Erm, no, the performance gains from HT are varying, it depends on the work load and what exactly it is that's being processed. Simple arithmetic can scale to 50% or more. You can only gain from threads if a program is multi-threaded for the available number of logical cores.
the Value of a 2500K has nothing to do with the validity of Hyper-threading. Programs simply don't scale well past 3 cores mostly.
What you're saying is that a PhenomII X6 has two useless cores because it's mostly the same performance as a Phenom II X4. :/
And what do you mean we cannot quantify mere nanoseconds? It's not theory. for every ms the Bulldozer is slower, it ends up being minutes and eventually hours if you leave it long enough. If I'm encoding a Blu-Ray disk, 2600K will be significantly faster than a 2500K, in the order of hours depending on the settings. It is the reason why GPU's are very fast at video encoding, it's simplistic fpu/simd instructions that can be threaded very well and it scales almost linearly with the number of available execution cores logical or physical.
BD isn't innovative, it's gone backwards. Narrower execution core, much deeper pipelines, higher latency caches, lower IPC as well. How is that in anyway like Netburst to Conroe? which had shorter pipelines, lower latency caches, larger cache, much higher IPC. Tell me how these two are anything alike? :/
IB is already a success, the talks you are reading where ever are untrue. IB is much faster than SB, in everything and if anything is true is that the Anandtech preview doesn't tell the whole story and IB is even more impressive.
There are no market expectations for IB. INTEL said, lower TDP, higher performance IGP, native PCIe support. They delivered on all of those. That it's faster than SB per clock is a bi-product of many other things.
