Chipmaking’s future: all of the nodes all of the time
The Monday morning keynotes at the 51st Design Automation Conference (DAC) in San Francisco underlined the collapse of Moore’s Law and the spreading effect it is having on how older process nodes are used.
Hossein Yassaie, CEO of Imagination Technologies, said: “Moore’s Law is really over from my point of view. It’s not that it can’t scale, it’s that the cost is not going down anymore. But that doesn’t matter because the various nodes are needed for different reason.
“The 28nm node will be here forever and 40nm will probably be the same,” Yassaie said.
In his Visionary Talk that preceded Yassaie’s, Antun Domic, executive vice president and general manager of the design group at Synopsys, said: “I believe that the established technology nodes are going to be driving the needs of advanced EDA tools. The new technologies initially targeted for advanced nodes will play a critical role for the established nodes.”
Domic said R&D for multi-patterning is being brought back to older nodes. “The effort that EDA industry put in place for double patterning brought huge advances in routing.”
The techniques reduced congestion on a 28nm design. “The best techniques for synthesis and floor planning were combined with place and route. You can do a complex chip with high utilization going forward.”
Similar technology was taken back to older nodes such as 160nm and 180nm. Domic showed a 160nm design with a large amount of mixed-signal circuitry. “Initially, it could not be routed using traditional tools. But taking pre-placement into synthesis we were able to do this. If you add other pieces, you can address reliability requirements, using features such as double vias and advances made with double patterning.
Domic added: “The EDA industry will have to provide tools that will enable people to differentiate by design. Differentiating by technology will be very difficult. There will be fewer sources and the life of established nodes will be longer. That should force innovation in EDA.”
Development for mature processes, Domic said, will need to run as part of the program for advanced technologies. “Given the financial constraints, we don’t have the possibility to develop parallel tools.”
