What might Westmere mean for AMD?

Now that both AMD and Intel have laid down their roadmaps for 2009 and 2010, we can have a little fun and start speculating about how the two competitors might match up. AMD has made it clear that it will stick with its 45nm Deneb core—the one that powers current Phenom IIs—through 2010. As for Intel, it's planning two waves of Core i7 derivatives: 45nm quad-core parts later this year (Lynnfield and Clarksfield) followed by 32nm dual-core CPUs (Clarkdale and Arrandale) in late 2009 or early 2010.

If the flurry of code names leaves you scratching your head, just check out the official roadmaps below:

What does this mean for AMD? Based on how current chips perform and what Intel has disclosed about its future 32nm processors, we can make some educated guesses. Here's how I predict things are going to go down on the desktop.

Right now, AMD's Phenom IIs are essentially competitive with Intel's 45nm Core 2 Quads. When Lynnfield comes out later this year with a Core i7-derived architecture, four cores, and eight threads, we can probably count on it outperforming the Phenom II overall—even if it's not quite as fast as the Core i7 clock for clock. AMD could foreseeably compensate by raising clock speeds, but I expect the firm is likelier to simply cut prices and offer slightly lower performance at a discount.

So, what happens when 32nm Clarkdale processors roll out, presumably early next year? With two cores, four threads, and a built-in graphics core, Clarkdale will likely inhabit a lower price bracket than quad-core/eight-thread Lynnfield and Core i7 offerings. And, if my first assumption is right, that would pit it right against the Phenom II.

I think we might very well end up with a close race between the Phenom II and Clarkdale. AMD's processors will have more cores, but the Intel chips should have better clock-for-clock performance and the ability to scale to higher clock speeds. Just look at current benchmarks: in Worldbench, the 3.33GHz Core 2 Duo E8600 is actually faster than the 3GHz Phenom II X4 940 overall. In several media encoding tests, a single 3.2GHz Core i7 manages to outrun dual Core 2 Quads running at the same speed.

I really wouldn't be surprised to see Clarkdale keep up with or even outperform quad-core Phenom IIs overall, with the AMD chips pulling ahead in heavily multi-threaded apps (like, say, 3D rendering). Assuming AMD can refine its 45nm process and keep power consumption low enough, that might not be such a bad matchup—and it certainly wouldn't be as bad as pitting Phenom IIs against 32nm Core i7 derivatives with four cores. Intel could clip AMD's margins by offering Clarkdale processors at bargain-basement prices, though, which shouldn't be hard with two small dies. (The cost of producing a big die like the Phenom II's is greater.) Also, no matter how much AMD refines its process, Intel should have a significant power efficiency advantage.

What about notebooks? I won't make detailed predictions here, because I'm not quite as familiar with the mobile end of the CPU market, but we know several things for sure: on the Intel front, Lynnfield and Clarkdale will both have mobile siblings. On the AMD side, a dual-core Turion X2 successor code-named Caspian will roll out this year, followed by a quad-core Champlain chip in 2010. I doubt shoving a 45nm quad-core processor into a notebook will be all that practical even next year, though, so I think AMD may choose to focus more on Conesus and Geneva—both dual-core chips for affordable and ultraportable notebooks.

We've already caught a glimpse of Conesus' potential in the form of the Athlon Neo, which is that chip's single-core little brother. The Athlon Neo powers HP's Pavilion dv2, a $699 notebook that weighs just 3.8 pounds and features a 12.1" 1280x800 display, Mobility Radeon HD integrated graphics, and Windows Vista. In many ways, that seems like the sweet spot between netbooks and full-featured laptops.

Intel will undoubtedly attempt to fill that same niche with its own batch of processors, but I think AMD will be in a uniquely privileged position here thanks to its integrated graphics. Even Clarkdale's mobile sibling, Arrandale, will have an IGP with the same lackluster underpinnings as today's GMA X4500. AMD should have no trouble pitching slower CPUs with faster graphics in a market where CPU performance really isn't that big a concern.

All in all, AMD will probably be in a very tough position until it introduces its next-gen Bulldozer architecture in 2011, but it seems like the company could manage to get by in the lower end of the market even against 32nm Nehalem derivatives (assuming it survives the current economic crisis, of course). Bulldozer will really need to deliver, though, because it should come out right as Intel hits its next "tock" with Sandy Bridge, a microarchitectural refresh based on 32nm technology.

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.