Article: AMD's Mobile Strategy
By: Michael S (already5chosen.delete@this.yahoo.com), January 13, 2012 6:39 am
Room: Moderated Discussions
Gregory (gregory2012@gmail.com) on 1/12/12 wrote:
---------------------------
>You wouldn't reuse the design for a car engine in a train just because "design reuse
>seems like a good thing". Similarly, you can't expect to have an optimal design
>if you reuse the same processor core in a notebook chip (Sandy Bridge) and a server
>chip (Sandy Bridge-EP).
Your analogy sucks, like all analogies.
Proper analogy would be: in many situations you doe use the same class of engines (diesel) both in cars and in trains. You start to use different class of engines only when environment is non-trivially different, e.g. you trains could be easily connected to electric grid, but your cars have no such luxury because they have to reach everywhere.
Same as when your server dedicated to limited class of tasks, e.g. throughput computing, it can use more specialized implementation (Sun/Oracle UltraSparc-T line, Intel's ever-coming LRB). But general-purpose servers are, in fact, have pretty close requirements and constrains to desktops, and even more so, to laptops.
> The complexity of x86 is forcing this design reuse.
>
Except when circumstances were sufficiently different, complexity of x86 did not prevent Intel from creating 3 distinct co-existing microarchitectures.
Even resource-constrained AMD plans for 2 co-existing microarchitectures and could easily and up with 3, in case Trinity proves himself significantly worse than Llano for some important market segment.
---------------------------
>You wouldn't reuse the design for a car engine in a train just because "design reuse
>seems like a good thing". Similarly, you can't expect to have an optimal design
>if you reuse the same processor core in a notebook chip (Sandy Bridge) and a server
>chip (Sandy Bridge-EP).
Your analogy sucks, like all analogies.
Proper analogy would be: in many situations you doe use the same class of engines (diesel) both in cars and in trains. You start to use different class of engines only when environment is non-trivially different, e.g. you trains could be easily connected to electric grid, but your cars have no such luxury because they have to reach everywhere.
Same as when your server dedicated to limited class of tasks, e.g. throughput computing, it can use more specialized implementation (Sun/Oracle UltraSparc-T line, Intel's ever-coming LRB). But general-purpose servers are, in fact, have pretty close requirements and constrains to desktops, and even more so, to laptops.
> The complexity of x86 is forcing this design reuse.
>
Except when circumstances were sufficiently different, complexity of x86 did not prevent Intel from creating 3 distinct co-existing microarchitectures.
Even resource-constrained AMD plans for 2 co-existing microarchitectures and could easily and up with 3, in case Trinity proves himself significantly worse than Llano for some important market segment.