It will also mean using gate-last rather than gate-first manufacturing and one of the ways that AMD might try to get around it is by doing a 28nm die shrink of existing Ontario/Zacate products.
This will lower Brazos’ power consumption and buy time for AMD to come up with a new processor.
If it presses ahead with this cunning plan a 28nm APU might see the light of day in 2012, but it will take at least 18 months to get a real 28nm replacement for the GlobalFoundries project which was codenamed Krishna and Wichita.
It is hell on toast for AMD because it means that any momentum AMD built for itself through 2011 has stalled and any connection between AMD and GlobalFoundries must be damaged.
AMD’s relationship with GloFo has been getting increasingly frosty lately. AMD has been shrinking its stake in the outfit and no longer has the right to designate a representative to the GlobalFoundries Board of Directors.
GloFo executives have fled the outfit in the past year and AMD has been getting increasingly close with TSMC.
AMD cancelled Krishna/Wichita because GlobalFoundries could not ramp the parts at volume and was unwilling to negotiate a new wafer agreement. At the moment AMD only pays GloFo for good 32nm dies, which means GlobalFoundries has been stuck building Llano at a loss for most of the year. This agreement ends at the end of the year which means AMD will have to start paying full whack for wafers. Not surprisingly AMD doesn’t want to pay a per-wafer fee for a slow, low-yield product ramp.
It looks like AMD and GloFo will collaborate on AMD’s main product lines but only because neither has any choice. Meanwhile AMD is looking towards TSMC for all its future projects.