Although Taiwan Semiconductor Manufacturing Company seems to have rather ambitious plans for its 28nm node, Nvidia Corp., one of the largest customers of TSMC, claims that the volumes of 28nm chips TSMC will be able to provide this year will not be enough to support a high-scale launch of the latest Tegra 3 "Kal El" product.
"28nm is not available this year, not until the very end of the year. I think, for us to ship [Tegra "Kal El"] production out in Q3, we have to start wafers in early Q2, right? So, 28nm is not an option. Secondly, 40nm is now in the third year of its production, and the yields are fabulous. [...] So 40nm process technology is absolutely the right approach. It is the most mature, and we can go into very high [volume] production very quickly because the yields are so great," said Jen-Hsun Huang, chief executive officer of Nvidia, during a conference call with financial analysts.
Tegra "Kal El" system-on-chip is based on four Cortex-A9 application cores, will feature a GeForce graphics processor with twelve stream processors and will have a new display and video engines capable of supporting Blu-ray disc video playback and stereo-3D graphics output. Other capabilities, clock-speeds, power consumption and die size of the novelty are not known previously.
Nvidia anticipates that its clients will integrate Tegra 3 into new products starting from August and it is natural that the company did not want to wait for 28nm process, which will become commercial only in the fourth quarter of 2011. Nvidia's intention to design the Tegra "Kal El" for 40nm products may signal that the company does not believe in TSMC's ability to supply enough 28nm chips this year.
According to TSMC, wafers made using 28nm fabrication process will account for 2% - 3% of its revenues in Q4 2011, which essentially means that the company aims to supply 28nm chips worth around $100 million at that time. The volume is not small for a brand new process technology, but is definitely not high too.
"40nm is actually more economical than 28nm this year and we will likely expect it to be so until about first half, maybe even the midpoint of next year," added Mr. Huang.
In fact, Nvidia is looking forward Intel's "tick-tock" model for transitions to new process technologies and architectures.
"You have noticed other companies that have used different ways of explaining their rhythm and some use one process node change. One as a architecture change, the next one is the process node change. Basically, it is every other year for a new process node. I think that rhythm is not a bad rhythm. I mean, that is basically how quickly the industry is changing process. So [this is] the reason why Tegra 2 and Tegra "Kal-El" will use the same process, and then the following ones would be 28nm," explained the head of Nvidia.