Phenom was disappointing when it was finally released as well after all the hype that went into "monolithic quad-core FTW" mantra...but they recovered (mostly) after a couple years.
I'm less worried about larrabee living up to the hype and more inclined to waste my time contemplating the momentum/trajectory of any given player as we walk it out a few generations into the future.
The single-precision results combined with core-count is data, the gap to the competition is data, granted the confidence-limits result in wide whiskers on the perceptual plot of the trajectory and momentum of each player but the "big picture" is clearer today (for me anyways) than it was yesterday.
I don't expect anyone/everyone to share this view, and grilling Intel over the failure of reality meeting up to the hype is completely valid and godspeed to anyone who wants to do that, it just doesn't do much for me is all. I'm excited about where this goes from here.
I too believe that Intel will have the most luck with subsequent iterations of Larrabee, than the first release. Really what I was trying to get at was that saying they aren't that far behind and as such are advancing on a rate where they will pass the competition isn't a fair assessment. I am not sure that is what you were getting at, but it did appear that you were projecting this at some future time period. They may get there at some point, but the one data point that we have is by no means sufficient to predict when or even if this will happen. (Plus this data point is somewhat questionable.)
My main point against this line of thinking is that Intel (or any other company) can always look at ATI/nVidia/3DFX Etc. tackled the issues they are facing while creating their GPU designs to help them them find ways around their own issues as they come up. Once they catch up to the point the competition has advanced, they will no longer have this ready knowledge base and will need to find their own solutions to issues from scratch. This isn't to say they cannot do this, as they most certainly can, but the process will be exponentially slower, and there is no way to predict that they will work through new challenges any quicker/slower than their competitors.
The size of their R&D budget might help, but do you even know that it is actually bigger than the competitions budget? You can assume that because they have greater resources, but that does not mean that they would choose to spend it on something with as small of a return as a discrete GPU. Even if they do spend much more on R&D, it does not equate to faster breakthroughs. I work on programs that have more resources than Intel will ever have available, yet I see huge inefficiencies in the design phases to the point that the biggest project I was working on was just cancelled. Throwing money at the problems generally made them worse, and there are good reasons for that as well. That being said, if research is underfunded it is a major detriment, so if the competition cannot properly fund their research, you may have a point.