Introduction
No sooner had I wrapped up my evaluation of AMD’s Phenom 9600 desktop processor than I got word that Intel was ready to start talking about its next-generation architecture. Though the Phenom delivers a reasonable value at its price, Intel is rocking the high-end right now. And while we don’t know much about what AMD has planned come 2009, it had better be good, because Intel’s forecast is aggressive.
At the same time, this is business as usual for Intel. The company is adhering to its architecture and silicon cadence—the tick and tock that enables Intel to move to a new manufacturing technology and then launch a fresh micro-architecture once manufacturing has been perfected. Intel’s 45nm Penryn family, represented on the desktop as the dual-core Wolfdale and quad-core Yorkfield chips, was the tick. Later this year, Intel will start production on the tock—an entirely new architecture currently called Nehalem.
Pat Gelsinger, Sr. vice president and general manager of Intel’s Digital Enterprise group, gave us a fairly in-depth look at the Nehalem die in anticipation of IDF Shanghai 2008. We also got an idea of how Intel will modularize the architecture to create a diversified product lineup targeting different price points. We even came away with information about the platform set to power high-end desktop and workstation Nehalem-based machines.
After he had covered Nehalem, Intel’s Gelsinger spent some time discussing Larrabee—a hush-hush graphics project expected in the 2009-2010 timeframe. Intel wasn’t ready to get into specifics about the chip’s architecture or specifications. In fact, we’ve learned more about Larrabee from reading Ars Techica’s
coverage of the
processor. However, there were a handful of interesting nuggets tossed around and an ultra-confident endorsement by Gelsinger. If all goes as planned, Larrabee stands a good chance of changing the GPU landscape for us gamers.
![Intel Discusses Nehalem, Larrabee, Dunnington [ Tukwila, the 2 billion transistor Itanium successor @ 366 x 544 ] > View Full-Size in another window.](images/01-s.jpg) Tukwila, the 2 billion transistor Itanium successor
|
|
On an unrelated, but still interesting note, Intel also introduced plans for its Tukwila and Dunnington designs. The former is an upcoming addition to the Itanium family. It’s a monster, made up of two billion transistors and comprised of four cores, 30MB of cache, and dual integrated memory controllers. Obviously, it won’t have any bearing on desktop gaming, but the die shot is incredible when you consider its complexity.
The latter, Dunnington, will be the final Penryn-based Xeon. Pat Gelsinger explained the company’s approach to Dunnington as a balance based on extensive workload analysis, which is why you’re seeing six cores and fairly large cache.
![Intel Discusses Nehalem, Larrabee, Dunnington [ Dunningtons eight cores and 16MB of cache @ 455 x 418 ] > View Full-Size in another window.](images/02-s.jpg) Dunningtons eight cores and 16MB of cache
|
|
“In this segment of the market, everything is already thread-enabled,” Gelsinger says. So, the company had the option to go quad-core with lots of cache or eight-core with less cache. Eight-core with lots of cache would be cost-prohibitive for a Xeon chip. The decision to implement six cores and 16MB of L3 cache yielded, in Intel’s collective mind, the best possible compromise. Dunnington is to be manufactured using Intel’s 45nm Hi-K node—good, since the design incorporates 1.9 billion transistors on a single die. To put that into perspective, the quad-core Yorkfield, which actually consists of two dual-core Wolfdales, sports 410 million transistors times two.
When Dunnington launches in the second half of 2008, Intel says it’ll be socket-compatible with the Caneland platforms already shipping. The quad-socket, dual independent front side bus chipset we know as the 7300 is decidedly enterprise-class, so don’t expect to see Dunnington make its way into gaming platforms any time soon.