"4 different teams at Intel said it wasn't even economically feasible"
You have to put that into it's historical context. Those statements were made, and the decision to go with what eventually became IPF were made at a time when it was clear that conventional, in order execution machines were hitting a way. The 21164 was supposed to scream through over 1 billion instructions a second, but it actually was averaging a whole lot less. Sure, there was that out of order processing stuff that people were talking about, but it wasn't clear that would actually do the trick. Why extend IA-32 if it was going to hit a wall? Even if it did, it was going to take transistors and it wasn't clear that they could economically make them a whole lot smaller. Better to go with a clean sheet design that could take advantage of all those new compiler techniques that is all the buzz...
However, as it turned out, OOOP actually did work. It took a lot of transistors, but that wasn't a totally intractable problem. OTOH, those new compiler techniques helped out the old architectures too...
News 
Market Data 
Discover 
