TSMC kicked off their annual Technology Symposium series in San Jose today. Founder, Chairman and CEO Morris Chang opened the proceedings by addressing concerns for how Japan’s disaster recovery will impact the foundry’s supply chain. Dr. Chang listed a number of issues that had been of concern; metal sputtering targets, CMP (chemical-mechanical polishing) slurry, raw silicon wafers, chemicals, tools and spare parts. However, he assured the audience, TSMC has been able to solve all supply problems and “everything is under control, no supply shortages will interrupt the production lines“.
TSMC does expect an impact on their customers, however, and their customer’s customers. Dr. Chang said that there will be some effect on the industry in Q2 and possibly into Q3, but the impact would be no more than two quarters. This will add to a slight softening in the world economy; including Chinese efforts to fight inflation, problems in Europe, some economic measures in U.S., with the end result being that 2011 will be less strong than TSMC had originally thought. The company is resetting their forecast from 7% growth down to 4%.
Dr. Chang then went on to describe how the semiconductor market drivers have changed; from PCs (1980-2000), to cell phones (2000 - 2010), to today’s “killer app” - mobile products, including smartphones, tablets, and any “devices you carry around“.
Moving to more technical topics, Dr. Chang predicted that the 20nm process that they have in development will provide 2X the performance over 28nm. TSMC has also begun investigation at 14nm, and “Moore’s law still has some distance to go“, according to Chang. He also noted that the company has “significant R&D” in 3D IC technology; including 2/2.5D interposers, TSV (through-silicon vias), and what the company refers to as “system-level scaling“.
Dr. Chang concluded by comparing TSMC’s capacity (2 - 12″ fabs produce up to 260,000 wafers/month), to competitors, stating that they are building capacity but they don’t have the technology, that what competitors have is not “effective capacity“. TSMC is also working on production of 450mm wafers. Pilot production is targeted for 2013-14, with an “intercept point” for production of 20nm in 2015-16.
The 20nm roadmap
Next up on the morning’s agenda was Dr. Shang-Yi Chiang, TSMC’s Sr. VP R&D, to go over TSMC’s advanced technology roadmap. Looking out to the next process node, Dr. Chiang said that the 20nm (20G) process will be available in Q3 2012. In the meantime, TSMC will offer4 versions at 28nm; LP, HPL, G, and HPM. HPM will combine high performance and low power, the sweet spot for ARM cores according to Dr. Chiang.
Dr. Chiang said that the migration to a new process node typically adds 2-3 process modules, but 5 new modules are needed for 20nm. The 20nm process node will be the last generation for planar transistors. The most expensive 20nm process module will be the change to double patterning. TSMC has developed software that will automatically apply DFM requirements to create the 2nd mask from design data.
TSMC’s 20G and 20SoC processes will offer >2X the performance and <0.75X the switching power compared to 28HP & 28HPM. A 20nm SRAM has been fabricated, but 100% yield has not been achieved yet. At 20nm, TSMC is seeing a sidewall interface effect that can dominate interconnect delays, now that line widths are approaching the mean-free path of electrons. According to Dr. Chiang, TSMC has developed a smoothing process that minimizes the effect of electrons literally bouncing off of the walls of interconnect.
TSMC says that they are innovating to extend immersion lithography to 20nm, but that NGL (next-generation lithography) will be required for nodes <20nm. Dr. Chiang said that TSMC plans to have an ASML EUV machine, the NXE-3100 at TSMC by mid-2011.