product value micromanagement.

        • kamikaziechameleon
        • 8 years ago

        I personally would rather see a product designed for the price point they aspire to sell it at rather than this goofiness. Its worse than what car companies do. In todays world there is almost no reason to segment processors into more than a sub 200 category, sub 1,000 category and over 1,000

        Server vs. consumer is also a load of BS most of these processors are near identical and could literally be sold as one processor with workstation/server settings optional. Its almost as bad as the software segmentation like what adobe and Microsoft are guilty of.

        I absolutely hate how fast intel cycles through sockets and the premium their mobos go for.

        Gosh I’m not currently to happy with the business practices of either company though I do enjoy some of their products.

    • derFunkenstein
    • 8 years ago

    It’s not really an exciting chipset, but that’s going to be true as you move things onto the CPU die. What’s worse, the boards themselves are not terribly exciting. And as far as motherboards go, that’s probably a good thing.

      • continuum
      • 8 years ago

      The chipset would be more exciting if the extra SATA/SAS ports actually worked… and had USB 3.0… and yeah while I’m at it, I’ll wish for world peace and a pony!

      • kamikaziechameleon
      • 8 years ago

      I would disagree. As stuff moves onto the CPU there is less overhead on Mobo manufacturers so we should see better bundled tech as a result. Built in 11n wireless, blu tooth, better audio or a bundled discrete sound card, Further support for different connections standards etc. With a product in this price range I’d expect all of these from those mobos but don’t see it strangely.

        • derFunkenstein
        • 8 years ago

        I can add all those on if I want them and don’t have to pay for them now. Otherwise what do I do with all the expansion slots? 😀

        I prefer the path they’re on with regards to better OC and stability due to upgrades in power regulation.

          • kamikaziechameleon
          • 8 years ago

          I would agree, I don’t think a 300 dollar mobo should be able to do both though. 300 bucks for OC potential kinda is a rip off. I’ve OC’d just fine on 80 dollar mobo’s.

          300 bucks for a mobo then another 200 for addon cards to get basic functionality that comes bundled in a 300-400 dollar net book is kinda BS.

            • kamikaziechameleon
            • 8 years ago

            Then again the last two TR recommended mobos I got have been bogus spent more than 200 on each and yet the 80 dollar mobo not recommended was best mobo i’ve bought in the last 5 years.

            • derFunkenstein
            • 8 years ago

            Yeah Iv’e got a $120 mobo with “good enough” onboard features and a nice set of voltage regulation circuitry and I’m super-dee-duper happy with it.

            • kamikaziechameleon
            • 8 years ago

            I’m really going to be hard on a 300 dollar product like this. I’ve never bought a premium mobo and been happy but been plenty happy with budget ones.

            • IlluminatiASUS
            • 8 years ago

            I think you are correct that a strong budget board can offer strong value but due to a limit budget that also means you are limited in regards to connetivity and a number of other aspects. I think higher segementation boards serve their respective users. As such if you do not need additional USB3 or 3 way support then you go for the board that offers 2 way with less usb3 as an example. It is alot more complicated than that though as smaller details like number of fan headers and the super I/O controller required to control them add cost to a board. In the end you pick the board that serves your needs.

    • MadManOriginal
    • 8 years ago

    Every time someone calls solid polymer capacitors ‘solid-state capacitors’ instead, God kills a kitten.

      • LaChupacabra
      • 8 years ago

      Wait, we shouldn’t say “solid-state capacitors?” Damn, because I find talking about “solid-state capacitors” nearly unavoidable sometimes. Infact, I probably say “solid-state capacitors” half a dozen times a day! In this reply alone i’ve said “solid-state capcitors” three times already! Crap, that was 4. See? It’s pretty hard not to say “solid-state capcitors.”

      =D

        • Bauxite
        • 8 years ago

        [domos chasing kitten.jpg]

          • derFunkenstein
          • 8 years ago

          I had to google it, but hey it’s pretty funny.

          [url<]http://jasonsgalaxy.com/images/domo%20&%20domo%20vs.%20cat.jpg[/url<]

    • green
    • 8 years ago

    page 7 sata performance, last paragraph:
    [quote<]Although we're only talking about differences of a few [b<]micro[/b<]seconds, the Intel platforms have slightly slower access times than the 990FX and its SB850 south bridge.[/quote<]

      • green
      • 8 years ago

      page 13 motherboard peripheral performance, HD Tune Serial ATA performance:
      [quote<]Asus P9X79 PRO (Marvell): Burst (MB/s) - 2250[/quote<] ... that seems a little high... edit: nevermind. re-read the paragraph under it again.

Pin It on Pinterest

Share This