RAM | Reseller Advocate Magazine: Issue 70

By Chris Angelini

Toward the end of the NetBurst microarchitecture’s life cycle, Intel was second place in the processor performance game. Nevertheless, its platforms were still rock solid. Business customers continued buying Intel because they perceived the value in going with a CPU matched up to a chipset on a like-branded motherboard. The stability, support, and, to a lesser extent, the speed were enough to cement repeat business in the face of AMD’s more elegant memory architecture and use of HyperTransport technology.

Now AMD is struggling to catch up to Intel, and the company is clearly aiming to duplicate Intel’s successes as a platform vendor rather than just a purveyor of processors. By acquiring ATI, AMD put itself on a trajectory to offer CPUs, chipsets, and graphics cards beneath one marquee. AMD will be unveiling its next-generation server and desktop chips within the next couple of months, setting the stage for a bitter battle in 2008.

Of course, Intel isn’t sitting around waiting for AMD to bring the fight to its doorstep. Early indications suggest that Intel’s first 45nm CPUs will surface late in 2007 for servers and workstations, with the desktop chips to follow early in 2008. Expect your customers to want these processors. The smaller manufacturing process is making way for more complexity and additional performance.

All of the updates made to Intel’s Core microarchitecture are wrapped up in the Penryn generation of CPUs, which will cover enterprise, desktop, and mobile chips. All three Penryn families will generally feature enhanced Intel Virtualization Technology, larger shared L2 caches than found on today’s SKUs, and SSE4 instructions. You can also expect faster clock frequencies and lower temperatures. Penryn-based Xeons will employ bus speeds as high as 1600 MHz, while the mobile Penryn parts conserve battery life through a deeper idle state.

When the 45nm Xeons drop, they’ll need updated core logic. The same goes for Intel’s Penryn-based mobile parts. Both of those platforms are currently in the works. For now, Intel is focusing on the desktop, where updated chipsets not only extend support for tomorrow’s 45nm CPUs but also build on some of the most popular motherboards you’re selling today.

Courting The Enthusiast

Most motherboard vendors recognize the enthusiast’s influence over mainstream buyers and the fact that the king of the hill can pull features from pricey flagships to create more budget-oriented derivatives. For a long time, Intel ignored the upper echelon of power users. While everyone else was adding adjustable voltage settings and front-side bus tweaks, Intel refused to sanction operating its own processors outside of spec.

MEMORY OF THE FUTURE
As DDR3 memory comes down in price, it’ll become a much more popular option for enthusiasts looking to ratchet up performance while cutting back on power consumption.

Around the same time Intel started doing battle against AMD’s Athlon 64 FX with its own Pentium Extreme Edition, it also began enabling advanced BIOS options in its highest-end motherboard models. Suddenly, you could use the same boards popularized by Intel and juice them with a little bit of silicon nitrous. Intel’s minor tweaks slowly evolved into more aggressive options. In fact, the current D975XBX2 rivals the best Intel’s competitors have to offer.

Charged with the task of one-upping its own best effort and facing an impending 45nm processor family launch, Intel is putting a big emphasis on its sexiest desktop chipset, the X38. Past flagships included an extra bus speed setting here and a faster memory timing there. The X38 actually adds a handful notable features that simply won’t be available anywhere else. Some of the extras, such as support for Core 2 Extreme overclocking enhancements, won’t be a big deal to bargain hunters who wouldn’t pay $1,000 for a quad-core powerhouse anyway. But architectural advancements like DDR3 and PCI Express 2.0 represent the future of core logic, so picking up on them early gives your customers the inside track on performance and future upgradeability.

Why would a power user even care about an updated peripheral interconnect? We’re not exactly pushing the boundaries of what a PCI Express x16 slot can do today. But if you’ve paid attention to the last couple of launches from AMD and NVIDIA, you probably picked up on language suggesting both vendors want their extremely parallel processors working on more than just 3D graphics. Moving more data over PCI Express for general purpose computing is going to require a wider pipe.

The PCI Express 2.0 spec introduced by Intel’s X38 effectively doubles the throughput of the interface as we know it. Each lane jumps from 2.5 Gbps of bi-directional bandwidth to 5 Gbps. PCI Express 2.0 slots also support cards with higher power requirements. So instead of connecting two auxiliary power connectors to your GeForce 8800 GTX cards, you’ll be able to run adapter-free or with a single power input. Best of all, 2.0 is backwards compatible with existing graphics cards, meaning you can expose the upgrade path without requiring customers to buy new hardware.

Shifting to a DDR3 memory subsystem isn’t as forgiving. The modules aren’t pin-compatible with DDR2, though both employ a 240-pin layout. As a result, customers buying boards based on Intel’s X38 will need new memory modules to go with their new platform. Where’s the value proposition? DDR3 runs at higher clock frequencies, potentially offering more bandwidth. The flip-side is that existing DDR3 products suffer less aggressive timings, pulling performance in the other direction as well. Pros and cons aside, DDR3 does offer better performance on a lower power budget and will become much more affordable in time.

Intel plans to launch the chipset by early October, according to Todd Garrigues, Intel’s North American channel marketing manager, with all of the company’s usual board partners shipping X38-based platforms. However, Intel’s own X38 motherboard is being held back until the 45nm desktop chips actually ship. Let’s just say it supports an unexpected front-side bus configuration beyond what Intel’s current Core 2 chips support. So while any of the current 65nm CPUs would run just fine, Intel wants to ship the board alongside a CPU capable of exploiting its fastest settings.

Managing The Mainstream

Just below the die-hard enthusiast eager to shell out a few extra dollars for bleeding edge hardware, there is a group able to appreciate the latest and greatest but unwilling to pay first-run prices. That’s where Intel’s P35 chipset comes into play, offering the same compromise between cost and functionality you would have found in last generation’s P965 platform.

NEW PROCESSORS CALL FOR NEW CHIPSETS
The move to 45nm means more advanced Core 2 processors with larger caches, new multimedia instructions, and invariably more performance.

When it comes right down to it though, the compromise is really quite minimal. Two of the chipset’s advancements are a faster front-side bus and official DDR3 memory support, which work together to improve system bandwidth. The 1333 MHz bus opens a 10.6 GBps path between Intel’s Core 2 Duo or Core 2 Quad CPUs, while two channels of DDR3 running at 1066 MHz combine to achieve 17 GBps. Compare that to the P965 and you’re looking at a lot more headroom to accommodate CPUs sporting four execution cores. Of course, the P35 chipset itself supports DDR2 as well, which most motherboard vendors are currently using to help keep prices down. Two channels of DDR2 800 push up 12.8 GBps—enough to keep pace with that new 1333 MHz frontside bus setting.

There’s no PCI Express 2.0. But the P35 does feature one standard PCI Express x16 link and six PCI Express x1 links branching off from the ICH9 controller hub. Other improvements to the ICH9 include a max of 12 USB 2.0 ports, six Serial ATA ports (including eSATA), Gigabit Ethernet, and High Definition Audio. Intel’s Matrix Storage Technology is available on motherboards with the ICH9/9DO/9DH, and the ICH9DH is one of your keys to Viiv compatibility.

Value In Integration

Although the P35 is a much more reasonable buy for mainstream customers, it still requires discrete graphics. SMBs who’d as soon pass on an entry-level DX10 card will instead want a motherboard based on Intel’s G33 chipset. The G33 and P35 are similar in every way except that the G33 sports an integrated graphics engine.

Onboard video is usually written off as more of a performance liability than anything. However, Intel’s GMA 3100 boasts enough horsepower to accelerate DirectX 9 software, including Windows Vista. The core isn’t Intel’s best. In fact, its specifications are nearly identical to the GMA 950, part of the aging 945G core logic. Software pixel shaders, software vertex shaders, four pixel pipelines, and Intel’s Clear Video processing technology at least lend ample speed to mainstream apps.

Intel’s DG33TL micro-ATX motherboard makes the most out of the G33’s list of features, including its integrated graphics. Several generations of Intel GMAs have quietly supported dual independent displays. In the past, you would have needed to install an ADD2 card for a second output at a cost of $25 or so. The DG33TL does away with the ADD2 option by integrating DVI outright. “Hopefully,” says Intel’s Garrigues, “the second output allows VARs to sell an extra LCD panel as well.”

There’s one more notable 3 Series integrated chipset: the G35. Though you’d assume Intel’s G35 would sit a notch higher than its G33, both chipsets offer unique features. The G35 boasts a more advanced graphics core, for example. DirectX 10 and OpenGL 2.0 support are both standard, as is a souped-up version of Clear Video able to accelerate VC-1 decoding. The G35 does support Intel’s upcoming 45nm desktop CPUs, but it doesn’t work with DDR3 memory modules. The G35 also employs Intel’s ICH8 instead of the more modern ICH9. While differences between the G35 and G33 are subtle, they’re notable enough to cause the uninitiated some confusion.

A LITTLE SOMETHING DIFFERENT
Perhaps you’re not used to seeing VGA and DVI outputs on the back of a motherboard. And where are the legacy serial connectors? Intel’s DQ35JO does away with old technology in favor of lots of USB, Gigabit, FireWire, and two display outputs.

The 3 Series Means Business

According to Intel, its DQ965GF motherboard is a best-seller. The business-oriented Q965 chipset it centers on contributes the right combination of enthusiast performance and SMB value. Full Core 2 Quad and Core 2 Duo processor support, up to 8GB of DDR2 memory, the GMA 3000 graphics engine, Gigabit Ethernet, and vPro compliance dovetail nicely together.

Resellers should expect the Q965’s successor to attract even more attention. Intel’s Q35 chipset incorporates the same general spec sheet as the G35, meaning a 1333 MHz FSB, DDR2 800 memory support, and a connection to the ICH9 hub. Instead of the G35’s DX10 graphics core, however, Q35 chipsets leverage the DX9-based GMA 3100 core. Little extras like dual independent display outputs, more SATA connectivity, eSATA, and up to 12 USB ports on top of the Q965’s value.

The first motherboard from Intel based on Q35 takes the chipset a step further. Passively cooled and completely silent, the DQ35JO adds the security of a Trusted Platform Module, an Intel 82566DM Gigabit Ethernet controller with AMT 3.0, and a compact microATX form factor.

All Aboard

We can’t blame you if you’ve struggled to keep up with Intel’s chipset refresh. The platforms are emerging in waves, each with a slightly derivative list of features. Perhaps the easiest way to keep the 3 Series straight is through each chipset’s letter designation. The “X” line is performance-oriented—the best of the best. The “P” parts are discrete-only mainstream platforms. All three “G” series chipsets incorporate some form of integrated graphics, each differing slightly from the others. Finally, look to the “Q” family for business-oriented features. If you’re still in doubt, check out the sidebar for a short description of the major 3 Series models and their key points.