NVIDIA recently announced that it has fully learned the lessons learned from the 40nm process and will not repeat the mistake at 28nm. After in-depth cooperation with TSMC, it has now obtained a 28nm chip that can work.
NVIDIA chief Huang Renxun said yesterday: “We are much better prepared for the 28nm process than 40nm, because we attach great importance to it. We have achieved success in a large number of different technological eras, so that everyone forgets how hard it is. This time, we set up an entire team dedicated to serving the new process. We performed a very large number of chip tests on the 28nm process and already had a working wafer."
TSMC's 40nm process used to be too low for long-term yield due to the technology itself, manufacturing issues, and design flaws. Unfortunately, NVIDIA was not fully aware of this when it first developed the 40nm GPU, so it encountered a series of unanticipated problems directly. This led to the delay of Fermi's Fermi architecture products. I believe this time NVIDIA will not fall again in the same river.
Huang Renxun added: “(28nm) looks great, much better than what we experienced at 40nm. Our extensive and intimate cooperation with TSMC guarantees that we can start production at the right time. In general, I feel good about 28nm."
Huang Renxun did not disclose when "Kepler" was put into production and released, and according to previous official statements, it will wait until some time in 2012.