Intel’s P45 Express chipset

Intel’s P35 Express chipset has been a revelation for the Core 2 platform, delivering the same excellent performance, low power consumption, and generous overclocking headroom of more expensive X38 and X48 variants at a much lower cost. Or at only a slightly lower cost, depending on whether you prefer your motherboard loaded with more frills and extras than the spawn of Pimp My Ride. But that’s the beauty of the P35; it’s equally capable of powering mainstream desktops, high-end gaming rigs, and the sort of balanced best-bang-for-your-buck systems that enthusiasts tend to build.

If you’ve been around for long enough, the P35’s success should come as no surprise. Intel’s mid-range chipsets have been consistently solid year after year, establishing a virtual dynasty in an industry that doesn’t always look kindly upon incumbents. It’s no wonder, then, that expectations are high for the P35’s successor, the new P45 Express.

Intel hasn’t messed with the formula much for the P45, but there are a few new tricks up its sleeve. A redesigned memory controller promises to take better advantage of processors with 1333MHz front-side bus speeds, for example. Second generation PCI Express connectivity has also been added to the mix along with a more balanced approach to CrossFire configurations. And ever the efficient manufacturer, Intel has managed to shrink the whole thing down on a 65nm fabrication process.

The motherboard market is about to be saturated with a tidal wave of new designs based on the P45 Express, each with its own set of unique features and capabilities. We’ve already taken an early first look at what Asus is bringing to the table. Now that Intel has made the P45 official, we’re diving a little deeper, this time with a couple of new motherboards from Gigabyte. Read on for a in-depth look at how the P45 Express fares against a collection of DDR2- and DDR3-equipped competitors ranging from its popular predecessor to Nvidia’s uber-high-end nForce 790i SLI Ultra.

A die-shrunk refresh
Intel employs a “tick-tock” cadence for processor development, rolling out new architectures on each tock and then shrinking them with more advanced process technologies with each tick. Judging by the all around awesomeness of the recent “Penryn” 45nm die shrink of the original Core 2 microarchitecture, the tick-tock strategy appears to be working well. So well, in fact, that I think it may have been applied to the P45 Express, which in many ways looks like a die-shrunk P35.

Like Penryn, the P45 packs a few new elements alongside its move to a finer fabrication process. The most notable among these is easily the inclusion of second generation PCI Express connectivity. 16 lanes of PCIe 2.0 are available in the P45’s north bridge chip, and they can be configured as either a single x16 link or as a pair of x8s for CrossFire. With multi-GPU configurations finally offering compelling performance at affordable prices, the P45’s balanced approach to CrossFire is much appreciated. The P35 chipset wasn’t capable of splitting its north bridge PCI Express lanes, forcing secondary graphics cards to hang off just four south bridge PCIe lanes and compete for limited interconnect bandwidth.

In addition to refreshing the P45’s PCI Express component, Intel redesigned the chip’s memory controller to better exploit the 1333MHz front-side bus speeds made common by the Penryn Core 2 refresh. The memory controller is still compatible with both DDR2 and DDR3 memory, and it now supports up to 16GB of the former. You’ll only be able to pair up to 8GB of DDR3 memory with the P45, which given current market prices, may be a blessing in disguise.

Like the P35 that came before it, the P45 only officially supports DDR2 memory speeds up to 800MHz. DDR3 memory speeds top out at a more generous 1333MHz. However, it’s worth noting that motherboard makers are largely ignoring Intel’s official specifications and loading BIOSes with the memory bus multipliers necessary to hit higher speeds.

Mobo makers are also advertising P45 boards capable of hitting 1600MHz (quad-pumped) front-side bus speeds despite the fact that Intel only endorses FSB speeds up to 1333MHz. Intel appears content to limit official support for the 1600MHz front-side bus required by its obscenely expensive Core 2 Quad QX9770 flagship to the high-end X48 chipset, which also doubles up on the P45’s PCI Express lanes.

With even the high-end X48 relying on Intel’s DMI chipset interconnect, it’s no surprise that the P45 uses the same link to tie together its north and south bridge components. DMI has been around for a while now, and its 2GB/s of bandwidth looks a little light next to the speedy 8GB/s HyperTransport interconnect used in nForce chipsets for Core 2 processors. Keep in mind, though, that Nvidia needs the extra bandwidth because it’s made a habit of hanging graphics cards off the south bridge.

While the P45’s north bridge component at least has some new hotness to brag about, the chipset’s ICH10R south bridge chip looks like little more than a die-shrunk ICH9R. The ICH9R certainly wasn’t hurting for an upgrade, though; six Serial ATA RAID ports remains the standard and Intel’s ability to mix and match array types with “Matrix” RAID configurations is still unique among even high-end chipsets.

Heck, we’re even over Intel’s decision to exclude an IDE port from its recent chipsets, in part because Vista plays well with the JMicron storage controllers most motherboard makers have been forced to use to provide “parallel” ATA support. And offering 12 USB ports still keeps the ICH10R up with the Jonses, even if it looks like those ports are the same ones you’ll find in the old ICH9R.

Like the ICH9 family, the ICH10s also feature a Gigabit Ethernet MAC. We didn’t see many manufacturers take advantage of this capability in the ICH9R, ostensibly because Intel didn’t have a necessary PHY physical interface chip ready at launch. The ICH9R’s integrated networking capabilities finally got a chance to strut their stuff on Intel’s DX48BT2 “Bearlake 2” motherboard, which offered great throughput and the lowest CPU utilization we’ve seen from any chipset-level GigE implementation. We can only hope more motherboard makers take advantage of the ICH910R’s Gigabit credentials this time around.

About the only thing really missing from the ICH10R is PCI Express 2.0 connectivity. The onboard peripherals and x1 slots normally serviced by south bridge PCIe lanes aren’t exactly begging for bandwidth these days, but it wouldn’t hurt to have generational parity between PCI Express implementations in the P45’s north and south bridge components.

Gigabyte shows its hand
We got our first look at Gigabyte’s P45 lineup in Taiwan last month when the company revealed plans to produce no fewer than 20 different designs based on the chipset. Two of those boards have now made their way to the Benchmarking Sweatshop: the EP45-DQ6 and the EP45-DS3R. The DQ6 carries a lofty suggested retail price of $260, which doesn’t exactly put it in the mid-range territory one might expect from a P45 board. You get what you pay for, though, and the DQ6 has a number of interesting features that warrant closer inspection.

Gigabyte dresses the DQ6 in turquoisy blue, complementing its trademark hue with a virtual rainbow’s supply of near-fluorescent trim. Love it or hate it, the look is unmistakably Gigabyte.

Of course, the DQ6 is more than just an aesthetic statement. Gigabyte has also done an excellent job with the layout considering how many slots, ports, and additional peripherals have been squeezed onto the board’s ATX form factor. Even the power plugs are right where we like to see them, located along the edges of the board where their associated cabling won’t interfere with typical chassis airflow.

Like most high-end motherboards, the DQ6’s processor socket is surrounded by an ornate array of heatsinks and heatpipes. The north bridge cooler doesn’t look all that extravagant (and it probably doesn’t need to be given the P45’s 65nm underpinnings,) but there’s still plenty of radiator fin surface area dedicated to the board’s voltage regulation circuitry. Despite the additional heatsinks, though, the DQ6 can still accommodate larger aftermarket heatsinks like Scythe’s massive Ninja cooler.

From this angle we also get a good look at the DQ6’s onboard components. The board is a part of Gigabyte’s Ultra Durable 2 family, which exclusively uses solid-state capacitors, ferrite core chokes, and low RDS(on) MOSFETs. Higher quality components have become a staple of high-end motherboards, and after losing a couple of older boards to busted or otherwise leaking capacitors, I couldn’t be happier.

The ICH10R’s six Serial ATA ports apparently weren’t enough for Gigabyte, because the DQ6 features four more tied to a mix of auxiliary storage controllers from JMicron and Silicon Image. With ten ports in total, clearance problems are bound to crop up with longer graphics cards. However, Gigabyte has paid special care to ensure that the ports most prone to blockage are edge-mounted to sidestep the issue.

While most motherboard makers are content to hang additional Serial ATA ports off a single auxiliary storage controller, Gigabyte’s approach with the DQ6 is a little more complicated. The four auxiliary ports are arranged in pairs, with each couple tied to a two-port Silicon Image SATA RAID chip. This chip supports RAID 0 and 1 configurations in addition to a few combinations of the two, ala Matrix RAID. Array-related calculations are performed in hardware by an on-chip RISC processor, and since arrays are presented to the operating system as standard Serial ATA hard drives, additional storage drivers aren’t required.

Gigabyte adds another layer of goodness by hanging each Silicon Image RAID chip off a Serial ATA port provided by a JMicron storage controller. This 300MB/s SATA link does introduce a potential bottleneck, but as long as you run the JMicron chip in native IDE mode, there’s still no need to install additional storage drivers. One can also set the JMicron chip to run in RAID mode, allowing effective RAID 10 or 0+1 arrays to be built from component parts connected to the Silicon Image chips.

This all sounds like an overly complex way to gain four extra SATA RAID ports, but because the default configuration (JMicron chip in IDE mode and Silicon Image chips running in RAID 1) doesn’t require drivers, it actually gives users a relatively easy path to fault-tolerant mirroring. Just don’t mind the fancy-looking heatsinks on the additional storage controllers; they’re just for show.

The DQ6’s gravy train of excess continues on the expansion slot front where we find all sorts of PCI and PCI Express connectivity. There are seven slots in total, including two PCIe x4s that are slotted to accept longer cards. Those two slots are actually connected to a PCI Express switch chip whose heatsink you can see just behind them. The switch chip has access to four south bridge PCI Express lanes, which it shares between the two x4 slots and the board’s networking chips. We’ll get to those in a moment.

In the meantime, note the presence of handy onboard buttons at the bottom of the slot stack. Power, reset, and clear CMOS buttons are present on a few of Gigabyte’s high-end P45 boards, although they don’t make the cut for some of the budget models.

We wish Gigabyte would move the CMOS reset button to the DQ6’s port cluster, but it doesn’t look like there would be room. This board has not one, not two, not even three, but four Gigabit Ethernet ports, each backed by its own (and brand spankin’ new) Realtek RTL8111C networking chip. The GigE ports can be run individually or teamed together, but this sort of networking excess seems more appropriate for server boards than one targeted at PC enthusiasts. I’d much rather see a couple of GigE options alongside a good Wi-Fi adapter.

To no one’s surprise, Realtek gets the call on the audio front, too. The DQ6 makes use of the crab’s flagship ALC889A codec chip, which is capable of encoding multi-channel Dolby Digital Live audio on the fly. Eight USB ports round out the cluster, with onboard headers provided for an additional four ports. Gigabyte also provides onboard headers for three Firewire ports.

With all the clutter, it takes a moment to realize that the DQ6’s port cluster is completely devoid of external Serial ATA connectivity. Gigabyte prefers to provide eSATA ports via PCI back plates that can be attached to any onboard SATA port, which given the DQ6’s variety of storage controllers, makes perfect sense.

A more affordable DS3R
Although the EP45-DQ6’s cornucopia of extras make a good case for the board’s higher price tag, few enthusiasts really need or even want the additional frills. For those folks, the comparably stripped-down EP45-DS3R looks like a much better option. This board carries a suggested retail price of just $150, which is much more in keeping with the P45 chipset’s mid-range appeal.

Despite its lower price, the DS3R shares a similar layout and features the same Ultra Durable 2 components that you’ll find on the DQ6. The DS3R does lack fancy heatpipes and voltage circuitry cooling, though, and it only feeds the processor with six power phases, as opposed to 16 on the DQ6.

Still, the DS3R features dual Gigabit Ethernet ports, Firewire (that’s conveniently located in the port cluster,) and seven expansion slots, including a pair of PCIe x16s. After running the DS3R through our application benchmark suite, we found that the board is every bit as fast as the DQ6, too.

BIOS options and tweaking software
Gigabyte has done a good job with its BIOSes of late, especially on the overclocking front. The EP45-DQ6 is no exception, and as you might suspect, the DS3R offers a nearly identical suite of BIOS options.

With front-side bus options up to a whopping 1200MHz—and that’s before quad-pumping—the DQ6’s BIOS sets the tone early. Gigabyte offers a decent array of memory bus speeds, too, providing dividers that can push DIMMs up to an impressive 1333MHz—that’s DDR3 territory, folks.

Gigabyte neatly integrates all of the DQ6’s most important overclocking and tweaking options on one page. Even the big four memory timings make an appearance, although you’ll have to duck into the advanced timing menu to adjust the DRAM command rate. There, you’ll also find access to a whole host of additional timing options that you probably shouldn’t be messing with in the first place.

It’s in the voltage department that the DQ6’s BIOS really shines, though. Control over an impressive nine different voltage settings is provided, and there’s plenty of range to play with. CPU voltage options are available all the way up to 2.3V in fine 0.00625-0.02V steps. You can take the memory up to a whopping 3.0V, too. Surely, those voltage ceilings should be enough for even liquid-nitrogen-fueled overclockers.

We can’t help but feel a little spoiled by the DQ6’s generous overclocking options, which is perhaps why the BIOS’s fan speed control section is a disappointment. Temperature-based fan speed control is limited to the CPU fan header, and users only have the option of toggling between three- and four-pin fan settings. Gigabyte really needs to provide automatic fan speed control for the board’s system fan headers, as well—ideally with explicit control over temperature targets and actual fan voltages.

All of Gigabyte’s P45-based motherboards feature Trusted Platform Module (TPM) chips ripe for the next wave of invasive DRM technologies. However, before you don your tinfoil hat, note that the BIOS provides an option to disable the TPM chip completely.

We rarely have kind things to say about the tweaking software provided with most motherboards, but the latest version of Gigabyte’s EasyTune application is much improved over previous iterations. Gone is the garish, over-sized interface and awkward controls, replaced by a GUI that actually looks like it belongs in Windows. The interface has a bit of a CPU-Z feel, and it’s quite easy to use.

Hardware monitoring is supported, too, and you can save and load configuration profiles should you wish to preserve or share a particular setup.

Specifics on specifications
We’ve distilled the differences between the EP45-DQ6 and DS3R into a handy chart below. As you’ll see, both pull from the same pool of chips; the DQ6 just takes things to a bit of an extreme.

Our testing methods
We’re focusing on the P45 Express’ performance today, so we’ve busted the chipset out in DDR2 and DDR3 configurations. The Asus P5Q3 Deluxe motherboard we looked at a couple of weeks ago will be standing in to fill out the DDR3 side of the equation.

To see how Intel’s latest mid-range chipset fares against its predecessor, we’ve included DDR2 and DDR3 flavors of the P35 Express. We also have a couple of X48 setups, an X38, and Nvidia’s nForce 780i and 790i SLI Ultra chipsets to fill out the pack.

While application tests were run on each board, we’ve consolidated our south bridge peripheral testing to only include one example of each chipset. In cases where we have DDR2 and DDR3 versions of a chipset, we’ve used the results from the DDR3 config.

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

Thanks to Corsair for providing us with memory for our testing.

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

Finally, we’d like to thank Western Digital for sending Raptor WD1500ADFD hard drives for our test rigs. The Raptor’s still the fastest all-around drive on the market, and the only 10K-RPM Serial ATA drive you can buy.

We used the following versions of our test applications:

• SiSoft Sandra Standard 2008 13.12
• WorldBench 6.0 Beta 2
• TCD Labs HD Tach 3.01
• Call of Duty 4 1.4
• RightMark Audio Analyzer 6.06
• NTttcp
• CASE Lab Euler3d CFD benchmark multithreaded edition
• Intel IOMeter v2004.07.30
• Crysis 1.1
• Enemy Territory: Quake Wars 1.4
• Half-Life 2: Episode 2
• CPU-Z 1.41

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

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

Memory performance
Memory subsystem performance doesn’t always track with real-world applications, but it’s a good place to check out the P45’s 1333MHz FSB-optimized memory controller.

The P45 doesn’t look as impressive as we might have hoped here. It’s not so much that the chipset is slow, but that we expected it to at least be faster than the P35 since we’re using a processor with a 1333MHz front-side bus. Nvidia’s nForce chipsets definitely have an edge when it comes to memory subsystem performance.

Memory controllers don’t always handle four DIMMs gracefully, so we popped an additional two memory modules into each system for another round of tests. In these tests, we had to back off to a 2T command rate for the nForce 780i SLI, X48 boards, and the DDR2 P45 configuration—a common adjustment for four-DIMM configurations.

With four DIMMs installed, the P45 proves to be a little bit slower. I suspect more BIOS refinement could improve performance here. The Gigabyte EP45-DQ6 we used in our P45 DDR2 system didn’t even work with four DIMMs installed until the latest BIOS release, and the Asus P5Q3 Deluxe we used for DDR3 testing is using a very early BIOS rev.

The following latency graphs are a little indulgent, so I won’t be offended if you skip them. They show access latencies across multiple block and step sizes, painting a fuller picture of memory controller performance with each chipset. I’ve arranged the graphs in order of highest latency to lowest. Yellow represents L1 cache, light orange is L2, and dark orange is main memory. These results are from our two-DIMM configurations.

As far as access latencies are concerned, Nvidia’s nForce memory controllers have a definite edge over Intel’s designs, whether it’s with DDR2 memory or DDR3. If you’re looking for differences between the P35 and P45, pay special attention to main memory with a step size of 128. There, the P35 has much higher access latencies than the P45, whose performance tracks more closely with that of Intel’s X48 chipset.

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.

While our DDR3-based P45 configuration scores well in Euler3D, its DDR2 counterpart doesn’t fare so well. Then again, neither does the DDR2-based nForce 780i, which was the fastest DDR2 setup in our memory subsystem tests.

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

DDR2 and DDR3 flavors of the P45 are evenly-matched in WorldBench, where they sit smack dab in the middle of the pack.

WorldBench’s multimedia editing and encoding tests don’t give the P45 much opportunity to stretch its legs. Scores are pretty close throughout.

The pack remains closely-bunched through WorldBench’s office and multitasking tests. The DDR3 P45 config is consistently faster than its DDR2 twin, but only by marginal, er, margins.

WorldBench’s 3dsmax rendering and DirectX modeling tests are largely gated by the system processor and graphics chip, respectively, so the chipset doesn’t have much influence here.

Given the tendency of Windows Vista’s intelligent disk caching scheme to affect WorldBench’s Nero test, I’m hesitant to make much of the results. There’s certainly no reason why the DDR3 P45 configuration should be so much slower in that test.

Gaming

A system’s graphics processor has much more impact on game performance than its core logic chipset, but the pack does spread out a bit in our gaming tests. Again, the P45 is consistently faster with DDR3 memory than it is with DDR2.

Serial ATA performance
The Serial ATA disk controller is one of the most important components of a modern core logic chipset, so we threw each platform a selection of I/O-intensive storage tests using a Western Digital Raptor WD1500ADFD.

IOMeter
We’ll begin our storage tests with IOMeter, which subjects our systems to increasing multi-user loads. Testing was restricted to IOMeter’s workstation and database test patterns, since those are more appropriate for desktop systems than the file or web server test patterns.

We’re using a slightly newer version of Intel’s AHCI driver for our P45 system, and that might explain why the chipset is just a little faster than the competition here.

The P45’s slight advantage is barely visible when we look at IOMeter response times, though.

CPU utilization is low across the board, with most of the chipsets staying well under 1%.

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

Burst and sustained read speed results are pretty tight, but HD Tach’s sustained write speed test teases a little separation from the pack. The P45 finds itself in the middle of the bunch there.

Random access times are all but identical, and CPU utilization doesn’t vary much from one chipset to the next. Keep in mind that HD Tach’s margin for error in the CPU utilization test is +/- 2%, too.

USB performance
Our USB transfer speed tests were conducted with a USB 2.0/Firewire external hard drive enclosure connected to a 7200RPM Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.

While its write speeds are still a couple of MB/s second sort of the standard set by Nvidia’s nForce offerings, the P45’s USB controller performs well. Write performance is much improved over the P35 chipset, at least.

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

PCI Express throughput looks pretty good, and with the lowest CPU utilization of the lot.

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

We’ve seen much better PCI throughput from our X38 and P35 motherboards in previous reviews, so I’m hesitant to make too much of these results. Something about the system configuration we used for this latest round of testing appears to be affecting PCI Gigabit Ethernet throughput with Intel chipsets, at least with this VIA-based NIC.

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

Power consumption
We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Power consumption was measured at idle and under a load consisting of a multi-threaded Cinebench 10 render running in parallel with the “rthdribl” high dynamic range lighting demo.

Gigabyte has rolled out a new version of its Dynamic Energy Saver (DES) power phase scaling software with its P45 motherboards, giving users the ability to run DES as a background service rather than a tray application. We’ve included results from the EP45-DQ6 with DES enabled to see how this latest power savings scheme works out.

Unfortunately, DES doesn’t do much for the DQ6’s power consumption at idle or under load. The DQ6 sucks quite a few watts overall, likely thanks to its wealth of integrated peripherals. Don’t take that as a knock on the P45’s power consumption, though. We put Gigabyte’s more sanely-equipped EP45-DS3R on the bench and it consumed just 97W at idle and only 171W under load.

On the whole, then, the P45’s power consumption actually looks quite good. The chipset is easily more power-efficient than Nvidia’s nForce offerings, and it’s more frugal than Intel’s flagship X48 Express, too.

Overclocking
For our overclocking tests, we dropped our CPU multiplier to 6X—its lowest possible value. The memory bus multiplier was also dropped to 2X to keep our DIMMs, which are rated for operation at up to 1066MHz, from becoming a limiting factor. Next, we turned our attention to the front-side bus, cranking it up and using a combined load of Prime95 and the rthdribl HDR lighting demo to test stability along the way.

Since we’ve already overclocked Asus’ P45 motherboards, we’ll be sticking to Gigabyte’s EP45s today.

The DQ6 had no problems scaling up to a 500MHz front-side bus without so much as a voltage tweak. Getting to 510MHz required feeding the north bridge an extra 0.1V, but 520MHz proved elusive regardless of what we tried.

Our DS3R was just as happy running a 500MHz front-side bus. No fuss. No drama. However, no amount of extra voltage, chipset cooling, or whispering sweet nothings would coax the system to post at 510MHz.

We’ve now had hands-on experience with four P45 Express boards, and each has coasted up to a 500MHz front-side bus speed without requiring additional voltage or cooling. Your mileage may vary, of course, but from here the P45 looks like the best Core 2 platform on the market if you’re looking for an easy overclock.

Motherboard peripheral performance
Core logic chipsets integrate a wealth of peripherals, but they don’t handle everything. Firewire and audio are farmed out to auxiliary chips, for example. Boards based on Intel chipsets tend to rely on third-party silicon for networking, as well.

To provide a closer look at the peripheral performance you can expect from the motherboards we’ve tested today, we’ve complied Firewire, Ethernet, and audio performance results below. You’ll notice that there isn’t much variance from one board to another, but there are a few things worth pointing out. Our P45-based motherboards are highlighted and in bold to make them easier to pick out from the crowd.

Realtek appears to have addressed the high CPU utilization of its previous GigE chips with the new RTL8111C, and that’s a good thing for the EP45-DQ6, which has four of ’em.

Firewire performance looks good for both our P45 boards.

Ahem. Yup.

Conclusions
It’s still early days for the P45 Express chipset, and based on our test results, motherboard makers may need some additional time to familiarize themselves with the chipset’s new memory controller and optimize their BIOSes accordingly. Even so, the P45 offers competitive application and peripheral performance that neatly keeps up with its P35 predecessor. But if the P45 isn’t any faster than the chipset that came before it, why bother?

PCI Express 2.0, for one. And not just for the faster signaling rate, but because the P45 can split its PCIe lanes evenly between a pair of graphics cards for CrossFire. Once written off as expensive gimmicks, multi-GPU configurations now deliver considerable value at affordable price points, making proper CrossFire support a rather attractive feather in the P45’s cap.

The P45’s die shrink to a 65nm process appears to be paying dividends, as well. The chipset’s power consumption is very low—provided your board isn’t burdened with excessive peripherals—and 500MHz front-side bus overclocks appear to be all but a lock. Now the P35 was hardly a slouch in the overclocking department, but it certainly didn’t overclock as easily or consistently as the P45 boards we’ve tested thus far.

In the end, then, the P45 looks like an incremental update to an already solid P35 foundation. It’s a worthwhile update, though, and one that should extend Intel’s mid-range chipset dynasty for another generation. Tick-tock indeed.

As for Gigabyte’s EP45 motherboards, we’re a little torn. The DQ6 is no doubt a solid offering, but at $260, it’s just too expensive unless you absolutely must have four Gigabit Ethernet ports or a couple of x4 slots that can accommodate longer expansion cards. The DS3R, however, looks like a fantastic option. You get everything you need, including dual GigE ports, Firewire, and a pair of x16 slots wrapped in a power-efficient package that overclocks like a champ and costs just $150. That’s a tough combo to beat, and one we’d heartily recommend.