I'm not saying I believe everything the road map saids as a certainty. Of course not, that would defy the meaning of a road map. I'm just pointing out that Intel finally has a decent plan for the future that doesn't include those 10Ghz P4s that you mentioned.
In any case, what I'm most concerned with is the near-term of that road map. And the near-term looks even more promising than what that road map saids. For example it saids that Presler is to be launched early 2006. In fact it the 955EE launches December 27, 2005, with the rest coming January 15, 2006. Similarly early 2006 for Yonah2 means New Years day. As well Dempsey is listed as being mid-2006 while in reality it is being pushed forward like Presler in order to correct the mistake of Paxville. The performance numbers are already out and a Q1 launch is more likely.
What's most interesting is that while Merom and family is listed as End 2006, a Q3 launch is more likely. The Inquirer has already stated that Merom is way ahead of schedule and its been taped out and in production since June of this year.
While I share your reservations about the loong-term viability of road maps, the processors listed for 2006 are believable and are in fact ahead of schedule. Besides, road maps try to incorporate buffer room to account for possible problems and to allow them to pat themselves on the back when they beat targets. For instance, the quad-core Kentsfield which is meant for the consumer desktop market is already taped out with prototypes in production while its launch date isn't until mid-2007.
And in regards to Tom's AMD DDR2 comments, I didn't find them that bad in context.
"Then there is the obvious question of how Intel's new technology is going to influence the current balance of power. We can't say yet, because there are too many variables that need to be considered. Will AMD be able to take advantage of DDR2 memory?"
They were just asking whether Merom and family will be successful, and saying that AMD's DDR2 implementation is a determining factor in that. That seems perfectly reasonable.
Besdies, the initial M2 processors are still K8-based. K8L is scheduled for 2007. That means they aren't redesigning the internal workings of the processor for the initial M2 socket. The only difference is the DDR2 memory controller. One way or another, the latency will still increase by going to DDR2. Their DDR1 controller was ultra-efficient and I have no doubt that the DDR2 controller will likewise be so. Cancelling the controllers out will still yield the increased latency going from DDR1 to DDR2. The thing that offsets this is of course the increased bandwidth of DDR2.
About Intel pricing its processors below AMD. I don't think its really that big an issue anymore. Going from 90nm to 65nm would have decreased their production costs especially since Cedar Mill is identical to Prescott2M. They wouldn't be losing profits by aggressively pricing since they're just passing on the savings to consumers. The situation is even better in the chipset market. Currently chipsets are produced in 130nm using 200mm wafers. They are now transitioning to 90nm on 300mm wafers.Going from 200mm wafers to 300mm wafers alone increases the number of printed dies by 240%. That isn't even including the transition from 130nm to 90nm. Intel could easily drop their chipset prices in half and still increase their profit margins.
http://www.xbitlabs.com/news/chipsets/display/20051124214142.html
Facebook
Twitter
Instagram