Article: AMD's Mobile Strategy
By: Mr. Camel (a.delete@this.b.c), December 27, 2011 8:50 am
Room: Moderated Discussions
I think AMD should implement a subset of the x86 architecture in a mobile CPU that is targeted at tablets and smartphones.
I am not completely sure what that subset would be but here are some things that I think they could drop:
- most (but not all) x86 instructions involving microcode
- string rep
- most of the protected mode segmented memory model mess
- transcendental fp instructions
- etc.
- self modifying code (i.e. don't support it)
It would not seem to be too paralyzing to drop these because most mobile platforms (iOS, Android) are currently tailored to architectures where those things don't exist anyway. So moving those to this new subset of x86 would be just as easy as moving them to the current x86 instruction set.
Also, removing these would make it easier for AMD to come up with a lower power design using fewer transistors.
Intel would never do this because they are paralyzed by the
antiquated thinking that 100% backwards compatibility is a must. This may be true for the traditional PC market (actually I doubt that it is true even here), but not for the ultra mobile market (smartphones and tablets). Perhaps Intel hopes that their 22 nm process technology lead will save them, but I am skeptical.
Any thoughts?
If this has already been suggested, then I apologize.
I am not completely sure what that subset would be but here are some things that I think they could drop:
- most (but not all) x86 instructions involving microcode
- string rep
- most of the protected mode segmented memory model mess
- transcendental fp instructions
- etc.
- self modifying code (i.e. don't support it)
It would not seem to be too paralyzing to drop these because most mobile platforms (iOS, Android) are currently tailored to architectures where those things don't exist anyway. So moving those to this new subset of x86 would be just as easy as moving them to the current x86 instruction set.
Also, removing these would make it easier for AMD to come up with a lower power design using fewer transistors.
Intel would never do this because they are paralyzed by the
antiquated thinking that 100% backwards compatibility is a must. This may be true for the traditional PC market (actually I doubt that it is true even here), but not for the ultra mobile market (smartphones and tablets). Perhaps Intel hopes that their 22 nm process technology lead will save them, but I am skeptical.
Any thoughts?
If this has already been suggested, then I apologize.