When the world moved from micro to nano, one of the myths which arose was that IDMs would win out over foundries.
It was argued that mis-matches between design and manufacturing would be so likely, and would result in such disasters, that only those companies with control of both design and fab would be able to successfully produce nanometre wafers.
Like nearly all generalisations about the semiconductor industry, and most of the predictions, this has turned out to be a nonsense.
UMC has been successfully shipping 65nm wafers on its high performance process for Xilinx since March, and TSMC started shipping 65nm wafers on its low-power process for Qualcomm in April.
TSMC says it is now shipping 65nm wafers to 15 customers, and will have supplied 7,000 finished 300mm 65nm wafers by the end of this year.
UMC is now shipping eight different devices for Xilinx and first full production parts are due in Q1 2007.
Subscribe to this blog's feed
Comments (2)
Hi David,
Just to confirm your data, we at Chipworks have looked at Xilinx Virtex-5 samples with date code 0617 (late April), and from Toshiba a little later with dc 0629 (July). Both were ES samples (engineering samples), but obtained through normal distribution channels.
More remarkably, we are currently analysing a 65-nm TI-designed Nokia baseband chip (out of a 2610 phone), fabbed by UMC with a July date code. There was some question whether TI had fabbed the front end and shipped the wafers to UMC for the back end, or whether UMC did the whole process; but just having returned from IEDM, "industry sources" confirmed that UMC are indeed doing the complete processing.
This is the first example I can think of where an IDM has used a foundry to produce new generation product in parallel with (or maybe even before) it's own fabs. We are still hunting for 65-nm TI product out of TI's DMOS6 or Kilby fabs.
Judging by the commentary from TI and at IEDM, this is a very conscious decision by TI - they appear to be intending to use foundries for the majority of their 65-nm product, and the new Richardson fab will be wound up on 45-nm.
A bit of a change from when the foundries were two generations behind!
Posted by Dick James | December 20, 2006 3:01 PM
Posted on December 20, 2006 15:01
This is very good research indeed!
Posted by Matt Cromie | January 2, 2007 10:08 PM
Posted on January 2, 2007 22:08