The key term here is "up to", which means that it could be a 0.000001 percent increase in performance and the "up to" would still be considered valid. There is always a lot of speculation out there about the performance of yet-to-be-released products, some of it accurate, and most of it is hype or FUD. Until Intel or AMD release their next generation parts, it is almost impossible to know what the performance will be.
The whole AMD K10 is the perfect example of this. The initial reports were indicating that AMD was in trouble because they couldn't get their next generation chips up in clock speed, and that 2.0-2.4GHz was the range we could expect. Now that the B2 steppings have been shown to get up to 3.0GHz and beyond, things look a bit better for AMD than many thought.
Just because pre-release parts don't look fast means nothing, and just because Intel hypes a product doesn't mean it will really be much faster. Intel has a LOT of room for overclocking the current processors, so even on current chips there is a lot of room for Intel to boost the speeds.
The article does indicate that there will be more cache on the new chips, which would account for a boost in performance, even if nothing else has changed. Without an integrated memory controller, having a large cache will boost performance by quite a bit. Extra cache isn't what I consider to be a real improvement to the design of a processor though.