After weeks of nothing but bad news from the semiconductor market, Intel's investment announcement and roadmap updates on Tuesday were a welcome blast of fresh air. Company executives acknowledged the state of the economy, but took the opportunity to affirm the manufacturer's commitment to investment and innovation over the long term. In this case, Intel wasn't just talking—the company plans to accelerate the ramp of its 32nm process, bringing it to market more rapidly than originally anticipated. Roadmap goodness awaits.
Say good-bye to Havendale, Auburndale
The 32nm version of Nehalem that's coming down the pipe is codenamed Westmere; it's the "tick" of Intel's tick/tock process where "ticks" are die shrinks and "tocks" are new microarchitectures. Originally—and as recently as CES—Intel was planning to launch a series of 45nm mainstream desktop and mobile processors in the second half of 2009. On the desktop, we had Lynnfield (quad-core, octal threaded) and Havendale (dual-core, quad-threaded); the mobile parts were Clarksfield (quad-core, octal-threaded) and Auburndale (dual-core, quad-threaded, on-package integrated graphics).
Of those four, two of them—Havendale and Auburndale—are being "deprioritized," to use Intel's terminology.
The Extreme platform at the highest end of the desktop doesn't change much; the new information here is that Intel will, at some point, launch a 32nm "Gulftown" hexacore processor. Technically, Gulftown will still use the same X58 Express chipset; time will tell if that translates into compatibility with present X58 Express motherboards.
Lynnfield is still queued for a launch in the second half of 2009 but 45nm Havendale has been replaced by 32nm Clarkdale. If you had your eye on a dual-core Core i7 (with or without integrated GPU), this is actually something of a disappointment; Clarkdale and its mobile cousin Arrandale will enter production by the fourth quarter of 2009 but probably won't ship in volume until Q1 2010.
Ironically, Intel's decision to bring 32nm technology to market more quickly could leave some customers (and I'm betting a few Ars readers) grumbling. Anyone with an eye on Lynnfield (desktop quad-core) is safe; nothing here has changed. Similarly, if you want a Clarksfield-sized desktop-replacement sort-of notebook (quad-core mobile), you'll see a part to fit your needs at some point in the second half of 2009. If, however, you were gunning for a 45nm integrated GPU + dual-core in either a desktop or a laptop form factor, Intel's acceleration is most likely your delay. There's still a chance we might see 45nm dual-core Nehalem in desktop or mobile, but Intel believes its 32nm technology is strong enough to render the products unnecessary.
So how about them graphics?
One of the advantages to waiting for Westmere before launching an integrated CPU+GPU is that it gives Intel substantially more real estate to play with. According to the details the company shared today, the first "fusion" products will combine a 32nm dual-core processor with a 45nm GPU + integrated memory controller (IMC). You're reading correctly: Intel's design calls for the integration of GPU and memory controller; the CPU connects to this pairing via a single Quick Path Interconnect link. (No word yet on that link's speed or width). Intel has always discussed Nehalem as a combination of "core" and "uncore" components, with the memory controller in the latter category, so this move isn't the result of any technological voodoo. A discussion of how efficiently the CPU, GPU, and IMC are able to communicate will have to wait until we know more.
One of the points the manufacturer emphasized is that the same physical socket—presumably LGA1156—will support both quad-core and dual-core + graphics processors. The in-socket GPU will utilize up to two channels of DDR3 memory and is capable of switching to a discrete GPU solution. Upgrading the video card on a dual-core Core i3 (or whatever) should be no more difficult than upgrading the GPU on a current-generation integrated system.
The GPU itself will be a die-shrunk version of the company's GMA X4500 series of products. Intel wasn't providing or didn't have additional details on the part, but indicated we could expect in-line improvements commensurate with a die shrink and power consumption reduction. Based on Intel's past history, that likely translates into a higher clockspeed and maybe a new unified shader processor or two. Hopefully the company will make full HD decode support standard across all of its future fusion processors.
Server roadmap
Intel will lead with the Tylersburg-EP platform through 2009, accompanied by its Nehalem-EP processor. We'll see a Lynnfield-based Xeon offering in the Xeon 3000 series (Lynnfield is dual-channel, Nehalem-EP is triple), and a client-branded Clarkdale (dual-core) product once that CPU launches. In the second half of 2009, we'll see octal-core Nehalem-EX and its accompanying Boxboro platform. Both the Boxboro and Tylersburg platforms will be compatible with the upcoming 32nm processors. No word here on Foxhollow or Lynnfield, but Intel seems to plan to keep its desktop Nehalem chips on 45nm for a while after it transitions more power-sensitive segments.
Conclusion
Good news is especially welcome at a time like this. Intel's announcements and updates aren't going to rejuvenate the market, but the company's anticipation of a successful 32nm ramp and a strong future for its products are an important reminder that there will eventually be a light at the end of this particular chamber.
Some of you are undoubtedly none too happy about Intel's decision to delay its mobile Nehalem launch and there's no denying that the company's 32nm acceleration could also be seen as a tacit admission that it either needs more time to clear Penryn inventories or is concerned that demand won't materialize for these parts by the fall of 2009. No matter the reason, Intel does have a plausible technical argument for holding back; Clarkdale and Arrendale will likely be more attractive than any Havendale/Auburndale counterparts the company might have offered.
Intel can afford to take its time with Core i7. AMD's Shanghai, as great a step as it is for that company, largely competes with Penryn. So long as Core 2 Duo technology is viable across the market, Intel will be free to set its own timetable. In this case, that timetable should result in a better class of notebooks, desktops, and even energy-efficient servers. Not a bad trade.
Especially for novice/amateur use, as you seem to be "guided" to how the netbook can be used more efficiently. This is maybe not so fun for advanced users who want the flexibility to "hack" and play with the system more, but for a novice user, heck, this could be even better than e.g. Windows 7 or Ubuntu for netbooks. 
