Elmer, beamer has it right. It's a shame that the Netburst legacy will be defined by Prescott, because things really were humming along under Northwood.
Prescott was a disaster. More than anybody outside of that project will ever know. Maybe in another decade I'll share some stories. Did anybody see that disaster coming? Well, yes and no. Anybody working in the trenches knew that the schedule was a joke (a brand new design on a brand new process using brand new aggressive LVS circuit topologies and brand new cell-based design methodology, all on a compaction schedule!), and the result was that any effort on reducing power was ZBB'd to get the damn thing taped out.
What came as a surprise was how much the part's frequency was limited by power. Modelling power before you have silicon wasn't much more than a guessing game at that point. The architectural improvements actually did their job, as the IPC was comparable to Northwood despite a deeper pipeline. Performance was supposed to come from higher frequency, but no extra frequency was gained because of the power limitations (not to mention the lack of schedule to push speedpaths as hard as they could).
If more sane decisions had been made about Prescott's scope and schedule back during its planning and inception in the 2000/2001 timeframe, the general opinion about Netburst might very well be much different today. While I don't blame Barrett for Prescott specifically (I reserve that blame for the division VP at the time, who was thankfully "retired" during the recent manager cut), I do blame him for taking his eye off the ball, and taking for granted all of the risks that must be assessed before approving a project as complex as a bleeding-edge processor design.
The good news is that things feel like they're back to normal, and the general mood is more like it was back in the Wmt/Nwd days. The project goals and objectives certainly feel more sane than they ever did on Prescott, which is a very good thing.
I have a friend who's been at Intel for ~24 years now in one of the research divisions, and for a while reported to a member of the senior staff. I once asked him over lunch how he felt about Barret as CEO, and he flat-out told me that Barret's and Otellini's roles needed to be reversed (Otellini was COO at the time). He said that Barret is a great operations guy, and always has been, whereas Otellini had more vision and fresh ideas for the industry. Personally, I'm much more comfortable with Intel's current leadership, both at the corporate and VP level.
AI
