[H] are saying it also, and its pretty much from TSMC themselves. Best case scenario is looking to be Feb/March for real supply, there might be hot lots available in minimal numbers a month or two before then IF the hot lots work.
Hot lots are highly risky waifers they put through the same time as the first batch of waifers. IE they put in their latest tweaked design, rather than make 10 waifers and test them to death and if they work fine, then go into final production. You make 100, or 1000 waifers instead, pray to every single god you can think of that they don't need a new revision, don't need a respin and retaping out and simply work. If that works, then you get 1000 waifers instantly rather than waiting 2 months. If they don't work you've spunked maybe 100 waifers, at $5k a waifer, which is $5million , down the drain. Hence its a huge risk and we can't really be sure they've done it.
THe chances they've risk anymore than that is immensly small, meaning even if they work great, the "hot lots" will be tiny and will in no way be enough for a realistic hard launch, meaning best case is a maybe 10k cards late January before mass production starts seeing final parts a couple months later.
Its looking like Jan 6th is the next "big" publicity event by Nvidia now, so its almost impossible they will launch (hard or paper) before then.
Its really boring when people doubt Charlie, he's not been wrong on anything Nvidia for two years, calling most of whats happened a year before it was confirmed, chipsets, bumpgate, the constant delays in the GT200b shrinks, etc, etc.
Production is simple, you can't speed it up, you can't change how TSMC works because you want to, and theres plenty of people at TSMC who want people to know what Nvidia have screwed up themselves rather than TSMC, because otherwise Nvidia blame TSMC and TSMC look bad. When you know the dates of tapeouts, you know when production starts, when a chip goes in to be made you can call the tapeout dates, as again Charlie did to within days for the A1 and A2 tapeouts.
You only need to look on his forums to see links to his past AMD bashing, over the 2900xt, and plenty of CPU stuff.
Nvidia fanboys convince themselves he's pro ATi and anti Nvidia, despite the fact he likes Intel a lot, thinks AMD chips aren't that great and when ATi were having a bad time he was mocking them mercilessly aswell. He'll laugh at the next company to screw up aswell, hardly his fault Nvidia have messed up so much lately
The other latest news is TSMC are said to be taking orders from ATi now, for production at 32nm, in Q1 next year. It should be an easier transition(ATi from 65 to 55nm) being a half node, but I can't see the refresh launching in Q1, but likely sometime in Q2 we'll have 5XXX parts, at 32nm, within a couple months of Fermi's real launch.
