A couple of Core i7 motherboards compared

By now you’ve no doubt read all about Intel’s new Core i7 processor. You know that the CPU otherwise known as Nehalem is based on an all-new architecture, complete with four cores (or eight, if you count Hyper-Threading), three levels of cache, an integrated memory controller, and a QuickPath Interconnect replacement for the front-side bus. You know that the Core i7 is unequivocally the fastest desktop processor around, and that in some cases, it’s a fair bit quicker, clock-for-clock, than Intel’s already impressive Penryn designs. Naturally, you want one. Badly.

Saddling up a Core i7 won’t be as easy as plugging one of the new chips into your current system, though. Nehalem is built on a new LGA1366 package that isn’t compatible with the existing LGA775 infrastructure. Without a front-side bus, the Core i7 also requires a chipset compatible with its QuickPath Interconnect: Intel’s own X58 Express. You’ll need a whole new motherboard to join the Core i7 party, then. But which one?

Intel has its own X58 board, of course—the latest in a series of attempts to effectively cater to overclockers and PC enthusiasts. As one might expect, Taiwanese giant Asus has an iron in the fire, too. We’ve had the P6T Deluxe in the labs for weeks now, and it’s laden with all the rich features you’d expect from a perennial favorite.

To find out which of these two is the best option for those eager to adopt Intel’s latest microarchitecture, we’ve run them through a grueling gauntlet of performance, peripheral, power consumption, and overclocking tests. Read on for a detailed look at the first Core i7 motherboards to make it into the Benchmarking Sweatshop.

Introducing the X58 IOH

The Core i7’s new interconnect requires a compatible chipset, and right now, the only game in town is Intel’s X58 Express. Only one component of this fresh core logic offering is actually new, though. That would be the north bridge chip, which Intel refers to as the IOH, or I/O Hub. Previous Intel chipsets have branded their north bridge components MCHs, or Memory Controller Hubs, but since the Core i7 moves the memory controller to the processor, IOH is much more fitting for the X58.

Intel manufactures the X58 IOH using 130nm process technology that feels a little antiquated given the company’s fabrication prowess. After all, AMD has been building north bridge chips on a power-efficient 55nm node since March. One might expect the X58 to present a thermal challenge; however, according to at least one motherboard maker, the chip is no more difficult to cool than Intel’s last flagship, the X48 Express.

Thanks to its lack of a memory controller, the X58 is a much simpler design than its predecessor. The chip is essentially a bundle of point-to-point links, the most important of which is a QuickPath Interconnect that hooks into the Core i7 processor. This interconnect is made up of a pair of 20-bit links (one for upstream and another for downstream) that push 4.8 or 6.4GT/s, depending on your choice of Core i7 processors. QPI only uses 16 bits for data (the other four are reserved for error-checking), giving the interconnect an effective 19.2 or 25.6GB/s of aggregate bandwidth—a much bigger pipe than even Intel’s fastest front-side bus.

The X58 IOH

All that processor bandwidth will certainly come in handy given the fact that the X58 IOH also sports 36 lanes of second-generation PCI Express. Four of those lanes are reserved for peripherals and expansion slots, with the remaining 32 dedicated to graphics. Naturally, those graphics lanes can be split evenly between a pair of full-bandwidth x16 slots. It’s also possible to arrange the lanes in a three-slot x16/x8/x8 setup and in a four-way-x8 config.

Like previous Intel chipsets, the X58 Express has full support for AMD’s CrossFire multi-GPU scheme, including three- and four-way configs. SLI will also be supported—a first for Intel platforms (not including uber-expensive Skulltrail systems). However, actual SLI certification will be done at the motherboard level rather than being tied to the chipset. Certified motherboards will have a special key embedded in their BIOS that Nvidia’s graphics drivers will check prior to enabling SLI. To date, Asus, DFI, ECS, EVGA, Foxconn, Gigabyte, and MSI have licensed SLI for their X58 boards. Intel is conspicuously missing from that list, though. When asked whether its X58 board would support SLI, Intel said it and Nvidia had “not found mutually acceptable business terms for certification.”

The last ingredient in the X58’s connectivity soup is Intel’s DMI interconnect. This link offers up to 2GB/s of bandwidth, which seems a little light given the massive pipes seen elsewhere in the Core i7. However, DMI does allow the X58 to hook into Intel’s existing ICH10 series south bridge chips.

Those familiar with Intel’s P45 chipset will no doubt recognize the ICH10, and more specifically, the RAID-equipped ICH10R. You’d recognize this chip if you’re familiar with Intel’s last generation of chipsets, too, since it’s little more than a die shrink of the old ICH9 series. There’s nothing new here, but that’s not necessarily a sign of weakness. After all, the ICH10R does pack six 300MB/s Serial ATA ports with support for novel Matrix RAID configurations and more traditional multi-drive arrays. You get a dozen USB ports, too, and an integrated Gigabit Ethernet MAC. Six gen-one PCIe lanes round out the package, offering more than enough bandwidth for onboard peripherals and expansion slots.

Overall, the ICH10R has feature parity (and in some cases, superiority) with its rivals from AMD and Nvidia. Even more importantly, the chip has proven itself to be remarkably reliable and free of niggling little issues. Given the current core logic chipset climate, that’s a very good thing.

Asus’ P6T Deluxe
A little something for everyone

Manufacturer Asus
Model P6T Deluxe
Price (MSRP) $309
Availability Soon

Perhaps more than those of any other motherboard maker, Asus’ enthusiast offerings have been the most consistently good over the years. The world’s largest mobo maker has been catering to the gaming, overclocking, and do-it-yourself crowd for a very long time, and it shows. It’s no surprise, then, that the first X58 board to hit our labs was the P6T Deluxe. Asus also has a Republic of Gamers Rampage II Extreme board ready for Nehalem, but it’s a little up-market from the P6T. Since the Deluxe already carries a hefty $309 suggested retail price, we thought it was the best one to tackle today.

The P6T Deluxe is appropriately dressed in orange and black for the season, with a splash of brilliant blue thrown in for good measure. If anything, high-end motherboards have toned down their aesthetics lately, replacing some of the flashier palettes of a few years ago with muted but no less attractive tones. Restraint is a good thing, in this case, because the P6T looks about as sharp as can be expected from a motherboard.

You can tell that Asus’ board designers have been at it for a long time because the P6T Deluxe’s layout is virtually free of clearance issues. That’s not easy to do, particularly when you have to squeeze in all the additional chips and slots associated with Asus’ Deluxe offerings.

Power plug placement is the first potential clearance quirk that I look for in a motherboard—you don’t want power cabling crowding the CPU socket or interfering with airflow between the processor heatsink and chassis exhaust fan. Asus does well here, placing the auxiliary 12V connector out of the way along the top edge of the board. This layout works well with traditional enclosures that place the power supply above the motherboard, although it’s a little less convenient for those new-fangled upside-down enclosures like Antec’s P180 series.

As one might expect from a high-end motherboard, the P6T Deluxe surrounds its shiny new LGA1366 socket with heatpipes and cooling fins. The heatsinks are passive, low-profile designs, so they won’t make any noise or interfere with larger coolers that fan out from the socket. Somewhat surprisingly, the cooler’s north bridge component doesn’t have much in the way of actual surface area, at least when compared to the arrays of thin fins that cover the board’s voltage circuitry heatsinks. That starts to make sense when you consider that the Deluxe is outfitted with a whopping 18 power phases; there are 16 for the processor core and another two for its memory controller and QPI component.

Next to the socket we find six DDR3 memory slots. The board supports up to 2GB of memory per slot, for 12GB overall. Naturally, Asus hasn’t been content to stick with Intel’s official 1066MHz memory speed ceiling; DDR3 memory speeds are advertised up to 1600MHz.

While we’re talking memory, it’s worth pointing out that the P6T’s DIMM slots are exactly centered relative to the CPU socket. Asus says this both shortens and equalizes trace lengths to memory, allowing for more stable overclocking.

Moving south brings us to the board’s cluster of Serial ATA ports. The six red ports branch off the ICH10R south bridge chip while the remaining two orange ones are tied to a Marvell 88SE6320 Serial Attached SCSI controller. Backwards compatibility built into the SAS standard allows plain old Serial ATA drives to work with the SCSI controller, and it’ll do RAID 0 and 1 arrays, too.

Asus arranges the P6T’s SATA ports in a tight cluster, with six of them facing the outer edge of the board. Doing so ensures plenty of clearance for longer double-wide graphics cards—something that surprisingly few high-end motherboards get right. Speaking of clearance, the low-profile south bridge cooler won’t interfere with expansion cards, either.

Over to the left of the SATA ports are onboard power and reset buttons. These are incredibly handy if you run an open test bench, but they’re largely useless otherwise. At least Asus knows how to make life a little easier for its internal testers (and motherboard reviewers).

The P6T’s slot stack packs a little something for everyone. Three physical x16 slots are available for graphics; the top slot gets a full 16 lanes of bandwidth, with the bottom two split in an x16/x1 or dual-x8 config. However, because the bottom two slots are side-by-side, three-way configs can’t accommodate a double-wide card in the second slot (or the third, depending on how your case is laid out). That’s not likely to be a serious limitation for most, but it’s one that could have been avoided.

Additional PCI Express lanes can be found in an x4 slot that’s notched to accommodate longer x8 and x16 cards. Asus also throws in a couple of PCI slots, one of which will be blocked by double-wide primary graphics cards.

From here, we can see a little riser card poking out from between the second PCI and PCIe x16 slots. This module stores the board’s embedded ExpressGate instant-on Linux distribution, which offers basic functionality like a web browser and Skype. ExpressGate is a nice little touch, but one whose utility would be improved vastly by the addition of a comprehensive system stress testing application for overclockers.

Forward-looking is perhaps the best way to describe the P6T’s port cluster. The only concession to old-school connectivity is a single PS/2 port. Otherwise, the board’s I/O payload is laden with USB, Firewire, Ethernet, and audio ports. We especially appreciate the inclusion of eSATA connectivity and two flavors of digital S/PDIF output, although it would have been nice to get a digital audio input, as well.

Asus backs the board’s Ethernet ports with a pair of GigE chips from Marvell. More intriguingly, the P6T’s audio codec is supplied by Analog Devices, which recently announced that it’s getting out of the PC audio business. Analog Devices codecs have long been featured on Asus boards, but this one doesn’t support real-time DTS or Dolby Digital Live encoding, restricting multi-channel digital audio output to applications that already have pre-encoded audio tracks—that’s fine for movie playback, but it doesn’t work for games.

In addition to the ports available in the I/O cluster, the P6T also has onboard headers for an additional Firewire port and six more USB ports. That brings the board’s USB count up to 14 (15, if you count the USB connection typically used by the ExpressGate module), which is a little higher than the dozen ports offered by the ICH10R. We suspect there’s some sharing involved, but Asus has yet to confirm exactly how it’s feeding the extra USB ports.

The P6T wouldn’t be Deluxe without a little something extra in the box, and Asus obliges with its OC Palm external display module. Introduced as the ScreenDuo with Asus’ Vista Edition motherboards last year, this little device hooks into the board via a USB port and offers a 2.5″ QVGA display, a four-way directional pad, and four input buttons. OC Palm is designed to interface with Asus’ Windows overclocking and hardware monitoring software, allowing users to tweak settings and keep tabs on system variables without consuming screen real estate or interrupting full-screen gaming.

While its interface is fairly simplistic, OC Palm allows users to adjust the board’s base clock frequency and voltages for the processor core, memory bus, and QPI link. On the monitoring front, the device tracks motherboard and processor temperatures, fan speeds, and the voltage running to the processor, 3.3V, 5V, and 12V lines. Someone at Asus, it seems, has been reading my blog.

OC Palm looks promising, and it’s certainly a unique feature in a market largely devoid of new ideas. I’d like to see its monitoring component expanded to support user-configurable graphs that track a collection of system variables on a single screen, and that may very well happen, since Asus expects to expand OC Palm’s functionality through software updates.

Intel’s DX58SO
Serving up a Smackover

Manufacturer Intel
Model DX58SO
Price (Estimated) $300-350
Availability Soon

With the exception of the lackluster gaming performance of its integrated graphics chipsets, Intel’s core logic offerings are widely regarded as the best in the business. Even notoriously fickle enthusiasts have flocked to them, dating all the way back to the venerable BX of yesteryear, and more recently, with the P35 and P45 Express. Given this chipset pedigree, it’s a little surprising that Intel hasn’t been able to build a motherboard that has really caught on with enthusiasts. Intel wasn’t interested in catering to overclockers back in the day, of course; tweaking clock speeds was officially taboo back then. These days, though, Intel’s high-end motherboards are loaded with overclocking options, littered with pictures of skulls, and branded with edgy code-names like Skulltrail and, er, Smackover.

Say what?

Smackover is indeed the code-name for Intel’s X58-based DX58SO motherboard, and it must be some sort of inside joke. Given the Core i7’s performance, Smackdown might’ve been more appropriate. But at least the board looks like a proper enthusiast offering. It comes draped in a black and blue color scheme that looks pretty good, and Intel hasn’t gone overboard with its little skull logos, which are tastefully applied.

The first thing you’ll notice about the DX58SO is its unconventional layout. The DIMM slots have been placed parallel to the top edge, the north bridge chip to the right of the CPU socket, and the processor’s voltage circuitry across the middle of the board. This arrangement is unique among the X58 boards that we’ve seen thus far, and some of it’s quite clever.

Take the placement of the auxiliary 12V power connector, for example. The plug sits mid-way down the board, but over to the right-hand side, where associated cabling won’t block airflow between the processor cooler and rear chassis exhaust. Because the 12V plug is equidistant from the top and bottom of the board, cable reach shouldn’t be an issue with either a traditional or upside-down enclosure.

The DX58SO’s socket layout actually reminds me a little of the BTX form factor, which also put the DIMM slots parallel to the path of chassis airflow. This should help memory modules run cooler than with more traditional layouts, although you’ll only be able to use four DIMMs at a time. The DX58SO still taps all three of the Core i7’s memory channels, but two of those channels are limited to a single slot each. Intel says it went with only four slots so that memory would be closer to the CPU, reducing latency. That’s also why the slots are exactly centered relative to the socket.

Putting DIMM slots so close to the CPU socket may create clearance problems if extremely tall memory modules like Corsair’s Dominators are paired with larger aftermarket coolers. The board’s north bridge cooler is a little on the tall side, too, and the manual suggests that it should be used with an included 40mm fan. We didn’t find the fan to be necessary when running at stock speeds, and it didn’t help us when overclocking, so you can probably do without. A high-end motherboard really shouldn’t have to rely on active chipset cooling, anyway.

By now you’ve no doubt noticed that the DX58SO is entirely without heatpipes—a pleasant change of pace given the tangled networks of copper we’ve seen on many enthusiast boards. I don’t expect the lack of plumbing will compromise the board’s ability to cool various components, especially since it won’t be sharing heat between them.

Considering that Intel dropped support for “parallel” ATA from its south bridge chips a while back, it’s no surprise that the DX58SO is entirely devoid of IDE ports. Given the popularity (and affordability) of Serial ATA optical drives, there’s really no reason to saddle a system with bulky ribbon cables anymore.

All of the board’s Serial ATA ports are lined up along the right edge, spaced in pairs of two. These SATA couplets are carefully spaced such that longer graphics cards—even those with bulky double-slot coolers—won’t interfere with cabling. The low-profile south bridge heatsink won’t get in the way of longer cards, either.

The DX58SO serves up an onboard power button like the P6T, but not a reset switch, which you may or may not miss. Intel hasn’t bought into the solid-state capacitor craze that’s sweeping the high-end motherboard market, either, so you’ll find plenty of electrolytic caps dotting the board. That’s not necessarily problematic, but it seems a little low-rent given the DX58SO’s price tag.

There’s only one PCI slot on the DX58SO, but plenty of PCI Express. An x4 slot sits atop the slot stack, and not only is it notched to accept longer x8 and x16 cards, there’s also a retention clip to hold those cards in place. You also get a couple of x1 slots and a pair of x16s.

While the DX58SO supports CrossFire, Intel wasn’t able to reach an SLI licensing agreement with Nvidia. As a result, this may be the only enthusiast-oriented X58-based motherboard that won’t work with the green team’s GPU teaming scheme (that is, until someone hacks an SLI certification key into the BIOS).

You won’t find any throwback touches in the DX58SO’s port cluster, which is fitting considering the board’s lack of other legacy connectivity options. Intel does provide eight backplane USB ports, though, and they’re joined by a couple of eSATA ports, Firewire, and a Gigabit Ethernet jack. The single networking option is a little sparse for a high-end motherboard, but it does tap the ICH10R’s integrated GigE controller—something that other motherboard makers have shied away from in favor of auxiliary Gigabit chips from Marvell and Realtek.

A collection of audio ports fills out the right side of the port cluster, offering a host of analog inputs and outputs alongside a digital S/PDIF output. These ports hook into a Realtek ALC889 codec chip that supports Dolby Digital Live, allowing users to encode multi-channel audio from any application into a pristine digital bitstream that can be passed to compatible receivers or speakers.

A whole new set of BIOS options

The Core i7 brings with it all sorts of new dials for enthusiasts to tweak and a whole new learning curve when it comes to overclocking. Fortunately, Asus and Intel have both done a good job of exposing all the important variables in their respective BIOSes for the P6T and DX58SO.

P6T Deluxe DX58SO
Bus speeds Base clock: 100-500MHz in
1MHz increments

DRAM: 800, 1066, 1333, 1600, 1866, 2133MHz
PCIe: 100-200MHz in 1MHz
QPI: 4800, 5866, 6400MHz
Uncore: 2133-5600MHz in 133MHz

Base clock: 133-240MHz in
1MHz increments

QPI: 4.8, 5.866, 6.4GT/s

Bus multipliers NA DRAM: 6X-12X in 2X

Uncore: 16X-30X in 1X increments

CPU: 0.85-2.1V in
0.00625V increments

CPU PLL: 1.8-2.5V in 0.02V increments
QPI/DRAM: 1.2-1.9V in 0.00625V
IOH: 1.1-1.7V in 0.02V increments
IOH PCIe: 1.5-2.76V in
0.02V increments
ICH: 1.1-1.4V in 0.1V increments

ICH PCIe: 1.5-1.8V in
0.1V increments
DRAM: 1.5-2.46V in 0.02V increments

CPU static: 1.0-1.6V in
0.0125V increments

CPU dynamic: 6-494mV in 6mV increments
QPI: 1.15-1.8V in 0.025V
IOH: 1.1-1.5V in 0.025V increments
DRAM: 1.5-2.5V in 0.04V increments

Voltage, fan status, and
Voltage, fan status, and

Fan speed control
CPU, system NA

For most folks—those who don’t shell out for thousand-dollar Extreme Edition processors—Core i7 overclocking will involve cranking a motherboard’s base clock. The P6T offers a greater range of options on that front, although the 240MHz base clock ceiling on the DX58SO should be more than high enough for all but the most extreme overclockers.

Since the base clock is used to drive the processor core and its uncore component, which includes QPI and memory controller elements, there are plenty of other components that need to be kept in-check when pushing the base clock higher. Asus provides a range of clock speed controls for the memory bus, QPI link, and uncore, er, core. Intel takes a different tact, providing QPI clock control and a set of memory bus and uncore multipliers. Both boards achieve the same end here, just through different means. It is worth noting, though, that the Asus BIOS offers support for higher memory bus speeds than the Intel, at least with a Core i7-965 Extreme. However, both BIOSes run out of memory clock at 1066MHz when paired with a 940 or a 920, due to the slower memory controller frequency of those chips.

Voltage manipulation is often necessary to stabilize more ambitious overclocks, and the Asus and Intel boards offer similar peaks for the more commonly-tweaked voltages, giving users more than enough overvolting allowance with which to hang themselves. In fact, despite Intel’s warnings that memory bus voltages higher than 1.65V risk seriously damaging Core i7 CPUs, the P6T and DX58SO each offer memory bus voltage options up to 2.5V.

To Asus’ credit, the Deluxe does offer finer granularity in its overvolting controls. The P6T’s BIOS also gives users a few more obscure voltages to manipulate, although most folks will probably leave those alone.

What the Deluxe doesn’t have, however, is fine-grained control over the Core i7’s “Turbo mode,” which cranks up the multiplier on active cores when others are idling. The DX58SO lets you set specific multiplier limits based on how many cores are active at any given time, which is a neat capability to have.

As one might expect in this day and age, neither the P6T nor the DX58SO are short on memory timing options. Both boards offer command rate control, and they were quite comfortable running at 1T, even when overclocked.

With most BIOSes offering more than adequate overclocking and memory tweaking controls, we have to turn elsewhere to poke holes. In my mind, comprehensive temperature-based fan speed controls are just as essential to a good BIOS as robust overclocking options. Unfortunately, neither board does well here. The Deluxe fares better than the DX58SO, which offers temperature-based automatic fan speed control for its CPU fan, but no BIOS-level control over how fan speeds ramp. With the Asus board, you get fan speed control for the CPU and system fan headers, and the ability to switch between a couple of fan speed presets. However, there’s no way to set explicit temperature targets, base fan speeds, or how aggressively fans will spin up in response to changes in temperature.

Specifics on specifications

We’ve consolidated all the key specifications of the DX58SO and P6T Deluxe in a handy chart below.

Asus P6T Deluxe

Intel DX58SO

CPU support
LGA1366-based Core i7 processors LGA1366-based Core i7 processors

North bridge
Intel X58 Express Intel X58 Express

South bridge
Intel ICH10R Intel ICH10R

DMI (2GB/s) DMI (2GB/s)

Expansion slots
3 PCI Express x16

1 PCI Express x4
2 32-bit/33MHz PCI

2 PCI Express x16
Express x4
2 PCI Express x1

1 32-bit/33MHz PCI

6 240-pin DIMM

Maximum of 12GB of DDR3-1066/1333/1600 SDRAM

4 240-pin DIMM

Maximum of 8GB of DDR3-1066/1333/1600 SDRAM

Storage I/O
1 channel ATA/133 via Marvell
6 channels 300MB/s Serial ATA with RAID 0, 1, 10, 5 support

2 channels Serial Attached SCSI with RAID 0, 1 support via Marvell 88SE6320

6 channels 300MB/s Serial ATA with RAID 0, 1, 10, 5 support
Audio 8-channel HD audio via Analog
Devices AD2000B codec
8-channel HD audio via Realtek
ALC889 codec
Ports 1 PS/2 keyboard/mouse

2.0 with headers for 6 more

2 RJ45 10/100/1000 via Marvell 88E8056
1 1394a Firewire via
VIA VT6308 with header for 1 more
1 eSATA via Marvell 88SE6111

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 digital TOS-Link S/PDIF out
1 digital coaxial S/PDIF

2.0 with headers for 4 more

1 RJ45 10/100/1000
1 1394a Firewire via
VIA Texas Instruments TSB43AB22Awith header for 1 more
2 eSATA via Marvell 88SE6121

1 analog front out
1 analog bass/center out

1 analog rear out

1 analog line in

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

Our testing methods

As you thumb through the performance data on the following pages, keep in mind that the system setup for this motherboard comparison differs significantly from that of our Core i7 processor review. Not only am I using different BIOS revisions, memory speeds and timings, and system components, I’m also running a 32-bit rather than x64 version of Windows Vista. As a result, the benchmark scores in this review should not be compared to those in our Core i7 CPU coverage.

Since 1066MHz is the fastest memory bus speed Intel officially endorses for Core i7 processors, that’s what we’ve used for testing. We’ll be pushing the memory bus a little higher with some overclocking tests in a moment, though. We also kept the Core i7’s Turbo mode enabled for testing, although we did disable it when overclocking the boards.

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

Processor Intel Core i7-965 Extreme
System bus QPI 6.4GT/s (3.2GHz)
Motherboard Asus P6T Deluxe Intel DX58SO
Bios revision 0703 SO2624
North bridge Intel X58 Express Intel X58 Express
South bridge Intel ICH10R Intel ICH10R
Chipset drivers Chipset:
Memory size 3GB (3 DIMMs) 3GB (3 DIMMs)
Memory type Qimoda PC3-8500U-7-XX-A0
CAS latency
7 7

RAS to CAS delay (tRCD)
7 7
RAS precharge
7 7
Cycle time
20 20
Command rate 1T 1T
Audio codec Analog Devices AD2000B

Realtek ALC889 with

Nvidia GeForce 9800 GTX
with ForceWare 178.24 drivers

Hard drive

Western Digital Raptor WD1500ADFD 150GB


Windows Vista Ultimate x86
with Service Pack 1

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

Finally, we’d like to thank Western Digital for sending Raptor WD1500ADFD hard drives for our test rigs.

We used the following versions of our test applications:

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

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

Memory performance

With the Core i7’s memory controller integrated right into the processor, one might not expect much difference in memory subsystem performance between motherboards. However, as we’ve seen with Athlon 64, X2, and Phenom platforms, motherboard makers do have some freedom to tune on-die memory controllers, and that can have an impact on memory performance.

Sandra’s memory bandwidth test shows the P6T Deluxe out ahead of the Intel board by a hair. However, the DX58SO’s memory access latency is a couple of nanoseconds quicker than that of the Asus. Perhaps there’s something to Intel’s decision to only run four DIMM slots, after all.

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.

Euler3D apparently prefers lower access latencies to higher bandwidth. Here, the DX58SO is about 8% quicker than the Deluxe.


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

The P6T Deluxe bounces back nicely in WorldBench, where its overall score is five points higher than that of the DX58SO.

Although scores are close through WorldBench’s multimedia editing and encoding tests, the Asus board is quicker than the Intel in each test.

In WorldBench’s Firefox and multitasking tests, the P6T is again out ahead of the DX58SO. The performance gap is wider here, particularly in the Firefox test, which usually prefers systems with lower memory access latencies.

The Deluxe is notably quicker than the Intel board in WorldBench’s 3ds max DirectX modeling test, too.

However, there’s little difference in performance between the boards in WorldBench’s Nero and WinZip tests.


Graphics horsepower tends to gate gaming performance at the resolutions and detail levels most folks actually use to play. We’re using lower resolutions and relatively modest detail levels here in an attempt to tease out performance differences between motherboards.

Asus and Intel are in a dead heat in Crysis and Call of Duty 4, but the Deluxe delivers a healthy frame rate boost in Quake Wars and Episode Two. In the past, we’ve seen differences in audio codecs and their associated drivers have an impact on relatively low-resolution gaming performance, so that may be what’s happening here. Yes, 1024×768 counts as a low display resolution these days.

Curious to see whether the performance gap between the boards persisted with more realistic settings, I cranked Quake Wars and Episode Two up to 1920×1200 with 4X antialiasing and 16X anisotropic filtering, and maxed out all the in-game detail levels. These settings yielded frame rates in the 60 FPS range, with less than one frame per second separating the Asus and Intel boards in each game. As far as gaming in the real world goes, then, these boards look evenly matched.

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. Results that fall under “No power management” were obtained with Windows Vista running in high-performance mode, while those with power management enabled were taken with Vista in its balanced performance mode.

Switching Vista’s performance mode doesn’t have as much of an impact on idle power consumption as one might expect, but then the Core i7 still throttles its clock speed even when running in high-performance mode. That said, there’s more than a 10W difference in idle power consumption between the Asus and Intel boards, with the DX58SO proving more frugal. The gap doesn’t hold up under load, where the Intel board actually consumes three more watts than the Asus. I suspect the P6T’s additional power phases are responsible for the board’s higher idle power consumption, but they probably also deserve the credit for the Deluxe’s higher power efficiency under load.


Since few enthusiasts are likely to shell out for an Extreme Core i7 with an unlocked upper multiplier, we’ve limited our overclocking testing to pushing the base and memory clocks. To tackle the former, we dropped the CPU and memory multipliers to their lowest values, adjusting the uncore multiplier to double the memory multiplier (the Core i7’s uncore clock can’t be set at any lower than double the memory clock). We also lowered the speed of the QPI link to its lowest possible setting, 4.8GT/s, in an attempt to take it out of the equation. Then we started turning up the base clock, testing for stability with an eight-way Prime95 load along the way.

Our base clock quest began with the DX58SO, and we quickly discovered that while the board does offer a robust suite of overclocking options, some of them don’t actually work. The DX58SO simply refused to POST with its QPI link set to anything other than 6.4GT/s, preventing us from removing interconnect speed as a potentially limiting variable. We also had problems with the uncore clock, which needed to be set at more than twice the memory clock for the board to boot. The DX58SO didn’t recover gracefully when we pushed it too hard, either. Although it would occasionally boot us back into the BIOS after a failed overclocking attempt, much more often than not, it simply refused to post, necessitating a BIOS reset. This process is more involved than you might expect, requiring that you cut the power to the system, flip a jumper to put the BIOS in configuration mode, fire the system up and change BIOS settings accordingly, then power down again, move the BIOS jumper back to its original system, and power up once more. Ugh. At least the BIOS has a nifty recovery mode that can be used to resurrect the board after a failed flash attempt with little more than a BIOS file burned to a CD, but that didn’t make overclocking any less painful.

We were able to get the DX58SO up to a stable 160MHz base clock—a 37MHz boost above stock, and enough to take a Core i7-920 up to 3.2GHz. However, the DX58SO wouldn’t post at higher base clock speeds, not even with its chipset fan installed and additional voltage applied to various processor and chipset components. I suspect our inability to lower the PQI speed may have held the board back.

Next, we turned our attention to the P6T, which was much less finicky when pushed beyond stock speeds. The Deluxe had no problems running a 4.8GT/s PQI link, and it recovered nicely when pushed beyond the brink, automatically rebooting into the BIOS after a failed overclocking attempt.

The P6T effortlessly sailed up to a stable 200MHz base clock without so much as a voltage tweak. We managed to get it to post at 210MHz, too, but at that speed the system quickly hard-locked under our Prime95 stress test. Since a 200MHz base clock will take a Core i7-920 up to 4GHz, that’s probably plenty of overclocking headroom for most folks.

Given the Core i7’s penchant for faster memory, we swapped in a pair of 1600MHz OCZ DIMMs to see how far we could push the memory bus on each board. These DIMMs are rated for operation at 1.65V with 8-8-8-24 timings, but they didn’t get along with the DX58SO, which refused to post with them installed. The P6T, however, didn’t put up a fuss.

Unfortunately, we weren’t able to get the OCZ modules stable under our punishing eight-way Prime95 stress test at much faster than their default speed. We did get them stable at an effective 1644MHz, but any higher and Prime95 errors would quickly cascade across all eight instances. Raising memory timings and the DRAM command rate didn’t seem to help matters, and neither did bumping the processor, chipset, or memory controller voltages. Rather than risk damaging our Core i7 processor, we heeded Intel’s warnings and elected not to boost the memory bus voltage above 1.65V.

Motherboard peripheral performance

Some of the biggest performance differences you’ll find between motherboards come on the peripheral front, where it’s easy to spot where mobo makers have skimped on auxiliary peripheral chips.

Ethernet performance
Throughput (Mbps) CPU utilization (%)

Asus P6T Deluxe (1)
944 4.0

Asus P6T Deluxe (1)
944 4.1

Intel DX58SO
939 3.4

There’s little difference in networking performance between the P6T and DX58SO. Both boards offer excellent throughput with low CPU utilization, with the ICH10R’s integrated GigE controller looking every bit as good as the Marvell chips used on the Asus board.

HD Tach
Firewire performance

Read burst

speed (MB/s)

Average read

speed (MB/s)

Average write

speed (MB/s)

CPU utilization


Asus P6T Deluxe
33.2 30.7 15.9 0.3

Intel DX58SO
41.0 34.9 26.2 0.0

Firewire is another story, however. The Intel board’s Texas Instruments Firewire chip delivers much higher burst and write speeds than the VIA chip used on the P6T Deluxe.

HD Tach
USB performance

Read burst

speed (MB/s)

Average read

speed (MB/s)

Average write

speed (MB/s)

CPU utilization


Asus P6T Deluxe
32.9 32.5 28.6 0.3

Intel DX58SO
33.3 29.6 26.9 1.7

With both boards sharing the same ICH10R USB controller, it’s a little odd to see the DX58SO lagging a little behind here. It isn’t much slower than the Asus board, but our results were consistent across multiple test runs.

HD Tach
Serial ATA performance

Read burst

speed (MB/s)

Average read

speed (MB/s)

Average write

speed (MB/s)

Random access time (ms)

CPU utilization


Asus P6T Deluxe (ICH10R)
250.9 110.5 110.1 7.1 1.7

Asus P6T Deluxe (88SE6320)
182.3 110.5 81.1 7.1 1.3

Intel DX58SO
221.7 108.0 108.4 7.4 4.3

These tests were conducted with a Western Digital VelociRaptor, and the DX58SO’s SATA performance is a hint slower than that of P6T Deluxe, particularly in the burst speed test. The Deluxe’s SAS controller doesn’t look too bad here, matching the ICH10R’s sustained read speed, but coming up a little short in the burst and write speed tests.

RightMark Audio
Analyzer audio quality

Overall score

Frequency response

Noise level

Dynamic range


THD + Noise

IMD + Noise

Stereo Crosstalk

IMD at 10kHz

Asus P6T Deluxe
4 5 4 4 4 3 4 5 4

Intel DX58SO
5 5 5 5 5 3 5 5 5

We used RightMark Audio Analyzer’s 24-bit/192kHz loopback test to probe each board’s analog audio signal quality, and the DX58SO scored higher than the P6T nearly across the board.


Intel’s DX58SO is certainly well-equipped for enthusiasts. The question is whether Intel has built a better enthusiast board than Asus has with the P6T Deluxe—a tall order given Asus’ history of enthusiast favorites.

The DX58SO certainly looks good, and it has a better Firewire controller and audio codec than the Asus board. It’s more power-efficient, too, consuming ten watts less than the Deluxe at idle. In a bit of a surprise, the DX58SO also offers more BIOS-level control over the Core i7’s Turbo mode, which is a nice touch. But that’s about where the goodness ends for Smackover.

Intel may offer a robust set of tweaking options in the DX58SO’s BIOS, but some of them don’t work properly, and that likely hindered our ability to overclock the board. We found more base clock headroom in the P6T Deluxe, which was much easier to overclock overall. The Deluxe was a little quicker in some of our application tests, too, notably pulling ahead by five points in WorldBench. And it delivers much more on the features front, with two more DIMM slots, one more PCI Express x16 slot, SLI support, a Serial Attached SCSI controller, and one more Gigabit Ethernet port. Don’t forget Asus’ handy little OC Palm external display and control module, either.

The Deluxe isn’t perfect, but based on Asus’ suggested retail price of $309, the P6T is a much better value than the DX58SO. Intel expects the DX58SO to run $300-350, which is too much to pay for a board that effectively offers less on so many fronts. So the P6T is our favorite Core i7 motherboard, at least among those we’ve tested, and that’s good enough for TR Recommended distinction.

Of course, by the time you can actually buy a Core i7 processor, you’ll have more than these two motherboards to choose from. We’ve already taken delivery of another X58 board, and more are on the way, ensuring the P6T plenty of competition in the coming weeks. Stay tuned.

Comments closed
    • tegl1n
    • 11 years ago

    On the specifications table, for the P6T “Storage IO” you write “2 channels 300MB/s Serial ATA with RAID 0, 1 support via Marvell 88SE6320” This should read “2 x SAS ports supporting SAS RAID 0 and 1”.

    In response to mjw, 15K SAS drives offer a major improvement in throughput and access times. However, when a SATA drive is used on a SAS controller, some loss in performance is to be expected.

    • mjw
    • 11 years ago

    Given that the throughput of the SAS controller seems to be lower than the SATA controller, is there any benefit in using SAS (e.g. 15K rpm) disks with the ASUS? Would they just end up costing more than 10K rpm velociraptor and perform worse as well? Thanks for your opinion.

      • Prototyped
      • 11 years ago

      Better access times, higher I/O per second count.

    • UberGerbil
    • 11 years ago

    These are workstation (and, eventually, DP server) CPUs being sold initially as “enthusiast” chips, so it’s not surprising the motherboards are suitably expensive.

    I have to hand it to Intel, getting enthusiasts to pay top dollar for the privilege of doing the final validation pass on their workstation/small server hardware

      • derFunkenstein
      • 11 years ago

      Won’t workstation/server hardware use FB-DIMMs, or was that just for Core2-generation Xeons?

        • UberGerbil
        • 11 years ago

        Not on that socket they won’t. The Gainestown DP servers are slated to use the same socket as Bloomfield, and the same regular old DDR3. In fact Gainestown may even be the same silicon, just validated for DP use (since Bloomfield already has the necessary second QPI link).

        The Beckton MP Xeons (on a new 1567 socket) are supposed to use FB-DIMM, and quad-channel to boot. Yeah, that’ll be cheap 😉

          • Prototyped
          • 11 years ago

          LGA 1156 is much more interesting to me as a desktop user. In fact, I’m interested in the January 2010 Havendale dual-core processors; quad-core is worthless to me.

          One thing I’m really hoping for is that Intel won’t cripple the VT-x, EPT, TXT or SSE 4.x support on the LGA 1156 chips. Intel has a history of only leaving VT-x support enabled on their “premium” parts (during the Conroe days, on the E6000 and Q6000 series, leaving the mainstream E4000 series and the budget E2000, E1000 and 4×0 processors without them; and during the Penryn days, on the E8000 and Q9000 series, leaving the mainstream Q8000 and E7000 series and the budget E5000 series processors without them).

          Sadly I expect that Intel will continue this highly annoying product segmentation behavior with the more affordable Nehalem desktop processors.

            • UberGerbil
            • 11 years ago

            Well, 1156 is supposed to be more interesting to more people: it’s Intel’s “mainstream” product.

            Yeah, Intel does assign different techs to different segments and makes you pay extra for the better ones (at least you get a more capable chip along with it). And that segmentation looks to increase in the Nehalem generation, based on their “componentization” slides at the IDF. They keep adding more knobs to turn to generate different variants — now the number of RAM channels, soon the on-die GPU, in addition to the usual clockspeed and cache sizes.

            I do hope the realize that virtualization is becoming a common computing task for a lot of people; perhaps Apple will help them see that (or MS, if they put VirtualPC into the OS at some point).

            • Prototyped
            • 11 years ago

            If Intel would quit frigging crippling their mainstream processors (in terms of architecturally visible details such as instruction set components, rather than microarchitecture adjustments like leaving off a memory channel), my optimal desktop platform would have: an LGA 1156 Havendale processor; a “Kingscreek” G55 (presumably) PCH; and two channels of DDR3-1333 at 1.5 V. It’d be ideal for HTPC use with a presumed GMA X5500 IGP, and gamers could slap in a PCI Express 2.0 x16 video card to utilize the integrated PCI Express controller. (Integrated onto the GMCH die on the processor package; Havendale does /[

    • Prototyped
    • 11 years ago


    • Momento Mori
    • 11 years ago

    I know it may be a lot to ask, but could you possibly do a head to head of the X58 offerings from EVGA (FTW edition), ASUS (Rampage II Extreme) and Gigabyte (GA-EX58-UD5)? Prices would also be greatly appreciated when going through the gamut of X58 boards. So far great coverage. 😀

      • crazybus
      • 11 years ago

      Gambit: this word does not mean what you think it means.

      Gamut perhaps? 😉

      Personally, I’m waiting for the EVGA Epic Win edition.

        • Momento Mori
        • 11 years ago

        Yes, you or correct. gamut is the word I am looking for, it would not make sense to go through a gambit of boards.

    • ybf
    • 11 years ago

    If I’m paying a kilobuck for the mill, I’m not complaining about the markup on the mobo. If an extra hundy means it won’t melt when I’m buzzing the memory at 1600 MHz, it’s worth it.

    • Voldenuit
    • 11 years ago

    It’s ironic that X58 mobos cost so much considering that the “northbridge” on the motherboard is essentially just a glorified electric router -_-.

    The intel DX58SO is a no show for me (surprise surprise), with overclocking features that don’t work, and zero effective headroom (not to mention lack of SLI support).

    The ASUS P6T is a more interesting picture, but still critically flawed in my book. The superfluous heatpipe arrangement eats up a whole expansion slot (have you seen how *thick* those fins are? They’re more ornamental than utilitarian), and the PCIEx16 slot placement leaves a lot to be desired (as mentioned in the review). The layout on the RoG board is better, but at its ludicrously inflated pricepoint, is a no-go.

    130nm? Multi chip solutions? Blegh.

    I sincerely hope nvidia gets its QPI license sorted out so we can see some real competition and innovation in the LGA-1366 space. They’re really been on a roll with their chipsets lately, even if this means precluding CF (which, as a 4850 owner, makes me sad).

    • flip-mode
    • 11 years ago

    Those are high prices. Two thumbs up for SAS on the Asus.

    I can’t see spending much more than $100 for a mobo these days. The $100 mobos get so much right, and still leave you with money in your pocket to spend on other things. But there are lots of people out there that have either more money or fewer expenses than I do, so these boards’ prices might be a-o-k for them.

    • AMDisDEC
    • 11 years ago

    Ahh..soon we will have the dual socket motherboard releases. Than I can finally upgrade my Opteron system. Shweet!

      • SubSeven
      • 11 years ago

      Funny, the very guy that has been spewing doom prophecies of AMD for the last God knows how many years, is actually running an AMD based platform? Someone correct me if I’m wrong, but doesn’t this fall in the hypocrite bin?

        • AMDisDEC
        • 11 years ago

        Explain please?

          • SubSeven
          • 11 years ago

          Gladly. Which part is hard to understand?

            • AMDisDEC
            • 11 years ago

            Barack Obama says, buy this part and an open OS.
            I endorse this message

      • DrDillyBar
      • 11 years ago

      Dual all the way 😉

      • Prototyped
      • 11 years ago

      Circa February or March 2009 for Nehalem-EP and the corresponding motherboards, from what I’ve heard.

    • Pax-UX
    • 11 years ago

    Looks promising, to bad this stuff is crazy money at the minute. Hopefully there’ll be some value to be had early next year.

    • dalisam
    • 11 years ago

    What no AGP slot? I see plenty of PCI slots on these motherboards. I thought it would be all PCI-E at this point.

    I mean from what I have seen from all the X58 mobos coming out is if you have 2 video cards with the blower fan type cooler 1x PCI-E NIC and a 1x PCI-E sound card ,you’re screwed.

      • nexxcat
      • 11 years ago

      With integrated NICs being so good, why would you have a PCIe NIC?

    • MadManOriginal
    • 11 years ago

    Thanks for this, it’s a real-world test that is nice to see:


    • Neutronbeam
    • 11 years ago

    I want an ASUS high-end shootout — Compare the *[

    • ilkhan
    • 11 years ago

    The boards cost as much as a quality DDR3 X48 board, plus a little for being new. No surprise, we expected $250-350 to be the stree prices for X58 boards.

    Also, I want to see the P6T compared with the X58-DS4 board, since those are the two Im likely to seriously consider on launch day.

    • shank15217
    • 11 years ago

    About 15-20% performance gain across the board with some very big outliers.

    • Nitrodist
    • 11 years ago

    Word to the wise, you can compare some scores from across different reviews on TR to see a preview of Core i7’s performance.

    • leor
    • 11 years ago

    holy crap!

    these mobos cost as much as server boards!

      • Prototyped
      • 11 years ago

      Or, er, enthusiast boards with the X48 chipset. The X in X58 should give away the fact that it’s meant for eXtr33m p30p1e.

    • Nitrodist
    • 11 years ago

    So, you can review the motherboards with the Core i7 chips in them BUT you can’t review the Core i7 chips yet?

      • lycium
      • 11 years ago

      yeah they even refer to a review of the i7:

      “As you thumb through the performance data on the following pages, keep in mind that the system setup for this motherboard comparison differs significantly from that of our Core i7 processor review”

        • moloch
        • 11 years ago

        check now 😉

Pin It on Pinterest

Share This