review intels p55 express chipset

Intel’s P55 Express chipset

For years, Intel’s core-logic chipsets have set the standard against which all others are judged—as long as we’re not talking about integrated graphics performance, anyway. You’d have to dig all the way back into the pre-Prescott era for an Intel chipset with serious issues. Finding flaw with relatively recent core-logic offerings from AMD and Nvidia is considerably easier, however. During Intel’s reign, AMD’s south bridge chips have struggled with all sorts of issues, from slow USB and PCI performance to generally flakey AHCI implementations. Nvidia’s nForce designs have had their share of problems, too, like bouts with data corruption and generally higher power consumption.

A relative lack of quirky issues has made Intel chipsets particularly popular among enthusiasts. The mid-range models have been the most iconic, not just because they’ve offered all the performance and stability of higher-end variants, but because they’ve generally taken well to overclocking. Such was the case with the P45 Express we’ve been enjoying for more than a year now, the P35 that came before it, the P965 before that, and so on.

Today, Intel’s latest mid-range chipset makes its debut alongside Lynnfield-based Core i5 and i7 processors. Ladies and gentlemen, I present you the P55 Express.

Yep, that’s it: one chip. It’s not really a chipset, is it? Intel calls this slice of silicon the P55 Express Platform Controller Hub, or PCH. The chip is fabbed using 65-nano process technology, and it’s tiny. According to my ruler, the PCH measures just 8.5 x 9 mm for a total die area of around 76.5 mm².

Intel can get away with such a small, single chip because Lynnfield moves much of the chipset’s traditional duties onto the CPU. Like its Bloomfield-based Core i7 predecessors, Lynnfield CPUs have integrated memory controllers. You’re limited to two DDR3 memory channels, but speeds are supported up to an effective 1333MHz. This latest Nehalem iteration also brings a sweet 16 lanes of second-generation PCI Express connectivity onto the processor package, consolidating CPU and north bridge components under one roof.

Lynnfield’s PCI Express lanes can be configured as a single x16 link or split between a pair of x8s for multi-GPU configurations. CrossFire support is universal, but as with Intel’s X58 chipset, SLI certification is handled at the motherboard level. All boards should be able to combine a single GeForce graphics card with a second one for dedicated PhysX acceleration, provided they have a second physical x16 slot. You can also add a dedicated PhysX card to an SLI-certified board running a pair of GeForces in tandem. If that wasn’t excessive enough, some uber-high-end P55 boards will also feature nForce 200 chips and support for three-way SLI. Such extravagance seems silly given the P55’s mid-range aspirations, but that’s how mobo makers roll these days.

Lynnfield has enough PCI Express lanes for most graphics configs. What about expansion slots and peripherals, though? That’s where the P55 Express PCH comes in. It’s equipped with eight gen-two PCIe lanes—two more than the old ICH10R south bridge. However, these lanes only offer signaling rates up to 2.5GT/s, which is the same speed as gen-one PCI Express. That’s unlikely to be a major impediment in mid-range systems, but it will limit the bandwidth available to high-end RAID cards, and the like.

Little has changed in the Serial ATA department, however. The P55’s six-channel SATA controller is largely the same as the one in the ICH10R, so it’s loaded with RAID functionality but lacking support for next-gen 6Gbps transfer rates. Intel says it’s looking at the new standard, but that only SSDs are likely to benefit from greater interface bandwidth. Not even today’s fastest SSDs can saturate a 3Gbps SATA link, anyway.

On the networking front, the P55 has a Gigabit Ethernet MAC that will consume one of the board’s PCI Express lanes if tapped. So much for saving an expansion slot with integrated GigE. Intel has made some changes in the P55’s USB component, as well, retaining the dual-controller design while adding two ports for a total of 14. Throw in an HD audio interface, and that just about does it for the P55.

The PCH is linked to the CPU via a 2GB/s DMI interconnect similar to the one Intel used to join the north and south bridge components of its last several chipsets. I suppose that makes sense given that this is still an I/O hub link, but the pipe doesn’t look nearly fat enough on paper. The DMI connection’s 2GB/s of peak bandwidth is equivalent to just four of the PCH’s half-bandwidth PCIe 2.0 lanes. And the P55 PCH has another four of those, plus SATA, USB, and other lesser chipset functions to clog things up. The average mid-range PC is only likely to have a few hard drives and USB peripherals, though, and it probably won’t be maxing too many of those PCIe lanes at once.

I have to admire the elegant simplicity of the P55 PCH, but there isn’t much to the actual chip. The same can’t be said about the first wave of P55-based motherboards to hit the Benchmarking Sweatshop. What the PCH lacks in flash, it makes up for by packing a heck of a lot of connectivity into a very small package, allowing motherboard makers to load up on expansion slots and peripherals to complement their ever-evolving arsenals of overclocking options and useful new features.

That ordnance is on full display in the first wave of P55 boards from the big three: Asus’ P7P55D Deluxe, Gigabyte’s GA-P55-UD6, and MSI’s P55-GD65. Rather conveniently, we’ve arranged an in-depth throwdown to see how they compare. We’ll dive into the results of our exhaustive suite of chipset and motherboard tests in a moment, but first, let’s properly introduce these Lynnfield-ready mobos.

Asus’ P7P55D Deluxe motherboard
Like, Xtreme Design, dude

Manufacturer Asus
Model P7P55D Deluxe
Price (Street)
Availability Now

As I sit here writing this article ahead of Lynnfield’s official unveiling, I can’t help but be amused that Asus’ P7P55D Deluxe is already for sale—and listed as in stock—at several online retailers. The board is currently running around $225, which is quite expensive given the ostensibly mid-range chipset. But then this is also a Deluxe, and just one of a whole collection of P55-based motherboards coming from Asus. Among those are a vanilla version of the P7P55D, an even more exotic flavor, and some love for the microATX crowd.

Today, the Deluxe is in the spotlight, and it looks very much the part. Users will only see the board briefly unless they’re running case windows, which is a shame, because it’s quite a looker. Asus has embraced the more muted tones that have come to typify modern motherboard design, and there’s a sophistication to the Deluxe’s palette.

The Deluxe looks surprisingly bare, which I suspect has a lot to do with the lack of a discrete north bridge chip hogging the middle of the landscape. Don’t be deceived by appearances, though. Asus has packed quite a lot onto this board, and into it. There are eight layers in total, with a full two ounces of copper between them, as is de rigueur on enthusiast-oriented motherboards these days. You’ll find fancy electrical components throughout, as well. In case you were wondering, some of these attributes fall under a new version of StackCool, which itself is a part of Asus’ new Xtreme Design marketing initiative. You know, because nothing says fresh like yet another example of Xtreme branding.

But I digress.

Many of the Deluxe’s electrical components can be found ringing the new Lynnfield LGA1156 socket. The board employs a 16+3 power phase design that uses a so-called T.Probe microcontroller to balance loads across available phases. Asus says spreading the load can help to lower motherboard temperatures and deliver more stable power to the CPU. The Deluxe also scales its power phases dynamically with the system load to reduce power consumption.

Asus covers some of the Deluxe’s voltage regulation circuitry with abstractly ornate heatsinks that do a good job of staying out of the way. This leaves plenty of clearance around the socket for larger coolers, although you might run into problems with taller aftermarket DIMMs. Speaking of memory, the Deluxe has a MemOK! feature that starts up RAM in stages, allowing you to change the timings, voltage, and frequency at POST if a module’s SPD settings are incorrect. That’s a neat feature if you’re struggling with off-brand DIMMs, but I have to wonder how many of those make it into high-end motherboards like this one.

The DIMM slots themselves add a new twist: tabless retention on one side. The missing tabs allow modules to slide in easily, even when longer graphics cards butt up against the slots.

Just below the DIMM slots sits a low-profile heatsink that has more square footage than my first dorm room. The size seems to be more about providing a big billboard for graphics than expanding the heatsink’s surface area. If Asus was really serious about lowering chipset temperatures, it probably would have left the heatsink’s underlying fins exposed to more airflow.

Of course, the chipset cooler’s footprint matters little, because it’s otherwise unobtrusive. The heatsink won’t interfere with longer graphics cards, and neither will the edge-mounted SATA ports hooked up to the P55. One of the three auxiliary SATA ports powered by various onboard JMicron controllers will be blocked by a third double-wide graphics card, though.

The gaping hole left by the P55’s assimilated north bridge component leaves room for a total of seven expansion slots. Only the top two are hooked up to the Lynnfield processor, but the board is SLI-certified for two-way configs. The third x16 slot gets four lanes of connectivity from the PCH.

We also find new retention tabs on the x16 slots. These are supposed to be easier to reach in tightly packed multi-GPU configurations, but the Deluxe’s excellent slot spacing is actually a more effective solution on that front. I suppose the new tabs would come in handy if you were running a third graphics card alongside a double-wide multi-GPU duo.

Asus tucks the usual power and reset buttons just under the slot stack. You won’t find a POST code display, though. Instead, the Deluxe has LEDs located throughout the board that light up next to components causing POST problems.

There are a couple of surprises to be found in the port cluster, which features a handy CMOS reset switch in just about the best place you could put it. Digital S/PDIF output is offered in two flavors, both driven by a Via VT2020 audio codec. The VT2020 supports Blu-ray playback, which is nice, but it can’t encode DTS or Dolby Digital Live bitstreams on the fly.

External Serial ATA connectivity is conspicuously absent from the port cluster. However, the box contains a PCI back plate adapter that can link any internal SATA port with the outside world.

Instead, Asus throws in a little overclocking remote for those who are somehow unsatisfied with BIOS- and Windows-based overclocking controls. I’d rather have eSATA, thanks.

I’d also like to have the Express Gate instant-on OS back on an onboard flash chip. With the P7P55D Deluxe, Express Gate has to be installed on a hard drive, which kills some of its appeal. The OS itself is largely unchanged, too; Express Gate still lacks the sort of stability testing and recovery applications that would make it really useful to enthusiasts, overclockers, and even the average PC user.

Fortunately, the P7P55D Deluxe’s BIOS is as good as you’d expect from a high-end Asus motherboard. The layout is excellent, and you can manually key in most voltages, timings, and clock speeds. Only the most extreme overclockers are likely to exceed the Deluxe’s voltage ceilings, and even then, you can flip onboard switches to inject a few extra millivolts into the CPU, its integrated memory controller, and the DIMM slots. Heck, it’s even possible to toggle between defining a target voltage or specifying a desired increment above the default.

Unlike the other boards in this round-up, the Deluxe doesn’t offer per-channel memory timing controls, just one set for both. The board’s fan speed options are also painfully thin: three arbitrary presets for the CPU and system fan header with no ability to further tweak temperature triggers or fan behavior.

Gigabyte’s GA-P55-UD6 motherboard
Two channels, six slots

Manufacturer Gigabyte
Model GA-P55-UD6
Price (MSRP) $250
Availability Now

The GA-P55-UD6 tops Gigabyte’s P55 lineup, and with a suggested retail price of $250, it had better. This makes the UD6 the priciest of the three boards we’re looking at today, although it’s not substantially more expensive than the P7P55D Deluxe. Like Asus, Gigabyte also has a varied array of cheaper P55 models, including more affordable SLI-certified variants, one of which has a microATX form factor. We’ll be getting to those boards soon, reserving our focus today for the UD6.

Gigabyte has made its trademark turquoisey-blue work with the more restrained motherboard aesthetics that have taken hold as of late. The UD6 certainly looks sharp, even if the racing stripes are a little odd. I’m usually a sucker for racing stripes, but these ones don’t really match or line up with each other.

A more traditional heatsink layout makes the UD6 look like it’s hiding typical north and south bridge chipset components. The P55 PCH is actually in the middle of the board where you might otherwise expect to find a north bridge chip. What looks to be a south bridge cooler is just covering a couple of auxiliary storage controllers. Gigabyte says this extra heatsink is there primarily to draw heat away from the chipset; the SATA chips don’t produce much heat on their own.

The UD6’s extensive array of heatpipe-linked coolers certainly contributes to the board’s crowded look. Also crowding the layout is an extra pair of DIMM slots, making a total of six, just like most X58-based mobos. You’re still limited to two channels and a maximum of 16GB of memory, but the extra slots at least make it possible to hit 12GB using relatively affordable 2GB sticks.

Like the other boards we’re looking at today, the UD6 uses high-quality MOSFETs, chokes, and solid-state capacitors. As a member of Gigabyte’s Ultra Durable 3 family, its circuit board is also laden with two ounces of copper. This is an eight-layer design, as well.

Gigabyte corners the CPU socket with a whopping 24 power phases—more than either of the other boards. The UD6 is smart enough to dynamically scale the number of power phases based on system load, too. However, the sheer volume of power regulation circuitry, combined with the close proximity of the DIMM slots and chipset cooler, makes for a more cramped socket than the other boards.

Moving south brings us to a total of ten edge-mounted Serial ATA ports. Six are supplied by the P55, while the other four are tied to auxiliary GSATA and JMicron silicon. Gigabyte had originally intended to equip the UD6 with a 6Gbps SATA controller from Marvell, but problems with the chip forced its replacement.

Obviously, these edge-facing ports won’t interfere with longer graphics cards. However, if your case’s hard drive cage happens to snug up right next to the motherboard tray, you might have to get creative with SATA cabling.

In case you’re wondering, that’s a CMOS reset button near the front panel connector block. I’d rather see this button in the port cluster where it can be accessed without popping case panels, but at least you won’t have to fiddle with a jumper.

The UD6 has the same slot complement as the Asus board, but Gigabyte stacks things in a different order. A potential conflict between DIMM slot retention tabs and longer graphics cards is avoided by putting the top x16 slot one down the line. However, the top x1 slot doesn’t have much clearance for longer expansion cards. The proximity of the chipset cooler even creates problems for my tiny x1 Gigabit Ethernet card.

Fortunately, the rest of the UD6’s slots are spaced intelligently. The board doesn’t feature an nForce 200 chip, but you do get a third x16 slot. This slot shares bandwidth with the x1s and the Serial ATA controller used to drive the eSATA ports, so you only get the full four lanes of bandwidth if you’re willing to give up the other stuff. Those who’d rather have all of the UD6’s x1s and peripherals active can switch the x4 to single-lane mode in the BIOS.

I’m not sure Gigabyte could’ve squeezed more into the UD6’s port cluster if it tried. There’s plenty of everything, including dual FireWire ports and a pair of digital S/PDIF outputs. More importantly, those ports are driven by Realtek’s ALC889A audio codec, the only one capable of encoding multi-channel digital bitstreams on the fly. Real-time encoding isn’t necessary for movie playback, but it’s the only way to pipe multi-channel game audio through a digital output.

The port cluster has a whopping eight standard USB ports, plus two more in hybrid eSATA/USB ports. New eSATA devices are apparently on the way, and they’ll be able to pull power from these ports, eliminating the need for inconvenient wall warts or auxiliary power cables.

Like most Gigabyte boards, the UD6 features dual BIOS chips. One serves as a handy backup, and the capacity of the chips themselves has been increased with an eye towards eventually storing crash information for troubleshooting purposes. For now, you can use the extra capacity to save passwords and “important dates” if you want those stored on your motherboard’s BIOS chip for some reason.

The BIOS itself is loaded with overclocking and memory-tuning features, including per-channel timing controls. Most high-end motherboard BIOSes let you arbitrarily key in values for clock speeds, voltages, and timings, and so does this one.

The UD6’s fan speed controls are sorely lacking, however. There’s no option for temperature-based control for the system fan header, and the CPU fan speed options are even more limited than on the Asus board.

Oh, the tinfoil hat crowd should probably be aware that the UD6 sports a TPM chip. Worry not, though; it can be disabled in the BIOS. The chip’s encryption key can be housed on a USB storage device or even a cell phone, should you wish to tie access to your data to either device.

MSI’s P55-GD65 motherboard
An affordable alternative

Manufacturer MSI
Model P55-GD65
Price (Street) $160
Availability Now

It seems that every time we cover a mainstream chipset launch, motherboard makers send us their most expensive examples of the breed. That’s the best way to show off the latest and greatest in mobo whiz-bangery, I suppose, but it doesn’t best represent the sort of mid-range motherboards most enthusiasts actually buy. Obviously, the Asus and Gigabyte boards we’ve looked at today occupy the opulent end of the P55 motherboard spectrum, and MSI has a direct competitor for them in the P55-GD80. We’re not looking at that board today, however. Instead, we have the $160 P55-GD65, which should give us a sense of what a more affordable P55 board has to offer.

At first glance, the GD65 definitely looks a little plain. MSI has gone with a simple color scheme, combining black, blue, and pewter tones on a board that looks more sparsely populated than the others.

The open layout nicely avoids troublesome clearance conflicts, which is easier to do with only two x16 slots and generally fewer onboard peripheral chips. That said, the GD65 is hardly lacking in expansion capacity or onboard connectivity options.

Despite its lower price tag, the GD65 still features fancy electrical components, including DrMOS MOSFETs that are reputedly 400% faster than traditional designs. Unlike Asus and Gigabyte, however, MSI doesn’t appear to be injecting boards with extra copper. That omission isn’t just an artifact of the GD65’s lower price tag; numerous less expensive Gigabyte boards have extra copper between their PCB layers.

MSI has also declined to compete in the power phase arms race that has Asus and Gigabyte constantly one-upping each other. The GD65 has a mere six power phases feeding the processor, and MSI assures us that’s plenty. Even the high-end GD80 model only has eight phases. Both models support dynamic power-phase switching complete with onboard LEDs that let you know how many phases are active at any given time. MSI says its design is quite efficient, too, delivering power to the CPU with 92% efficiency. We’ll see how the board’s actual power consumption measures up in a moment.

With fewer power phases ringing the socket and only modest voltage circuitry cooling, the GD65 should be able to easily accommodate larger aftermarket coolers. Keep in mind that the LGA1156 socket uses a new hole pattern for heatsink retention brackets. Your old LGA775 cooler won’t work on this or the other boards we have under the microscope.

Like Asus, MSI puts the P55 PCH where one might expect to find south bridge silicon. A small heatsink keeps the chip cool, but it’s a low-profile design that won’t interfere with longer graphics cards. Longer cards won’t brush up against the DIMM slot retention tabs or block access to any of the board’s Serial ATA ports, either. Of course, putting the SATA ports right on the edge of the board can complicate cabling in particularly tight enclosures. You win some, you lose some.

The blue Serial ATA port sitting just behind the edge-mounted cluster is predictably tied to a JMicron controller that also powers the IDE port. All three of the boards we’ve explored feature IDE connectivity, which is somewhat surprising in this day an age. IDE functionality was dropped from Intel chipsets years ago, and SATA optical drives have since become widely available with little associated price premium.

Rather than equipping the GD65 with a third PCI Express x16 slot, MSI opted for an open-ended x4 with enough clearance to accept standard graphics cards. The slot spacing won’t allow for a three-way double-wide config, though. Given that the GD65 is only certified for two-way SLI, that’s probably not a big deal.

The P55’s eight lanes of PCIe 2.0 connectivity are just enough to cover the GD65’s array of expansion slots and peripherals, so there’s no need to switch around lane configurations to keep all slots running at full steam. I’ll get to the collection of buttons at the bottom of the slot stack when we probe the GD65’s overclocking performance a little later in the review.

While not as loaded as the UD6’s port cluster, the GD65’s has all the right boxes checked: FireWire, dual S/PDIF, hybrid eSATA/USB, and gobs of USB. Realtek’s ALC889 codec sits behind the audio outputs, but MSI hasn’t sprung for the SoundStormesque Dolby Digital Live encoding option. That’s a shame, because with real-time encoding on the motherboard and a compatible digital receiver, there’s little need for a discrete sound card.

The GD65 isn’t equipped with much in the way of additional accessories, but MSI has put a little something extra on the board in the form of probe points for voltage monitoring. Your average enthusiast probably won’t bust out a multimeter to monitor CPU, memory, or chipset voltages. However, I can see serious overclockers appreciating the ability to tap those lines.

Overclockers will also find plenty of options in the GD65’s BIOS. Memory timings can be adjusted on a per-channel basis, too. The three boards we’ve rounded up today have some minor differences in their performance-tuning options, but they all offer more than enough range and granularity for even picky enthusiasts. Asus and Gigabyte are often lauded for having excellent BIOSes, but based on the P55-GD65 and other recent models, I’d say MSI’s are now every bit as good.

In fact, MSI’s BIOS-level fan speed control options are better than what you get on either of the other boards. Users can define a target CPU fan speed between 40 and 70°C and set a minimum speed between zero and 87.5% in 12.5% increments. The two system fan headers can only be set to run at 50, 75, or 100% of full speed, so they’re not controlled by system temperature changes.

The devil’s in the details
This wouldn’t be TR motherboard coverage without a painstakingly detailed assessment of each board’s BIOS options and specifications. These details don’t exactly lend themselves to eloquent prose, but you should be able to find what you need in the tables below.

Asus P7P55D Deluxe Gigabyte GA-P55-UD6 MSI P55-GD65
Clock speeds Base: 80-500MHz in 1MHz
DRAM: 800,1066,1333MHz

100-200MHz in 1MHz increments
QPI: 4270, 4800MHz
Base: 100-120MHz in 1MHz
90-150MHz in 1MHz increments
Base: 100-600MHz in 1MHz
90-190MHz in 1MHz increments
Multipliers CPU: 9X-21X in 1X increments CPU: 9X-20X in 1X increments
DRAM 6,8,10X
QPI: X23, X36
CPU: 9X-21X in 1X increments
DRAM 3,4,5X
QPI: 16,18X
Voltages CPU: 0.85-1.7V or
1.25-2.1V in 0.00625V increments
CPU PLL: 1.8-2.1V in 0.1V increments

IMC: 1.1-1.9V in 0.00625V

: 1.5-2.55V in
0.0125V increments

: 1.05-1.15V in
0.1V increments
DRAM data A/B: 0.395-0.630X in 0.005X increments
DRAM address A/B: 0.395-0.630X in 0.005X increments
CPU: 0.5-1.9V in 0.00625V increments
CPU PLL: 1.5-2.12V in 0.02V increments

QPI/VTT: 0.8-1.94V in 0.02V

: 1.3-2.6V in
0.02V increments

: 0.85-2.08V in
0.02V increments
DRAM data A/B: 0.53-1.18V in 0.01V increments
address A/B: 0.72-0.98V in 0.01V increments
CPU: 0.87-2.07V in
0.005-0.006V increments
CPU VTT: 0.451-2.018V in 0.005-0.006V increments

0.906-2.405V in
0.006-0.007V increments

PCH 1.8
: 1.0-2.4V in 0.01V increments
PCH 1.05
0.451-1.953V in 0.005-0.006V increments

DRAM data A/B:
0.825-1.225V in 0.025V increments
DRAM address A/B: 0.825-1.225V in
0.025V increments
Fan speed control CPU, system CPU CPU, system
Asus P7P55D Deluxe Gigabyte GA-P55-UD6 MSI P55-GD65
CPU support LGA1156-based
Core i5, i7 series processors
Core i5, i7 series processors
Core i5, i7 series processors
Chipset Intel P55 Express Intel P55 Express Intel P55 Express
Interconnect DMI (2GB/s) DMI (2GB/s) DMI (2GB/s)
Expansion slots 3 PCI Express x16
PCI Express x1
2 32-bit/33MHz PCI
3 PCI Express x16
PCI Express x1
2 32-bit/33MHz PCI
2 PCI Express x16
1 PCI Express x4
PCI Express x1
2 32-bit/33MHz PCI
Memory 4
240-pin DIMM sockets
Maximum of 16GB of DDR3-800/1066/1333 SDRAM
240-pin DIMM sockets
Maximum of 16GB of DDR3-800/1066/1333 SDRAM
240-pin DIMM sockets
Maximum of 16GB of DDR3-800/1066/1333 SDRAM
Storage I/O Floppy disk
1 channel ATA/133 via JMicron JMB363

6 channels 300MB/s Serial ATA with RAID 0, 1, 10, 5 support
channels 300MB/s Serial ATA via JMicron JMB322 with RAID 0, 1 support
channel 300MB/s Serial ATA via JMicron JMB363
Floppy disk
1 channel ATA/133 via

Gigabyte GSATA2
6 channels 300MB/s Serial ATA with RAID 0, 1, 10, 5 support
2 channels
300MB/s Serial ATA via Gigabyte GSATA2 with RAID 0, 1 support
2 channel
300MB/s Serial ATA via JMicron JMB362
Floppy disk
1 channel ATA/133 via JMicron JMB363

6 channels 300MB/s Serial ATA with RAID 0, 1, 10, 5 support
channel 300MB/s Serial ATA via JMicron JMB363
Audio 8-channel HD audio via VIA
VT2020 codec
8-channel HD audio via Realtek
ALC889A codec
8-channel HD audio via Realtek
ALC889 codec
Ports 1 PS/2 keyboard
1 PS/2 mouse
2.0 with headers for 6 more

1 RJ45 10/100/1000 via Realtek RTL8111L

1 RJ45 10/100/1000
via Realtek RTL8110

1 1394a FireWire via
VIA VT6308P with header for 1 more

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 S/PDIF out (TOS-Link)

digital S/PDIF out (RCA)

1 PS/2 keyboard/mouse
2.0 with headers for 4 more

2 RJ45 10/100/1000 via 2
Realtek RTL8111D

2 hybrid eSATA/USB via JMicron JMB362 with RAID 0, 1 support
2 1394a FireWire via
Texas Instruments TSB43AB23 with header for 1 more

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 S/PDIF out (TOS-Link)

digital S/PDIF out (RCA)

1 PS/2 keyboard
1 PS/2 mouse
2.0 with headers for 6 more

2 RJ45 10/100/1000 via 2 Realtek RTL8111DL
1 hybrid eSATA/USB via JMicron JMB363
1 1394a FireWire via
VIA VT6315N with header for 1 more

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 S/PDIF out (TOS-Link)

digital S/PDIF out (RCA)

Our testing methods
The P55 Express is the only chipset compatible with Intel’s new Lynnfield CPUs, so it’s essentially without direct competition. Our closest bet is the other Nehalem platform: the X58 Express. Since these two chipsets can’t share a common CPU, we’ll be looking at somewhat of a lopsided comparison. The X58 gets an eight-thread Core i7-920 running at 2.67GHz while the P55 must make do with a Core i5-750 that has the same clock speed as the 920, but lacks Hyper-Threading, and is thus limited to four threads.

Of course, we’re primarily focusing on chipset and motherboard performance today. Our X58 system’s more powerful CPU won’t affect those too terribly.

I should point out that our P55 rig will have a slight advantage over the X58 in one area: memory speed. Our X58 platform sports an engineering sample Core i7 CPU whose memory clock is capped at 1066MHz. Lynnfield CPUs can crank memory up to 1333MHz, and since OCZ sent over a pair of DDR3 DIMMs that are quite happy at that speed with tight 7-7-7-20-1T timings, we couldn’t resist running the P55 a little faster. Besides, the X58 platform’s extra memory channel more than makes up for the MHz difference. Incidentally, the OCZ modules are rated for operation at up to 1866MHz on just 1.65V; hitting that speed with a Lynnfield CPU requires overclocking the processor’s base clock, though.

Each motherboard was run through a full suite of memory, power, peripheral performance, and overclocking tests. I also put the X58 platform and the Gigabyte P55 board through a second wave of chipset-specific application, gaming, and peripheral performance tests.

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

Processor Intel Core i5-750 2.67GHz

Intel Core i7-920
ES 2.67GHz
CPU/chipset link DMI QPI

Asus P7P55D Deluxe Gigabyte GA-P55-UD6 MSI P55-GD65

Gigabyte GA-EX58-UD5
Bios revision 0504 F3 100 F3

North bridge
Intel P55 Express Intel P55 Express Intel P55 Express Intel X58 Express

South bridge
Intel ICH10R
Chipset drivers Chipset:
Memory size 4GB (2 DIMMs) 4GB (2 DIMMs) 4GB (2 DIMMs) 6GB
(3 DIMMs)

Memory type






DDR3 SDRAM at 1066MHz
CAS latency (CL) 7 7 7 7
RAS to CAS delay (tRCD) 7 7 7 7
RAS precharge (tRP) 7 7 7 7
Cycle time (tRAS) 20 20 20 20
Command rate 1T 1T 1T 1T

Audio codec
VIA VT2020 with drivers Realtek ALC889A with 2.31 drivers Realtek ALC889 with 2.31 drivers Realtek ALC889A with 2.31

Nvidia GeForce GTX 260 896MB
with ForceWare 190.62 drivers
Hard drive
Western Digital Raptor X 150GB
OS Windows 7 Ultimate RTM

Our test system was powered by a PC Power & Cooling Silencer 750W power supply unit.

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 Intel moving the memory controller into its Nehalem-based CPUs, motherboards and chipsets have less influence on memory performance. Still, this is a good place to start. I ran Stream with four threads on the P55 motherboards and eight on the X58. Because these systems were configured with tighter memory timings than the ones Scott used in his Lynnfield CPU review, the results aren’t directly comparable.

The X58 board has a predictable edge in our memory bandwidth test, even with the Lynnfield systems running their DIMMs at a higher clock speed. However, the P55 platforms have a decisive advantage when we look at memory access latency. A third memory channel apparently doesn’t help much in CPU-Z’s latency test.

There’s very little difference in memory performance between the three P55 boards, which is to be expected. Gigabyte appears to be pushing the CPU’s onboard memory controller a little harder than the others, though.

Application performance
Chipsets and motherboards don’t affect application performance as much as they used to. Still, we’ve used a small collection of application tests that simulate common tasks to highlight the differences between our P55 and X58 systems.

Interestingly, the P55 scores better in the first pass of our x264 encoding test. The X58 pulls ahead in round two, suggesting that the second pass may take better advantage of more than four cores. Or maybe it’s just hungrier for memory bandwidth.

7-Zip can spin up eight threads on the Core i7, and that translates to a healthy lead for our X58 system. The scaling from four threads on P55 to eight on X58 isn’t nearly linear, though.

Only a second separates the P55 and X58 systems in our panoramic stitch test. I suspect the X58 config’s superior memory bandwidth is responsible for its slight edge.

Games and multi-GPU scaling
With the X58 offering two full-bandwidth x16 slots to multi-GPU configurations via a proper north-bridge chip, and the P55 offloading that responsibility to Lynnfield’s dual-x8 onboard PCIe, I was curious to see whether game performance differs between the two. And so I fired up a few recent titles to find out. These games were run at 1080p with 4X antialiasing, 16X anisotropic filtering, and all their eye candy options cranked. I used in-game timedemos with Left 4 Dead and Far Cry 2 before moving onto FRAPS to log my time with Call of Duty and GRID. When using FRAPS, I recorded five 60-second gameplay sessions to ensure consistent, repeatable results.

Each platform was tested with a single GeForce GTX 260 896MB and Radeon HD 4870 1GB, and then with two of each in CrossFire and SLI, respectively. Our purpose here isn’t to pass judgment on which single- or multi-GPU solution is superior, though. Instead, we’re concentrating on how these graphics teaming schemes scale on each platform.

This one’s just about a wash. With just one of either graphics card installed, the X58 and P55 are locked in a dead heat. The X58 does prove faster in Call of Duty and GRID with CrossFire and SLI, but only by slim margins.

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

We’ll begin our storage tests with IOMeter, which subjects our systems to increasing multi-user loads. We used IOMeter’s workstation and database test patterns, since those are more appropriate for desktop systems than the file or web server test patterns. This particular test makes good use of the Native Command Queuing capability built into the AHCI specification.

The P55’s SATA controller is virtually the same as the one inside the X58’s ICH10R south bridge, so it’s no surprise to see nearly identical transaction rates here. Both chipsets scale IOps nicely as the load increases.

With Hyper-Threading presenting eight cores to Windows, our X58 platform will have a definite advantage over the four-core P55 rig when we look at CPU utilization results. Here, the P55 system pulls more than twice the admittedly few processor resources required by the X58.

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

Interestingly, the P55’s read burst speed is a good 40MB/s slower than the X58’s. This appears to be a problem with the Gigabyte P55 motherboard rather than a limitation of the chipset itself; the Deluxe hit burst speeds of 254MB/s with the very same hard drive. Its sustained read and write speeds were also faster at 110.5 and 109.6MB/s, respectively.

The X58’s random access times are a hair quicker than those of the P55. I wouldn’t get too worried about that or the difference in CPU utilization, though. HD Tach’s margin of error for the CPU utilization test is +/- 2%, which just about evens things out.

USB performance
Our USB transfer speed tests were conducted with a USB 2.0/FireWire external hard drive enclosure connected to a 7,200-RPM Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.

Intel’s USB controller tweaking pays dividends for the P55, which offers higher transfer rates than the X58 across the board. The X58’s CPU utilization is predictably lower, of course, but the P55’s isn’t exactly high at just 4%.

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

PCI performance
To test PCI performance, we used the same NTttcp test methods and a PCI Intel GigE NIC.

Again, we see the X58 system’s higher core count skewing CPU utilization test results. The P55 and X58 are evenly matched in the PCI Express GigE throughput test, but when we switch to a PCI networking card, the P55 pulls up a little bit slower.

Power consumption
That covers the chipset-specific portion of today’s festivities. Now it’s time to switch gears and explore variables more dependent on motherboard attributes than core-logic components. First up we have power consumption tests. Here 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 render alongside the rthdribl HDR lighting demo. Windows 7’s “Balanced” performance profile, which uses CPU power management features, was enabled for this all of our other tests.

Power consumption has become a hot topic among motherboard makers of late, and each board has a few different configuration options tied to BIOS switches and additional Windows software. The GD65 takes the simplest approach: either enable active power-phase switching (APS in the graphs below) or leave it off. With the UD6, you can run with or without Dynamic Energy Saver software installed. There’s a Dynamic Energy Saver option for the UD5, too, but it didn’t work right with the BIOS we used for testing. The Asus board has three options: an Xtreme Phase mode that’s enabled by default and apparently does power phase scaling, an Xtreme-less normal operation mode, and a third mode calibrated by EPU software for Windows.

The GD65 draws less power than the other boards, but then it also has fewer peripheral chips and power phases. The UD6 isn’t far behind, and you certainly don’t need to bother with Gigabyte’s DES software. That isn’t the case with the Deluxe, which pretty much requires Asus’ EPU app to get into the same power consumption ballpark as the other P55 boards.

Under load, the P55 pack tightens up quite a bit. MSI still comes out on top, but the Asus board draws fewer watts than the Gigabyte.

Our lone X58 board consistently draws more power than the P55 models, nicely highlighting the Lynnfield platform’s power-efficiency advantage.

Motherboard peripheral performance
Core logic chipsets integrate a wealth of peripherals, but they don’t handle everything. FireWire, Ethernet, and audio are farmed out to auxiliary chips, for example. To provide a closer look at the peripheral performance you can expect from the motherboards we’ve tested today, we’ve compiled Ethernet, Serial ATA, USB, FireWire, and audio performance results below.

HD Tach
FireWire performance

Read burst
speed (MB/s)

Average read
speed (MB/s)

Average write
speed (MB/s)

CPU utilization

Asus P7P55D Deluxe
41.6 37.3 28.5 2.0

Gigabyte GA-EX58-UD5
30.5 28.8 17.2 1.0

Gigabyte GA-P55-UD6
40.9 35.6 25.6 2.0

MSI P55-GD65
31.8 28.6 18.7 3.3

The P7P55D has the fastest FireWire implementation of the lot, trailed only slightly by the UD6. MSI’s GD65 shows some weakness here, with notably slower read and write speeds than the other P55 offerings.

HD Tach
USB performance

Read burst
speed (MB/s)

Average read
speed (MB/s)

Average write
speed (MB/s)

CPU utilization

Asus P7P55D Deluxe
35.1 34.5 33.7 2.7

Gigabyte GA-EX58-UD5
33.9 32.5 28.7 1.3

Gigabyte GA-P55-UD6
35.7 34.1 32.3 4.3

MSI P66-GD65
32.0 29.0 24.2 9.3

Despite using the very same P55 PCH as the Deluxe and UD6, the GD65’s USB transfer rates are also a little slower. Curiously, the MSI board’s USB CPU utilization also looks a little high when compared to the rest of the P55 pack. We alerted MSI to this issue, and they’re looking into it.

HD Tach
Serial ATA performance

Read burst
speed (MB/s)

Average read
speed (MB/s)

Average write
speed (MB/s)
Random access time

CPU utilization


P7P55D Deluxe (P55)
254.0 110.5 109.6 7.2 3.0

Asus P7P55D Deluxe (JMB363)
176.3 110.5 80.5 7.1 6.0

Asus P7P55D Deluxe (JMB322)
146.9 110.5 81.1 7.1 3.7

Gigabyte GA-EX58-UD5 (ICH10R)
253.6 110.5 111.0 7.1 1.3

Gigabyte GA-EX58-UD5 (GSATA)
145.6 110.5 81.5 7.1 2.0

Gigabyte GA-P55-UD6 (P55)
212.8 108.0 108.3 7.3 5.0

Gigabyte GA-P55-UD6 (GSATA)
181.9 110.5 81.2 7.0 4.3

MSI P55-GD65 (P55)
205.1 106.7 109.6 7.4 6.3

MSI P55-GD65 (JMB363)
163.8 108.5 76.2 7.2 7.0

All of these boards have at least one auxiliary storage controller, but none of those chips can match the performance of the P55’s own SATA setup. Interestingly, the slow P55 burst speeds exhibited by the Gigabyte board also afflict the MSI.

NTttcp Ethernet
Throughput (MBps)
CPU utilization (%)

Asus P7P55D Deluxe (RTL8111D)
941.3 6.1

Asus P7P55D Deluxe (RTL8110)
712.5 6.2

Gigabyte GA-EX58-UD5 (1)
940.8 2.4

Gigabyte GA-EX58-UD5 (2)
941.5 2.3

Gigabyte GA-P55-UD6 (1)
935.3 4.7

Gigabyte GA-P55-UD6 (2)
941.2 4.7

MSI P66-GD65 (1)
938.9 9.0

MSI P66-GD65 (2)
941.5 5.3

Gigabit Ethernet performance is pretty even across the field with one exception: the P7P55D Deluxe’s secondary LAN port. It’s driven by a Realtek controller that rides the antiquated PCI bus, whose limited bandwidth caps throughput more than 200Mbps shy of what’s achieved by PCIe Gigabit chips.

RightMark Audio
Analyzer audio quality

Overall score

Frequency response

Noise level

Dynamic range


THD + Noise

IMD + Noise

Stereo Crosstalk

IMD at 10kHz

Asus P7P55D Deluxe
4 5 4 4 5 3 4 4 4

Gigabyte GA-EX58-UD5
5 5 5 5 5 4 5 4 5

Gigabyte GA-P55-UD6
5 5 5 5 5 4 5 5 5

MSI P66-GD65
5 5 5 5 5 3 5 5 5

Our RMAA results give us a quick assessment of analog audio signal quality with a 24-bit, 192kHz loopback test from the front-channel output to the line input. The UD6 and GD65 both score a little higher than the Deluxe, which isn’t terribly surprising given that they both use similar ALC889 codec chips. We haven’t seen much of the Via audio codec Asus is using here. In fact, it’s not even listed on Via’s web site.

For enthusiasts, one of the best things about Intel’s mid-range chipsets is how amenable they’ve traditionally been to overclocking. Lynnfield takes the chipset right out of the overclocking equation, though. As a glorified I/O hub, the P55 Express is unaffected when the CPU is pushed beyond its default clock speed. Since current Lynnfield models have locked upper multipliers, most folks will probably be overclocking the old-fashioned way: by increasing the system’s base clock. For our first round of tests, we’ve isolated the base clock on each motherboard to see how far it can be pushed. To take other variables out of the equation, I dropped the CPU and memory multipliers to their lowest possible values before cranking the base clock beyond its default 133MHz speed. Along the way, I tested stability with a four-way Prime95 load with the rthdribl HDR lighting demo riding shotgun.

The P7P55D Deluxe sailed up to a 210MHz base clock without so much as a smidgen of extra voltage or a hint of instability. 220MHz refused to post, however, and no amount of voltage fiddling seemed to help. I even relaxed the system’s memory timings, but that didn’t make a difference.

To Asus’ credit, the Deluxe needed a CPU voltage of just 1.064V to reach a 210MHz base clock. The other boards pushed the CPU voltage to over 1.3V at their maximum base clock speed.

Like the Asus board, the Gigabyte effortlessly ramped up to a 210MHz base clock without additional tweaking. And like the Deluxe, it stubbornly refused to post at 220MHz, regardless of how much extra voltage I pumped into the processor or other system components.

The GD65 made it up to a stable 200MHz base clock, but at 210MHz, it consistently hung during the Windows 7 loading process. I tried more voltage, looser memory timings, and threatening language, but none of that managed to coax the board into Windows at 210MHz. 200MHz is certainly plenty of base clock headroom for the sort of rational overclocking enthusiasts seem to prefer.

Overclocking via the BIOS appears to be a dying art given how hard Asus, Gigabyte, and MSI are pimping the auto-overclocking features built into their new P55 boards. I suppose they’re trying to bring mainstream users into the fold, but there’s also something to be said for automating some of the trial-and-error that enthusiasts typically go through when testing the limits of a new CPU. Curious to see how well these automated overclocking approaches actually work, I let each board try to overclock itself.

Asus’ TurboV auto-overclocking scheme can be invoked in the BIOS or launched by a new Windows app (which, I might add, was developed in North America rather than Taiwan). The application’s interface looks pretty slick, and you can choose whether to let it manipulate voltages or change DRAM frequencies. Once the tuning process begins, TurboV slowly turns up clock speeds, rebooting and apparently verifying system stability at each step. I didn’t have any problems with the Windows app, but it wasn’t terribly aggressive. TurboV eventually settled on a measly 137MHz base clock speed that, combined with a 21X multiplier, pushed the CPU to 2.82GHz on 1.264V.

Unimpressed, I fired up the BIOS and ran TurboV from there. Some time later, it hard locked the system and forced me to reset the CMOS to get the board to boot. Ugh. Asus says this auto-overclocking feature is a work in progress, which is good, because it could certainly use some refinement—and perhaps a dose of optimism.

I’d be more willing to forgive TurboV’s uninspired overclock if Gigabyte’s Smart QuickBoost software didn’t produce much better results. Unlike TurboV, Quick Boost doesn’t loop through multiple attempts testing stability. Instead, it’s been programmed with ideal clock speed ranges based on Gigabyte’s experience overclocking Lynnfield CPUs. You can choose between three levels of, er, boost, and feeling cocky, I gave Twin Turbo a shot. A reboot later, and the system was stable at 3.36GHz on a 160MHz base clock with just 1.232V running to the CPU. Now that’s more like it.

Smart QuickBoost still requires Windows software, which is a bit of a drag. MSI’s OC Genie auto-overclocking scheme is much simpler: toggle the Genie button on the board, boot it up, and the system will overclock itself. Then, if you’re feeling really keen, you can adjust the base clock speed via onboard + and – buttons.

OC Genie was the only auto-overclocker to lower the CPU multiplier, honing in on a 197MHz base clock and 17X multiplier that pushed the CPU to 3.34GHz. Hitting that speed required 1.352V according to CPU-Z, which is more juice than the Gigabyte board needed to bring itself up to roughly the same speed.

If you’re in the market for a Lynnfield system, you’re stuck with Intel’s P55 Express chipset. And that’s a good thing, because this plucky little mid-range PCH has enough I/O goodness and expansion capacity to anchor high-end motherboards well above its pay grade. In a sense, the P55 Express feels like an updated south bridge chip; an ICH11R, if you will. It’s an evolutionary step up from the ICH10R, employing the same excellent Serial ATA RAID controller, plenty of PCI Express connectivity, and an improved USB component. Plus, it’s available in a tiny, power-efficient package.

My only complaints about the PCH are that it doesn’t support 6Gbps SATA or full-bandwidth PCIe 2.0. Even I’ll admit that 6Gbps SATA would be a stretch given the chipset’s mid-range aspirations and the peak performance of today’s fastest solid-state drives, though. Plus, few integrated peripherals and expansion cards really require PCIe signaling rates faster than 2.5GT/s. In reality, then, the P55 Express is well-tailored for Lynnfield and enthusiast-oriented motherboards.

We pored over P55 boards from the big three today, and they all look pretty good. There are few performance differences between them, and their BIOSes are nearly as well-equipped. Heck, the three even have similar peripheral payloads and expansion slots. They pushed our Lynnfield CPU’s base clock to within 10MHz of each other, too.

To tease our favorites from the bunch, we have to get picky. Attention to detail is usually an Asus strong suit, and in many ways, the P7P55D Deluxe is as refined an enthusiast board as we’ve seen. The layout is nearly perfect, for example, and we needed much less voltage to hit our maximum base overclock than with the others. But there are also things about the Deluxe that, well, aren’t perfect. The auto-overclocking scheme needs work, and so does Asus’ approach to power saving. Windows software shouldn’t be required for competitive power consumption. I’m also not crazy about the secondary PCI-based GigE chip or the lack of hybrid eSATA/USB connectivity. This is supposed to be a high-end motherboard, after all.

With a $250 MSRP, Gigabyte’s GA-P55-UD6 has an even greater premium to justify than the Deluxe. For the most part, it does a better job. You get more of everything: six DIMM slots, 10 internal SATA ports, two hybrid eSATA USB ports, and an audio codec that does a heck of a SoundStorm impression. All is not flawless, of course. One of the x1 slots is crowded by the chipset cooler, the BIOS’s fan speed controls are frankly inadequate for a high-end motherboard, and something’s up with SATA burst rates.

That leaves us with MSI’s P55-GD65, which has the distinct advantage of costing just $160. This is a true mid-range offering, and as we saw today, it doesn’t force you to give up much. With the exception of slower USB transfers and some odd SATA burst behavior, the GD65 was every bit as fast as the other boards. Plus, it’s quite power-efficient and has simple but effective power management and auto-overclocking features. Heck, the BIOS even has decent fan speed controls.

It’s tempting to name the GD65 our Editor’s Choice on the strength of its more attractive value proposition. However, I’m going to reserve judgment until we’ve had a chance to test a few other boards in its price range. Expect more on that front soon.

Between the high-end Deluxe and UD6, I think Gigabyte gives you more for your money. However, I don’t think either makes a compelling case for spending $225 or more on a Lynnfield board. You’re much better off checking out the mid-range boards Asus and Gigabyte have cooked up with the P55. We’ll be exploring those options shortly.

Update — We’ve now had a chance to look at a couple of other affordable P55 boards. Against these more direct rivals, the MSI P55-GD65 picked up a TR Recommended award.

0 responses to “Intel’s P55 Express chipset

  1. Do you get any idea of the stability and compatibility with memory modules? I don’t want to pick up anything unstable..

  2. Four cores and 32nm = Sandy Bridge. Unless Intel changes its plan.
    Or you could go AMD, I guess, whenever they get their 32nm product out the door.

  3. Yes, the datasheet confirms this. Well you’re right it doesn’t really change much, still I think it’s disappointing that not full-rate pcie 2.0 is supported. With graphics, everyone switched to pcie 2.0 very quickly, though benefits were limited at best, but for everything else we’re STILL stuck at pcie 1.1 speed.
    Clearly things like 10 gb/s ethernet adapters (ok they aren’t quite mainstream…) and storage controllers could benefit – can’t even push sata II throughput through pcie x1 at 2.5GT/s rate, and sata III requires more than one pcie 1.1 lane to even make sense at all.

  4. Looks like it’s PCIe 2.0 limited to 2.5GT/s, which wasn’t clear from Intel’s original briefing materials. I’ve updated the review accordingly, although our overall conclusions on the chipset remain unchanged. We’re awaiting comment from Intel.

  5. Can’t believe the socket 775 era is truly over, with P945,965, P35, P45 all now ‘obsolete’…long live the new king P55 with its PCH and everything. Sure it’s evolutionary, but this looks like another solid product from Intel.

    Adios FSB on mainstream componentry.

  6. I hope you’re right! Core2 mITX solutions are cramped for board space and have components on the underside, no doubt contributing to higher cost. Whereas H55/H57 “chipset” might snugly fit. There is still the issue with 65W to dissipate.

    On another note, we may soon be comparing apples to pears (45nm lynnz vs 32nm clarx). Perhaps TR could run some tests with constrained thermal design. E.g. pit 4C/4T lynnfield against 2C/4T clarkdale, one of them under/overclocked to match the system power draw.

  7. Maybe not 50 lanes of PCIe 2.0, but in 2010, there’s a very real possibility of an X68 with PCIe 3.0, which is double the bandwidth and power all over again.

    It’s pretty much inevitable, considering that Sandybridge should be out the same time that PCIe 3.0 is ready to go.

  8. That is why I will be waiting a little while longer for this. Like MMO said, let’s wait for the P57 and see what happens. And I’m sure there is a X68 on the way. Something has to support the SATA III and USB 3.0 and 50 16X PCIe lanes! No, no, sorry, got carried away, the last one is wrong. But you get my drift…

  9. Unfortunately, this particular chipset is more like an X58 with things taken away, rather than anything new. No SATA III, no USB 3.0, no north bridge, no GPU, two memory channels, 16 PCIe lanes…just nothing of interest. I think it’s understandable that no one really cares lol.

    If Intel has actually cancelled Braidwood, the motherboard market is going to be pretty lame for pretty much the next 6 months. There’s been nothing but hold ups and disappointments as of late, on all sides.

  10. I guess CPUs are 6 times more popular [based on comments] than the chipsets that help run them. Who’da thunkg{

  11. mATX supports more than a single slot — typically a couple of PCIe and a PCI, or vice versa. Sure, those will be connected to the chipset rather than the CPU (like every system prior to Lynnfield), but they won’t be used for high-bandwidth GPU boards so it doesn’t matter. And even that is more expansion capacity than I need for most applications, since so much is integrated on the mobo, and it’s fine even for HTPC applications with a tuner and discrete audio cards.
    §[<<]§ §[<<]§ ITX will be great for Clarkdale, for really compact nettops that still offer more hauling power than Atom.

  12. mITX will be great when Clarkdale is out, otherwise right now you’d be forced to use the only expansion slot for a graphics card. Not that that’s terrible but it’s less than ideal imo. Of course it means less potential in CPU power, hmm maybe it rally is too bad that NV may not have chipsets. Can’t believe I just said that :p

  13. Please include an Intel P55 board in the mix at some point. My current ASUS board is reliable but the number of things that don’t work are significant so I am probably going to go all Intel on my next rig.

    My A8NSLI-Premium would not mix SATA and IDE drives (instant corruption on the IDE drives) and will not sleep without corrupting the last partition on my 3rd SATA HD (works fine with only 2 HDs).

  14. “Crippled” is poor term. How is it crippled? Bloomfield has one less memory channel than Beckton, so I guess it is “crippled” also. Wait, Beckton could have infinitely more memory channels — omg, it’s crippled too! The chip is only crippled if the memory subsystem becomes the gating factor in most (or even any) common tasks performed by its intended customer base. The reviews show that isn’t the case; hence, it’s not “crippled.” Don’t let Snakeoil take over your brain.

    I agree most of these mobos are overkill, and the Gigabyte one in particular looks like they decided to just reuse their existing stock of X58 parts — perhaps for time-to-market reasons, perhaps because they had excess inventory they were trying to eliminate, perhaps because they thought a board at the price-point they were targeting looked too bare and un-featured if it didn’t have all the whizzy heat pipes. (And on that last point they might be right — I’ve noticed the folks eye-balling boards on display at Fry’s seem to always gravitate to the one with the most heatsink bling).

    Undoubtedly we’ll see cheaper, more sensible boards — they’re already starting to appear, and I expect TR will feature a round-up. Personally I’m looking forward to a comparison of mATX boards, which seems to be the perfect match for Lynnfield since I have zero interest in more than one x16 slot, SLI/CF graphics, or a huge HVAC case to house it. That’s where the good bang/buck deals for a Lynnfield system likely will be found. In fact, given the sparse board layout the P55 allows, we may even see someone like Zotac do an ITX P55 board, and wouldn’t that make the “Pocket Swiss Army Knife” build rock?

    Of course “budget” means a slightly different thing here at TR, where folks generally turn up their noses at the bottom end of the performance heap and don’t consider “cheap quad” to be an oxymoron. And “Budget Bloomfield” is another category altogether. But the fact remains that Intel doesn’t consider Lynnfield, or the P55/P58 platform, to be its “budget” offering.

  15. MSI has the same range of boards in price and ‘features’ they just happened to test a middle range one here.

  16. I’m still hoping for more decent mid-low 100’s boards. I still have yet to spend more than about $130 for a motherboard in almost 20 years of PC building. MSI board isn’t half bad so far, prices will come down in the coming months I’m sure. If there are some decent $120-140 boards come spring I might jump on this platform.

  17. Well, its basically a budget Bloomfield. On that roadmap its one level down on all the Bloomfields and its also lower mainstream too. A crippled version having 1 less memory channel, and fewer PCIe lanes. Less chips on the mainboard. It makes no sense to make a Bloomfield type mainboard like the Gigabyte in this review. Its got heatsinks for chips that don’t exist. Its got 3 PCIe slots for graphics when the max should be 2.

  18. Pull out a SIMM thats been in for years (which would be all of them these days, if any are left) and check the pcb, that uneven retention does things.

    Bad design is bad.

  19. I suspect just like with fb-dimms, there will be a latency penalty, hence just like fb-dimms it wouldn’t make sense for the desktop. I think that current xeon-mp platform using fb-dimms can also already use at least 4 dimms per channel.

  20. Never had problems with SIMMs. They work right, the trick was using the proper angle to insert the modules.

  21. I remember some of the very first motherboards with DIMM slots, who also (typically) had SIMM slots; this was probably in the Super 7 K6 days. The DIMM retainer clips on those were horrible, in that they were hard to engage and didn’t hold onto the DIMM very well.

  22. Folks, Intel does not consider Lynnfield a “budget” processor so it’s not surprising the boards for it aren’t “budget” either. There will be (and already are) cheap(er) P55 mobos, but naturally the makers would rather tempt you with their higher-end (and more profitable) offerings first. This is also why Intel has been introducing Nehalem from the top down, to try to tempt the early adopters into buying at least one tier higher than they otherwise would if all the options were available immediately. And as soon as everybody who’s going to has bought the P55 boards, Intel and the Mobo makers will be back to tempt you with P57 and Braidwood.

    Lynnfield — and the mobos for it — fit into the segments Intel calls “Peformance” and the top of “Mainstream”… and even the latter is not what consumers in the real world would call “budget.” The budget processors fall into what Intel calls “Essential” and “Value” and Clarkdale is destined for those. That’s where you’ll see the real /[<]§ (That slide dates back to before Havendale was canceled and replaced by Clarkdale, but the segments remain).

  23. The EX server version (Beckton) supports up to 4 DIMMs per channel (and four channels, so 16 DIMMs per socket), thanks to “Intel Scalable Memory Buffers” (aka “Buffer on Board” or BoB — essentially the relevant parts of FB-DIMMs moved off the DIMMs and onto the mobo; AMD was showing something similar and calling them “Microbuffers” though that term also has a history with Rambus and AMD seems to have dropped the idea or at least stopped talking about it).

    Of course, the tech almost certainly won’t appear in desktop systems; 4GB DDR3 DIMMs will be cheap soon enough.

  24. So you’re waiting for (dual core) Clarkdale? Or you’re planning on spending the big bucks for X58/1366 with (six core) Gulftown? Otherwise you’re waiting for Sandy Bridge in about a year.

  25. SIMM retainers really sucked, if thats what you mean.

    It always felt like you were going to break something and it tended to put bowed/angled pressure across the whole pcb.

    I don’t remember any real issues with the first DIMM slots, aka later EDO/early SDRAM days, besides the fact you could buy the same memory type in multiple formats and the confusion that brought. They have been refined since then obviously but moving away from SIMM was a real joy.

  26. No big news here.
    MSI have an P55 GD80 model, look at guru3d.
    For an upgrade, I will wait for the 32nm tech.

  27. Not exactly I bet you could have 4 ram slots too per channel if all of them are populated with single-sided ram. There is very little to no difference from the chip point of view between one double-sided and two single-sided modules (it might be problematic though due to the longer traces needed on the pcb however, not to mention of course it makes even less sense than 3 slots per channel).

  28. Well, Nehalem supports 3 DIMMs per channel in server version, but desktop chips don’t, it’s just 2 (or rather, it’s really 4 sides) per channel. This is similar to Opterons – these are the same chips as Phenoms basically, but the former supports up to 4 dimms per channel while the latter only 2. Reason is that server chips use registered memory which makes it electrically easier to install more modules in parallel (and that they actually have reduced memory frequency if you install more than 2 modules certainly helps too). For desktop chips, it seems everybody has given up making more than 4 sides work per channel – there used to be chipsets which supported that a LONG time ago (SDR/DDR1 timeframe), and it very rarely ran stable…
    But yes, a quick look in the manual from that gigabyte board certainly confirmed it.

  29. Mobo makers are just trying to bilk the early adopters for a month or two. It’ll get better.

  30. What a joke paying $250 for a budget board. Gigabyte is crazy. And a 3rd PCIe slot offering 4x … what’s the point in offering 3 way SLI at 8/8/4?

  31. All of this is mentioned in the manual/memory compat. page on their website.

    I know i was more excited about this board until i actually read the details/limitations on using 6 dimms.

  32. I knew Nehalem supported three DIMMs per channel (so 9 total for 1366 or 6 for 1156) but didn’t know about the distinction between sides of the RAM modules. Couldn’t find anything about it via Google but I didn’t look very hard.

  33. So Nehalem supports three DIMMs per channel but only four sides of memory per? Screwy. In any case, the RAMdisk is appreciably faster than my SSD. If the 6 DIMMs can’t be populated to 12gb I’ll probably just hold out until 4gb modules get cheaper. I’m not in a hurry to upgrade, just speculating.

  34. P55 can be considered to be ICH11. Everything makes more sense if you consider it to be the successor to the ICH10R — the low bandwidth of the PCI Express link (DMI is said to be a PCI Express 1.1 x4 interface with proprietary signaling), the behavior of the GigE interface (yup, ICH8 onwards have all integrated a MAC, although the cost of the PHY seems to deter most makers aside from vPro systems), the SATA ports, and so on.

    Architecturally this makes sense since the functions of what used to be the northbridge are by and large now integrated on the LGA 1156 CPU — memory controller, integrated graphics, PCI Express Graphics links, IOMMU (VT-d). (As with 965, 3-series and 4-series, the lower-bandwidth PCIe 1.1 x1 and x4 slots are fed by the southbridge, i.e. the P55.)

    The integrated-graphics, dual-core Clarkdale looks even more like Core 2 and earlier plus Intel integrated chipset architecturally. There’s a discrete 45 nm chip with the memory controller, DMI link and integrated graphics, connected internally via an interconnect (QuickPath interconnect rather than the FSB thankfully) to the actual processor cores on a 32 nm chip. The difference is just that the 45 nm northbridge is on the processor package this time, not the motherboard. But from a system architecture it smells a lot like Core 2.

    And of course, Clarkdale is paired with a 5-series “chipset” (really a southbridge). H55 and H57 for people who want switchable graphics (between the low-power low-performance integrated GMA graphics on the 45 nm GMCH die on the processor socket, and any discrete part you may have installed), Q55 for people who want AMT/vPro (also with integrated graphics), and presumably P55 for people who don’t need or want to be able to use the integrated graphics in favor of discrete graphics alone.

  35. I think 6 dimm slots would be possible, there are bloomfield micro-ATX boards out there with 6 dimm slots even.
    However, 6 dimm slots on Lynnfield based boards are completely pointless and won’t help you (read #7).
    Maybe you want to get a SSD instead…

  36. Woah there. I was referring to the physical slots themselves, not their connectivity. I just want a sub-$200 mATX board with 6 DIMM slots, one or two graphics slots and one x1 slot. All the other fluff on these boards doesn’t appeal to me in the slightest (on-board power/reset is handy but I’d only use it a few times, if at all). The rest of the crap –various networking and storage chips, gaudy heatsinks, excessive backplate connectors, etc– it’s useless to me.

    As the reviewed MSI board indicates, less feature-bloated mainstream boards are just as stable and just as good at overclocking as their overpriced flagship brethren… which is why I tend to scoff at the inclusion of mega-top-end gear in these types of “enthusiast” reviews. Enthusiasts != people who spend money like morons. “Here’s a board that costs 100% more than a sensible mainstream offering and performs 1% better!”

  37. The Gigabyte mATX boards do look nice since they don’t go psycho premium ROG pricing like Asus did. I just wich that they didn’t have TWO PCI slots on the less expensive one in this day and age.

  38. Yeah, but it’s taken a year for X58 boards to get to that level. The P55 boards are already fairly reasonably priced, and it should go down even further over the next few months.

  39. arnt some of the x58 boards running in this price range atm? These P55’s were supposedly gonna be “budget” boards.

  40. Yes if board uses that x4 pcie link on x58 it’s probably just a x4 pcie slot for storage adapters and the likes.
    Anyway, intel datasheet for p55 is out. This confirms that, the P55 PCH pcie ports are indeed pcie 2.0, but, run at 2.5GT/s, which is the same speed as pcie 1.1. That’s similar to some USB devices which come advertized as “USB 2.0” or “USB 2.0 Full Speed” which for all practical purposes is USB 1.1… Aside from the possible doubled to 5GT/s throughput, pcie 2.0 doesn’t really have anything worth mentioning in the spec over 1.1 (ok link management for power saving purposes but that’s not going to do much with x1 links).
    It’s a bit a shame since most if not all cheap add-on card sata III controllers will be pcie x1 only, but possibly support the faster transfer rates. But coupled with any chipset only supporting 2.5GT/s transfer rates, there’s pretty much zero point in using sata III anyway since the transfer rate of the pcie bus is already (slightly) slower than what sata II provides… It’s only really a potential problem with SSDs though.

  41. There are a few boards on Newegg from Asus and Gigabyte at or below the price of the MSI. To me the best value in P55-Land is the Gigabyte mATX boards at $110, the GA-P55M-UD2. I personally will never buy a ATX board again so if I were to pick up a Core i5 that would be the board right now.

  42. i5 plus the MSI board is going to be tough to beat in its price range when looking to build a new system. Let’s see what else comes along to compete in that P55 price bracket.

  43. I am rather surprised that ASUS board does not have DIMM retainers. I suppose that the board has a different retaining mechanism, otherwise it is going drive its users crazy with seating issues.

    I remember the horror stories with first-generation of DIMM retainers. Damm retainers blew chunks and it was a PITA to seat the DIMMs properly.

  44. Whoever at Asus thought it was acceptable to have a $200+ motherboard without eSATA deserves a kick in the shins.

  45. what are the chances that the Asus has memory come loose, since there’s no tab to retain the RAMs?

  46. True about those additional lanes on the x58 Northbridge. I suppose if some mobo makers were smart they could have used them for…I’m not sure what. 10GigE or multiple GigE controllers? A high network load plus high disk traffic is the situation I can think of easily where the DMI link might bottleneck. Probably they just used them for 1x or 4x PCIe slots, with the latter it might be useful for RAID cards.

  47. This is true though X58 also had an additional x4 pcie-connection (apart from the 2×16 for graphics) directly off the chip hence “freeing” the DMI from at least one high-bandwidth consumer. Don’t know if boards actually used that however…
    I just don’t quite understand why DMI didn’t get a boost, along with these pcie 1x links to pcie 2.0 speeds (DMI still seems to operate at pcie 1.0 speeds too otherwise that would be 4GB/s).
    And you’re right in practice it probably usually won’t really matter.

  48. 6 memory slots on the gigabyte are useless.
    Contrary to what the article states, this does not allow you to use 6 cheap 2GB modules for 12GB. Since it’s only possible to install 4 _sides_ of memory per channel. But all those cheap 2GB memory modules are double sided (because single sided would require 8 2gbit chips, IOW you could just as well get those expensive 4GB modules with 16 such chips).
    The only (very very slim) advantage this provides is you can use 2GB/1GB/1GB per channel (since AFAIK 1GB DDR3 modules are always single-sided using 8 1gbit chips) instead of 2GB/2GB if you happen to have some 1GB modules lying around (which seems rather unlikely).

  49. gen-two pcie x1? Don’t think so. If the bandwidth figures are consistent with the pcie-16x figures, this is total bidirectional bandwidth, so 2x250MB/s, in other words, pcie 1.0. (Plus you’d think intel would proudly write “2.0” there wouldn’t you…). That’s a bit disappointing, though given the paltry DMI speed maybe not really a big problem.

  50. *nevermind, I thought P55 was 16x CPU + 4x chipset lanes. Still, hanging one GPU off the southbridge with one off the CPU would probably be bad and you can’t do two 16x electrical slots anyway without PCIe lane splitters (or wahtever they’re called) which introduce latency negating the lane advantage anyway.

    Personally I’d be fine with one 16x electrical and the rest 1x electrical with the option for 8x/8x for CF/SLI, not that I’d use SLI/CF but that’s how it should be done. What I don’t like is when a 1x card in a secondary physical 16x slot cuts the main 16x slot to 8x. It may only make a tiny bit of difference for that main slot graphics card but it tweaks my anal retentive nature.

  51. The MSI’s sparseness makes it sexy; it’s like a plain black suit or a simple black dress. Enjoyable article, as usual. I’m more interested in the remaining sub-$200 offerings you’ve yet to write up, though.

    Could 6 DIMM slots fit on an mATX P55? I wish somebody would make one like that with two x16 and one x1 (and none of the extra storage & networking nonsense). I’d settle on one x16; I like the extra slot for “teh fyoocher” but it seems you can always trade up and get the same kinda performance from a beefier single card. A board like that for a buck fifty? Sign me up! Probably just a wet dream…

    I want six RAM slots because I’m spoiled by Ramdisk Plus. Loading programs from RAM feels instantaneous and everything (Windows included) seems more responsive if it uses the RAM for temp storage & cache. Great as all that is, none of it compares to the bliss of saving during a painting and not waiting for damned ever to get back to it. How did I manage?

    I wonder what 4gb RAM modules will cost this time next year…

About Geoff Gasior