At the same AMD press update, director of manufacturing technology Tom Sonderman hit on several myths that are circulating around AMD's capacity to ship its quad core Barcelona chip in volume to OEMs on time (this summer) and on the company's transition to 300mm wafer and 65nm process technology. Sonderman also pointed out that at the end of 2006, AMD had 20% share of the server market with only one 200mm fab versus four advanced 300mm fabs for Intel. Quite impressive I must say. And it just bodes well for the future as AMD will finish the year with three more fabs instead of just one.

"We don't just believe that you have to go out and build fabs to be successful. We believe that you have to maximise the investments that you put in place, leverage everything you can out of these investments... We now migrate to 300mm manufacturing and also go from being a single fab company to a multiple fabs company... It's important because basically it costs the same amount of money to build a fab, whether it's AMD or Intel".

So here are some of the myths Sonderman was eager to quelch (by the way, they are all wrong... of course):

Myth #1: AMD is behind on its 65nm transition. By year end, "everything AMD produces will be on 65nm technology". And that includes, AMD's own Dresden, Germany, fab, IBM Microelectronics and Chartered Semiconductor Manufacturing.
Myth #2: Quad core is not manufacturable. For Sonderman, everything is on track to deliver Barcelona in volume this summer. OEMs machines from Dell, HP and IBM are expected to ship in the fall.
Myth #3: AMD is 18 months behind Intel on 45nm. Just 6 months behind said Sonderman. AMD is already running 45nm pilot lines in its Germany plant (they showed a 45nm wafer during the press briefing) and expect to introduce 45nm in summer 2008, 18 months after AMD introduced 65nm. AMD seems to stick to an 18 months cycle for its processing transitions. Will see if they can keep that pace with 32nm, 22nm, etc.
Myth #4: Immersion lithography is too risky at 45nm. AMD is getting equivalent yield with immersion lithography than their current dry lithography. Using immersion lithography is fundamentally more efficient than Intel's approach (double exposure) which increases the number of masking layers, the cycle and the cost of the products because it's taking many more process steps to produce the same result, said Sonderman. He expects the rest of the industry to follow AMD steps at the 32nm generation.