I do have to admit that Intel isn't doing anything "bad" by designing the P4 as they did, with a lower instruction per clock. The same thing has been done by both Intel and AMD every time they change designs. They have to do it in order to allow the chips to reach higher clock speeds. The problem is that without a tremendous boost in clock speed, the first batches will always end up being slower than the previous design. They can't just make the first chips slightly faster than the most recent of the old design (well they CAN, but shouldn't), because the increase in clock speed isn't enough to make up for the reduced IPC. Essentially, all new chip designs will be slower than the previous design if you dropped them both back to the same clock speed (excluding optimized processes like the P4 with MPEG encoding), but the newer design is able to reach such significantly higher clock speeds that they're able to far outstrip the older design, once you reach the "turnover point", where the higher speed makes up for the lower IPC. For the P4, it seems that there's a very large difference in clock speed needed before the P4 outdoes the P3 or Athlon for most applications.