Common Platform Keynote Speaker Panel

February 6, 2012, Common Platform Technology Forum, Santa Clara, CA—After their keynote presentations, the presenters sat in a panel session to elaborate on some of the topics that arose in their speeches. Michael Cadigan and Gary Patton from IBM, Kwang Hyun Kim from Samsung, and Mike Noonan from Global Foundries answered questions from the press.

Change in technology to carbon nanotubes?
Patton responded that it will take at least another decade for any new technology to replace an incumbent, and then only when the existing technology hits some wall.

RRAM versus embedded DRAM?
Patton offered that multiple technologies are open.
Noonan added that MRAM is one of the other contenders, but they have no plans for product in the next two years.

EUV timetable and existing challenges?
Patton answered that it is tough to judge, there are a lot of hard physics and engineering challenges that have to be solved before the technology is available for production. The unknown nature of the solution means they have to work on extending immersion for another generation. The facilities and investments for the technology are available. we will need triple or quad patterning if we don’t have EUV, and the costs will be unacceptable.
Cadigan noted that 7 nm will be the entry point but hopes the technology is available for the 10 nm node. Unfortunately, it is unlikely that the technology will be ready in time. The building for the equipment is ready, but the equipment makers have to solve some intricate physics. We need alternatives to EUV as backup, just as FDSOI was an alternative to FIN.

Costs?
Noonan said they have to make the investments so the solution meets designer requirements. 14 nm advances the performance of the FinFETs with low cost metal stacks and other variants. It is not only geometries, but technologies have to advance across all parameters.

Milestones for EUV?
Patton offered light source, resists, masks, and multiple other activities are on the map.
Cadigan added that the industry has chosen the base technology from ASML, but they needed help so the common technology group invested in the eximer technologies. Early beta tools are in evaluation. The industry needs EUV, because more than double patterning is going to be too expensive.

Mature process and timing?
Noonan commented 32 is in production, 28 is ramping up, 20 is starting. 14 is in the experimental stages, so there are many technologies in parallel, which is breaking the 2-year technology cycle.
Cadigan noted a change in approach, but there is still power in the common technology platform. Now, time to market and partner flexibility is the main consideration, so companies move and shift in developments. Albany is getting more and earlier involvement to get equipment to the partners. In the past, Albany moved results to Fishkill, and then to the partners. Now Albany is providing their outputs to all partners at once.

Compare your learning curve to Intel?
Patton stated that they are accelerating the dispersion of equipment to the partners.
Cadigan suggested that cost is the main focus. Samsung focuses on manufacturing with the common platform. Fishkill is a low volume fab, so any results are not like those at Samsung.
Noonan agreed that ramping quickly helps the speed of learning and aids in transferring knowledge.

450 mm wafers post EUV?
Cadigan considered that the industry would start to absorb 450 late this decade. Volume will come later, since EUV affects the 450 adoption. The industry needs EUV on 300 mm before moving to the 450 wafer. There isn’t enough time and money to to EUV in parallel with 450. Lithography costs are important.

Extend immersion?
Patton noted that programmable light sources permit a lot of resolution without double patterning. They are working on optimizing colors and patterns to avoid 3 and 4 sets of masks.

20 nm in ’13, 14 nm in ’14, short technology lifetime for 20?
Noonan said that some will want to move from 28 to 20, while others will go directly to 14. the BEOL and MEOL processes are the same, but you need new technologies to implement either. There is no capital expenditures projected for this.
Kim chimed in that it depends upon the customer needs, but the company has no preference.

Low power and ARM, no high performance variant?
Noonan postulate that everything will be mobile or mobile-like and will have to address performance and power density. The new processes offer the opportunities to explore new architectures, and by moving to more cores, one can improve performance in other means than faster transistors. Some will choose a 3-D stack while others will go for higher integration, and a few will do both. The 20 common platform scales in the design space and is ok for many super high performance functions.

Fabless companies not moving to 20?
Patton opined that companies are always looking for alternatives to higher costs. If you continue to get 50 percent area scaling added to a 35 percent cost increase, it still results in a 15 percent cost benefit. That benefit is decreasing, so the choices get harder. The expectation is that you cannot extrapolate the costs for 20, 20 FIN, and 14 at this time.

Goal was to have fab parametric compatibility ?
Cadigan agreed, but now the fabs are evolving. The number of companies that transferred production has been approaching zero. Now the main lines are working on common features. Samsung will tune the process for their own production, and Global has some large customers who want process customization. Albany is setting the baseline only.

Cadigan observed that the common platform has worked well since ’04. The changes in the group now have the volume producers calling the shots. Changes in the model now focus on time to market and the speed to the next node. The group will have new business models in the ’14-15 timeframe.

Patton noted the concern for cost scaling issues. EUV needs a lot of work, so the plan “b” is to work more on source optimization. 20 nm will call for cost sharing for double patterning while the new structures for 20 FIN and 14 will have new structures and greatly increased complexity. 10 will be a true 14 shrink.

Kim commented that the common platform has been successful in the past and is still viable. The partnership reduces the costs of development for al of the partners.

Noonan added that the common platform created a critical mass for the equipment suppliers and helped to manufacture products for the benefit of the customers. The common platform gives a head start to compete profitably.