Both ATI and NV said at the beginning of the year they would aim to refresh, for ATI it was "second half" and for NV it was "late 2010".
NV also said (before GF100 launched) that other products weren't delayed due to that delay.
One problem with all this is that NV parts have been "delayed" compared to where early rumours situated their release, so it might not be too surprising if the Fermi refresh also slipped a bit (maybe to 2011).
Part of the problem I have with the idea of NV suddenly massively increasing performance though, is the fact that they don't have much chance to make a bigger die, so they will have to work by making their architecture more efficient in terms of 'speed per transistor'.
ATI seem to have managed that, but we knew they had an often 'inefficient' architecture in many ways, and significant weakness in many areas of DX11, so it will be interesting to see how much of an increase NV can get without increasing die sizes (since their is already very large, and they might not want to go too much bigger).
One thing in their 'favour' is that their current shaders seem much weaker in many situations than the GT200 shaders (e.g. GTS250 with 128 shaders beats 192 shader GTS450 in many situations), so maybe they will be able to make some tweaks ot make it a more effective architecture, but who knows.
