heyheybooboo
Diamond Member
- Jun 29, 2007
- 6,278
- 0
- 0
Originally posted by: Cogman
Originally posted by: Idontcare
Originally posted by: Nemesis 1
I here this allthe time. Iuse to agree . Idon't care knows this stuff I don't . But the way I read it . AMDs 45nm process is designed to go right to 32nm . So they may catch intel napping . Plus the new fab company owners can push up 32 development. Who knows . But the exact same applies to Intel from 32 to 22nm.
Other way around, Intel greased the skids to transition from 45nm to 32nm by adopting HK/MG at 45nm.
AMD's got a LOT of work to do in the production environment with HK/MG just to get up the learning curve that Intel has alreday climbed. I'm talking device reliability and characterization, mastering new yield loss mechanisms that simply don't exist on traditional SiON/poly-Si gate stacks, etc.
AMD's plus is that they have put immersion litho into production before Intel. This will be one less hassle for AMD as they transition to 32nm but there is absolutely no comparing the technical difficulties and differences between implementing a robust HK/MG integration scheme and implementing a turn-key supplier provided piece if equipment (ASML will sell the immersion litho equipment to anyone, AMD does not own the technology).
I thought that AMD was going fabless though? (IE they are not building any more new fabs in the near future). I feel so out of the loop.
It will be interesting to see how this price war pans out. Hopefully it doesn't end with AMD being out.
As I understand the whole 'deal': AMD has discussed the option of high-k/metal-gate in future spins at 45nm. With their limited resources it would be difficult to see any benefit in reaching full-production but who really knows what AMD has up their sleeve other than simple survival.
It would not be out of character for them to do some wacky 'proof of concept' in their process before moving down.
I think the 'Fab Deal' requires ground to be broken this spring on the Malta facility for 32nm. Where this fits in AMD's Grand Plan I'm not sure anyone knows.
UMC validated high-k/metal-gate at 45nm last month with the intention of 32nm in 2010. I don't know the current status of their relationship with AMD only to say that over the last eight years or so there have been exclusive 'arrangements' between UMC & AMD.
When you factor in TSMC announced a half-node (40nm) SOI last summer it looks like the dots are aligning for 'Shrike'. Shrike is the mobile platform with the 'Swift' microprocessor. The Swift CPU is the basis for the whole fusion thing and I think it's first run is to include 2 microprocessor cores and a GPU core on die.
It looks like the overriding issue with Swift is that folks have been slow thus far to incorporate any of AMDs SSE5 instructions (which is required to take advantage of the parallel processing power of the GPU on the die).
Got that? You will be tested later