[Forwardthinking] Globalfoundries 14nm in 2014

Well this is good news for AMD. Still, I wonder if they mean mass production in 2014.

Many tiny things shall come of this. Maybe for once, AMD can do something right.

Amd has done plenty of things right in the past.

It's been a while. While they've made good choices recently, Intel's beaten them in pretty much every category, with the exception of GPU muscle. However, that lead will be disintegrating next year. Can't expect much in the area of drivers from Intel, but AMD needs to do something revolutionary. They're losing the evolutionary game.

But Global Foundries isn't even part of AMD anymore.





Whats with their timeline, 28nm in 2011? Are they really shipping products on 28nm? Cause I thought it was late 2012 and we were still waiting on volume 28nm.


Edit: Ah ha. Its seems the first 28nm was taped out "Q4 2011". So that timeline isn't talking about shipping products.

If we went off tape out dates, intel is already at 14nm in 2012.
Edited by erunion - 9/24/12 at 5:45pm

I have to agree. While I didn't jump on the bulldozer wagon, I'm enjoying my APU and think that AMD is heading in the right direction.

I think 7970s were launched in late dec 2011 with limited qualities. I think, or something like that.

lol wow, really puts things in perspective...so Intel's lead is really a good two years?

Globalfoundries is already converting the new fab8 in New York to 14nm. This means that they already have working samples. They plan on tapping out customer's chips in late 2013 with mass production being 2014 (they have already made process development kits available according to this site). Intel and GF will launch 14nm products at almost the same time.

Intel's lead is a somewhat debatable item anyway (especially in the long term as I wrote about below). This timescale (leapfrogging to launch 14nm at almost the same time as Intel) will hurt Intel significantly. Intel will be pushing out 22nm Atom processors (soon) to compete with ARM and MIPS, but they will probably be only marginally faster and about the same in performance per watt despite being on a smaller node (1/2 node smaller than 28nm). Intel has been relying on it's node advantage to compete with better architectures. Take this away and Atom will be crushed in both power and performance. This announcement says that Intel won't have the node advantage that it MUST have in order to compete in the very-low-power segment for long enough to win big. Despite the little coverage on OCN, this is huge.

Another thing to consider is after 10nm (further nodes won't offer full shrinks, only partial shrinks of the parts that are still a little bigger than one or two atoms in size). While Intel is putting some resources into replacing silicon (this must be done within a few months of 10nm else the bottom will drop out of Intel stocks as they can no longer ensure their "Moore's Law" business model), the majority of investments into the long-term future of fabrication (carbon nanotubes, memristors, optics, graphene, etc) is being done by the joint venture that is HP, GF, IBM, samsung, etc. With a long line of IP protecting them, Intel will have quite a battle (even now) to enter the market without paying loads of cash to license patents. Even if the license the technology, I doubt that Intel has what it takes to catch up with the R&D that's currently being done.
Edited by hajile - 9/24/12 at 7:42pm