Dear Cirruslvr:
I do believe it, but like all things, it is subject to change as more information is gained. Now, do you believe anything you write? So far, my RAMBUS thinking has proven quite correct (its a LOSER), Intel is not yet finished with the 0.18u ramp and people (I am assuming that you agree with this given your reply) think it will start the 0.13u copper ramp with no glitches? They could sweep the glitches found in the 0.35u and 0.25u ramps under the carpet due to their being on top in performance, with AMD ahead, the harsh light of outside scrutiny is now showing that they too have glitches. Not much is now said about their "notch" process now. It did lead to a delay in the production ramp of their high MHz CPUs. The speed ramps have done to P3, what was done to K6, only the 1.13GHz fiasco has shown it for all to see. AMD spent 12 months getting copper right on 0.18u. Elmer ragged on AMD. Now Intel is faced with it. Their MO has been to run a pilot line (due to the contamination problem, it must be in a separate FAB and I do not think one is yet ready to begin testing the equipment) and then copy that line exactly in all the fabs to be converted (this is where some, including myself, believed the notch process was causing the 0.18u ramp to slow down but, Process Boy came up with another good reason, it takes a long time for the equipment makers to produce tens of copies of all the equipment (this still does not refute the first part because the resulting lines could not seem to produce the calculated (by those who attempted it) increase in usable CPUs)). Since the first part of the MO has not yet started, (and although Elmer ragged about the 12 months it took AMD, Process Boy did not think it that unusual (and he has the experience)), I take it that Intel, if it started such a line today, would estimate it would take 9 months to a year to get the first line into production worthy state. Then, it takes 12 weeks to make sure that it is production worthy (I guess this part could be cut short if, really necessary (and it could very well be)) and then the ramp up begins most likely by more lines in the same FAB (cu contam), proceeding to all other needed FABS. Given Intel's old MO on when a processor becomes shippable, they need at least a year to release a CPU (shipments over 100K/qtr) from start of testing. Thus Q4 2001 is the earliest production quantities of 0.13u copper are available for CPUs and that assumes that at least one production line is ready to begin testing now. Although, they could pull another paper launch (ie 1G P3) in Q2 and follow with volumes in Q4.
What information have you to refute the above?
Pete |
