For the last few days, I've been mulling over the top semiconductor stories of the year. Clearly, the settlement by Intel and AMD of their ongoing antitrust and patent/licensing disputes is the biggest business news. Picking the tech champ is tougher, because it's not about where we've been in 2009, but rather where we're headed in 2010.My sense is that 2009 has mostly been a transitional year, design wise. Both AMD and Intel pushed their architectures forward, and I don't want to give them impression that there weren't some bold advances, because that's not true. Intel unveiled Nehalem and AMD enhanced Istanbul.
As well, both companies took clear leadership in kicking the multicore calorie count up a notch. By this I mean that the quad cores, which have become the sweet spot of the industry, are soon going to give way to six- and eight-core designs. That's fine, and of a piece with the general trend of consolidation through proliferation.
I don't mean to be flip here; I just find it ironic that the impetus towards server consolidation has led to servers which themselves have so many more physical-plus-virtual processor instances than possibly even the data centers out of which they were originally consolidated.
And, as I've recently written, this will soon give rise to data centers on a box, backplane, chip, or some other tightly woven packaging, until it all tilts back in the other direction (possibly driven by the inability of software to properly exploit such densely linked hardware, or dashboards to decently visualize it). But I digress.
Anyway, the reason I'm so blase about what we saw in 2009 is because it's going to pale against what I think will be the most significant and exciting trend, namely integrated graphics. (Strictly speaking, integration is an umbrella phrase for true on-core integration as well external graphics which work alongside the traditional CPU as part of the platform.)
AMD's Fusion architecture, which is the big fruit of the scrappy semiconductor vendor's acquisition of ATI, is the big example to watch here.
Clearly, graphics architectures are not without their growing pains. I should actually flip that, and write that graphics architectures are all about growing pains. Fusion won't be here until around 2011. Intel's advanced graphics efforts evidenced the difficulty of productizing a true high-end design, with the delay of Larrabee as a standalone graphics chip.
On the other hand, but Fusion and Larrabee play at the high end. Integration is not quite so difficult at the low end, as evidenced by the improved graphics coming soon to Intel's "Pine Trail" Atom netbook processors.
OK, enough for one post. Let me know your take, by leaving a comment below or e-mailing me directly at [email protected].
Follow me on Twitter: (@awolfe58)
Like this blog? Subscribe to its RSS feed: (here)
Alex Wolfe is editor-in-chief of InformationWeek.com.
InformationWeek has published an in-depth report on the public cloud, digging into the gritty details of cloud computing services from a dozen vendors. Download the report here (registration required).