Otellini answered a question from an Intel staffer who asked whether the company's strategy on dual core desktop design was as well thought out as AMD's.
While Otellini wouldn't give specific details of Intel's architecture, he said his firm's strategy is on optimising the designs it has.
AMD's design, he said, uses an integrated memory controller on single core chips and uses hybrid transport. That will lead to a problem because AMD doesn't have shared caches, but separate memory caches.
Intel's approach, it appears, will exploit shared cache architecture which gives better performance and doesn't tie the chip giant to just one generation of memory. Intel didn't want to have to be hogbound by having to change the CPU every time memory architecture changed.
That may be true for desktop dual cores, but it's not necessarily so for server architectures.
Intel, he claimed, has had multicore and dual core products in its internal roadmaps for a fair while, and believes that it will be able to deliver higher performance using cooler devices.
The lack of benchmarks for multicore processors means Intel is working on desktop client benchmarks, perhaps similar to its Centrino Mobile Mark. The benchmarks are already there for server chips.
Otellini said in answer to another question that Opteron isn't touching the Itanium market and won't in the foreseeable future. So Intel will beef up its Xeons, although he admitted that the strategy for the next 12-18 months is defensive.
By early 2006, Intel will be ready to go on the offensive again. He said Intel doesn't accept AMD's assessment that it can capture 30 per cent of the server market by 2007.
He described the next 12 months as being a "classic street fight", before its new family of server processors take the battle back to AMD. µ