One thing that is starting to clear up is that Charlie's latter stories ended up being wrong, wrong, wrong. Dedicated physx hardware? Most likely not. Slower than pitcairn sometimes? Most likely not. Paper launch? Doesn't sound like that either. 350mm^2 die? Yeah no, 320mm^2 with heat spreader, 294mm^2 without.
Gotta read between the crayons with Groo the Wanderer's articles. He was right when he saw performance and blurted out "they win on every metric!!!". For the last years since he was at INquirer, He was correct about gt200 and gf100 on about a lot of things. Only things i got out of his recent drawings are that it was 100 was scrapped, 110 delayed, launch around april 2012, very fast new uarch. 7 and 8 unit skus. The rest was crap about samsung & apple and DOOM and GLOOM for Nvidia. No paper launch, Kepler rocks. Early sept '11, he wrote look to put Kepler in your computer April 2012, now 7 days before April 1, 2012, here it is. He obv knows whats going on at TSMC. it's the phsyx hardware part was funny, that was rumored to be a misdirection and he wrote an aritcle on it. I couldn't understand why hbe wrote "Oh it's sooo fast!!!" then the next day "could be slower than pitcarin" and stick to 7900 for max potential. Probably because he realized he hates Nvidia and everyone was going to read his new article saying Kepler is fast as sh!t and quit buying AMD 7000 cards, therefore derailing his number one objective: to destroy Nvidia's reputation. Break news and dis Nvidia at the same time proved difficult with a new efficient design, which is why his focus shifted to samsung, apple, notebook partnerships, tegra sucking, etc. Look for his future crap to be about GK110 conspiracies. Big kepler single die problems, yields, manufacturing changes and delays and a big story about the 104/110 developments unfolding bringing the 680 to market.