You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I believe this was one of the downsides of 5N4Y. Less time for tool evaluation and reducing expensive process flow decisions. 14A should be better in that respect with much less time pressure though some of that work was almost certainly started under Gelsinger much of it has been under the direction of Lip-Bu Tan. I'm curious to see what 10A is like since that will be a node that will be developed entirely under Lip-Bu Tan's oversight.
"14A should be better in that respect with much less time pressure".
Due to the fierce competition in AI, I think 14A will face a even higher pressure on the decision clock window. All the major potential customers must make decisions soon if they haven't already made ones.
Due to the fierce competition in AI, I think 14A will face a even higher pressure on the decision clock window. All the major potential customers must make decisions soon if they haven't already made ones.
"14A should be better in that respect with much less time pressure".
Due to the fierce competition in AI, I think 14A will face a even higher pressure on the decision clock window. All the major potential customers must make decisions soon if they haven't already made ones.
When the smoke clears, I think we will find 18A and 14A are like 20A and 18A were supposed to be . 18A is a checkout on two unbelievably complicated architecture changes and 14A is a optimizations and shrink. not sure what 18AP is but if it runs in Fab 52, the tool set is done.
"14A should be better in that respect with much less time pressure".
Due to the fierce competition in AI, I think 14A will face a even higher pressure on the decision clock window. All the major potential customers must make decisions soon if they haven't already made ones.
Intel ran through several process changes (not necessarily 5 full nodes, but still a much more rapid pace than standard) in those 4 years. Intel had to do more in less time to get back to being competitive at the leading edge. And doing more with less time means having less time to spend on things like tool selection and cost effective process flow development.
Now they are only looking to do a single node change at a standard pace, so they have more time to focus on getting the tool selection and process flows optimized before they have to move to designing the next node. And as MKWVentures pointed out, 14A is focused on refinement, not on completely new architectural changes, which should mean fewer tool and process changes in that time frame. That gives Intel the opportunity to improve the cost structure.
Intel ran through several process changes (not necessarily 5 full nodes, but still a much more rapid pace than standard) in those 4 years. Intel had to do more in less time to get back to being competitive at the leading edge. And doing more with less time means having less time to spend on things like tool selection and cost effective process flow development.
Now they are only looking to do a single node change at a standard pace, so they have more time to focus on getting the tool selection and process flows optimized before they have to move to designing the next node. And as MKWVentures pointed out, 14A is focused on refinement, not on completely new architectural changes, which should mean fewer tool and process changes in that time frame. That gives Intel the opportunity to improve the cost structure.
Yes, Intel 14A provides an opportunity for Intel to improve its manufacturing capabilities. However, the time pressure is enormous because neither the growing demand from Intel Foundry's potential customers nor the competition from Intel's competitors can wait, and they won't wait.
Intel Foundry's survival depends on its ability to advance at a pace that meets or exceeds market demand and competitive pressures. Competition is accelerating and the amount of time Intel has to work out the nuts and bolts of its manufacturing process is rapidly shrinking.
