I don't think the 20-stage Northwood design was bad at all. In fact, it gave AMD's K7 architecture quite a bit of trouble. People expected Intel to shrink this design and add enhancements, but that's when they came out with the 31-stage Prescott monstrocity. I don't think Grove had anything to do with that, and Barrett wouldn't know enough about processor micro-architecture to make an educated decision about it.
I disagree and here's why: If what you say is true then we must believe that there was a major disconnect between modeling and actual silicon. I don't think things were that far off. At least not so far off that many people didn't see it long beforehand. Prescott missed the performance target and over ran the power budget. There must have been people who saw this coming and it was their job to present their case to management who looks at the reasoning, their thinking process, their discipline and determines if it is sound, without needing expertise in the nuts and bolts. That's the CEOs job and those are the kind of stories one hears about Grove handing people their heads on a platter because of their process, not only the details. So the CEO doesn't need to know anything about processor micro-architecture but the discipline and the process by which decisions were made. Clearly that process was broken with Prescott.
Market Data 
Markets 
AI
