It seems that NV is fully taking advantage of this period of bad press about intel to cement their place as (at least perceived) leader in GPGPU areas:
http://www.xbitlabs.com/news/other/...in_Heterogeneous_Multi_Core_Technologies.html
David Kirk: nVidia corporate fellow said:
We find that most problems, if not all, are a mix of serial control tasks and parallel data and computation tasks. This is why we believe in heterogeneous parallel computing - both [parallel and serial] are needed. CPUs are commodity technology and there are multiple CPU vendors that we work with. In my opinion, Intel has no particular advantage in developing a hybrid system - in fact, they have had little success historically in designing either parallel machines or programming environments
I think in the highlighted parts you see a general technical strategy why nv thinks it has a serious leg up on intel. The necessity of heterogeneous computing at some point in the future makes it a necessity that each hardware vendor has access to mature and high-performance technology on both superscalar OoO microprocessors, as well as SPDM streaming microprocessors.
The way that intel has been forcing progressive lock-in on vendors will not make it easy for their partners to adopt these technology vectors. This is primarily done in three ways, (1) through some financial or logistic incentives that give OEMs an easier road when locked-in on the intel platform; (2) through actual production of microprocessor packages that are closer to SoC designs and progressively moving co-processor functionalities on die and use the current dominance of their general purpose uprocessor to force vendors to choose; and (3), refusing to license any of its own technology that eases the integration of third party chips into its own platform (QPI, DMI, etc).
So what nv has been saying is that #1 is basically illegal, has been proven illegal in the past, and is continuing to be deemed illegal by just about everyone except intel and its partners that benefited from them (Dell execs are again in hot water this past couple of weeks for their past deals with intel, see:
http://bits.blogs.nytimes.com/2010/07/23/dells-trouble-kicking-the-intel-habit/ ).
And in regard to #2, if you believe nv's PR campaign, it doesn't really hold water either, because in its view, Intel is basically missing half of the necessary ingredients to make heterogeneous computing work. Larrabee has had serious issues with power consumption, its coherence modes, driver issues with software kernels fully utilizing vector pipeline throughput, as well as inherent inefficiency of requiring an x86 front-end to decode into RISCy mmicroinstructions. And nv probably thinks that it is legally dubious anyways, given intel's dominant position in commodity MPUs in the world, to force vendor lock-ins with this method (sort of like, but not exactly like, IE being preloded on every windows machine). But NV doesn't have a solid legal argument in the US at the moment; nor does it think it needs to, since it thinks that it wins on technological ground.
For #3, it is the most uncertain where NV is going with on its pursuits. Perhaps that's where they have failed to come to a deal so far. What exactly is NV looking for in regard to licensing of technologies from intel; what would FTC have to do to satisfy NV, would a simple small royalty DMI license make nv happy, or are they really shooting for high-end/server playing field; or perhaps even more so do they really want a piece of the x86 pie some time down the road to ensure that they are not shut out by vendor lock-ins in the future.
We will have a lot better idea on these questions in about 10 days or so.