>But Paul, Elmer, and Wanna_BMW swear that AMD has been shipping .13 ever since Athlons first started shipping from Dresden, about 2 years ago - and that that's the only reason why Athlon has equaled or beaten PIII and P4 in performance for the past 3 years - though they claim that the first year of that was due to AMD's .25 process "actually" being a hybrid .25/.18 process.<
In a way, they are sort of right. Both AMD and Intel launched their 0.19-micron processes with some minimum feature sizes smaller than 0.13-micron. Furthermore, the CTX core of the K6-2 used 0.25-micron interconnects with 0.22-micron device layers. The K7 was not a hybrid process however. It used 0.22-micron for both device layers and interconnects.
Increasingly, what is important when talking about a process is the optics. Intel muddied the water a bit by using Numerical Designs mask technology with 248nm lasers and optics, then calling it a 130nm process. Sure this works, but it also means Intel has shot their wad with 130nm.
Consider this:
Intel originally planned on using 193nm lasers to get to 130nm. Because of the delays in the raw materials for the optics, they were forced to go with numerical design's phase shift mask technology using their old 248nm equipment to get to 130nm. The new 300mm wafer fabs will have 193nm lasers, and therefore be able to use Numerical Design's phase shift mask technology to reach 90nm.
AMD originally planned on getting to 130nm by using phase shifting optics on their 248nm lasers. The plan was to get to 130nm in Q1 2001. It turned out that some critical layers can not be produced satisfactorily using optical phase shifting, so they had to install 193nm lasers, thus creating a hybrid process for 130nm. Some mask steps will use 248nm lasers and optical phase shifting to produce 130nm design rules devices/lines while others will use 193nm lasers to produce 130nm design rules devices/lines. Now they plan on going to 90nm design rules using 193nm lasers and optical phase shifting. Why would they be any more successful using this technique for 90nm than for 130nm? Chances are that they will need 157nm lasers for some device layers at 90nm. However, there is a potential that they could, like Intel, be able to use phase shift masks for some layers. If AMD can successfully get to 90nm using phase shift masks and 193nm lasers for those layers that could not be shrunk to 130nm using phase shift optics and 248nm lasers, they may be able to get to 90nm without the need for any 157nm lasers. In this case, AMD might chose to go with an Intel style phase shift mask approach all the way around, or they might use some hybrid process whereby some layers are produced using optical phase shifting while others are produced using mask phase shifting technology. In either of the immediately preceding cases, all that would be needed is 193nm lasers all the way around. And, in either of these two cases AMD could begin the shift to 90nm in Q2 of 2003. But, if they have to use 157nm lasers for even one mask step, they will probably not be able to begin the transition to 90nm until late Q3 of 2003, and the transition will not be completed as early.
I think that, regardless of what AMD may claim, history is on my side. |
