Big news that took several quarters to fulfill
In what might be some of the most interesting news this
week, sources at Taiwan Semiconductor Manufacturing Company are now able to
report that yield issues with its 40nm half-node process have finally been
resolved. Over the past few quarters, a significant portion of the
semiconductor manufacturing industry has kept an observant eye on the
company’s 40nm transition progress, waiting for what has seemed like ages for a satisfactory transition to the new lithography process since its original announcement in March 2008.
As many observers have witnessed, the
unsatisfactory yield rate of the 40nm process node over the past few months
violently struck the world’s next-generation desktop graphics processor market
in more fiscally diminishing rates than had originally been expected. The
semiconductor industry at large managed to significantly effect its North
American customer operations as corporate giant AMD struggled to produce a
sufficient output of its RV870 chips due to the low yield rates. While the
company did its best to cope with enthusiast end-user demand for the new GPUs,
it simply could not maintain a market equilibrium adequate for the overreaching
demand of its customer base.
Meanwhile, Nvidia decided to hold back on its efforts to
compete with AMD in the high-end desktop graphics processing space during the entire
second half of 2009. Internal corporate strategies and engineering misfortune
lead the green giant to delay production of its highly anticipated GF100
Fermi-based desktop graphics processors for its high-end consumer base. In this case, however, it was
not a matter of whether or not TSMC could provide an adequate amount of 40nm
yields – Nvidia simply was not ready with final production samples of its 40nm
GF100 (GT300) architecture. As far as we know, lab testers and quality control
engineers are just now receiving risk wafer batches of A3 silicon from the
oven, and the next step is to ensure that they are ready for volume production
over the next few weeks.
According to Mark Liu, Senior Vice President of Operations
at TSMC, the semiconductor giant has finally managed to get its 40nm yield
rates on par with its 65nm production line. Back in October, CEO Morris Chang stated,
“lots of new tools were brought in and they added a complexity to production.
The problems that [had] cropped up had to do with two chamber matching. We
consider these to be logistical problems and we [were] in the process of
resolving them very quickly.”
It is important to note that the company originally stated
it would be able to completely resolve the issue by the end of 2009. At the
same time, it expected 40nm wafers to account for 10-percent of total wafer
revenue by the end of the year. Whether or not that outcome is true remains to be
observed.
What the company’s GPU and FPGA customers now await is an
obligatory verification that its 40nm yield issues have in fact been resolved
to reasonable extents. We look forward to expecting volume production of 40nm
chips over the remainder of Q1 2010 and remain observant in the outcome of AMD’s
RV870 marginal revenue numbers as well as Nvidia’s production of
next-generation GF100 single-GPU and dual-GPU desktop cards.