Re: Copper interconnects have not been necessary for Intel's .18u manufacturing process to reach high frequencies.
No? Well then, let's take a peek at what was necessary:
Intel's 0.18-micron process technology normally produces transistors with a 0.13-micron gate length, but in his presentation Ghani said a method had been devised that allowed a notch to be introduced at the bottom of each transistor's polysilicon gate, thereby narrowing the gate length to 100 nm.
eet.com
Although a few tricks remain at 0.13 micron to counteract some of these problems, at 0.10 micron and below they become unmanageable. Intel, for example, in a paper presented at last December's IEDM, described a notched-poly technique that undercuts the gate poly to reduce overlap capacitance. IBM says it shies away from such stopgap solutions, however, because they do not scale well to shorter channel lengths. IBM says that even at 0.18 micron, notched poly is more trouble than it's worth. The problem is that ultraprecise control over the etch is required to achieve consistent gate lengths, but such precise control is difficult because of factors such as the proximity of other structures, which create unavoidable local variations in the effectiveness of the etch.
mdronline.com
The second article, in particular, has a nice overview of many of the current controversies.
Remember that AMD didn't suffer a yield crash in late 1999 [note IBM's comments about "not worth it" and "ultraprecise control required"] - while Intel, driven to use a blind etching to go beyond the limits of its .18 lithography process, and clearly lacking the necessary "ultraprecise control" did suffer such a yield crash.
Somebody already used up a good part of the benefit of moving to .13 while still on .18, but maybe you should read up a little more on "notched gates" before you decide which of the two main Windows compatible processors that description is best applied to. |
