The Wiretap Intercept No. 111027
opinions and skeptical speculations too small to fit into an Industry Gadfly column
Subject: Xilinx CEO says TSMC 28 nm "fabulous", but 20 nm "challenging"

In a Monday article in ElectronicsWeekly.com, Moshe Gavrielov, the CEO of
Xilinx said about TSMC:

   "I think 28 nm is going to be a fabulous node.  A lot of learning
    went into 28 nm to avoid the problems of 40 nm.  Four flavours,
    aimed at various markets, will give it tremendous longevity."

        - Moshe Gavrielov, CEO of Xilinx, (EW 10/24/11)

Because of unrelenting pressure on FPGA's to keep up with ASIC's and SoC's,
Xilinx is one of TSMC's most technologically demanding customers.

This comment came on the same day that TSMC announced that their 28 nm node
was in volume production.  The TSMC press release claimed that their 28 nm
had 80 customer tape-outs -- which they said has "more than doubled" the
number of tape-outs it's ill-fated 40 nm node had.


TSMC'S EARLIER 40 NM YIELD PUBLIC RELATIONS NIGHTMARE

TSMC's 40 nm yield problems were so serious that an engineer brought his
worries into the DeepChip.com discussion back in early 2010:

   Our project is almost ready for tapeout but our VP is afraid to commit
   to TSMC 40 nm because of their off-and-on again yield issues.

   07/06/08 -- Rumors of first parts at 40 nm scheduled for late 2008,
               then delayed to Jan 2009, then delayed to March 2009.

   02/09/09 -- Charlie Demerjian reports TSMC 40 nm has leakage problems
               and power saving at 40 nm will be "minimal to negative".

   04/30/09 -- TSMC CEO Rick Tsai admits in a conference call to
               analysts that "There have been difficulties with the
               yields.  40 nm is a difficult technology to manufacture.
               We understand the root of the problem."

   06/11/09 -- Morris Chang takes back the TSMC CEO job from Rick Tsai.

   06/12/09 -- Goldman Sachs downgrades TSMC from "buy" to "neutral".

   07/01/09 -- FBR Capital Markets reports for 40 nm "we believe yields
               are as low as 20 percent to 30 percent, which may explain
               the recent management change."

   07/31/09 -- DigiTimes reports "In response to TSMC's 40 nm yield issues,
               Chang said rates have improved from 30% to 60%" at a July
               investors conference.  (i.e. the yield problem is fixed.)

   10/07/09 -- UBS downgrades TSMC from "buy" to "neutral".

   10/30/09 -- DigitTimes reports "TSMC said it has seen yield rates for
               its 40 nm node drop to 40% due to chamber matching issues."
               (i.e. the yield problem is back!)  And "CEO Morris Chang
               pledged the issue will be solved within the quarter."

   11/10/09 -- Nvidia confirms in the conference call that they were supply
               constrained due to TSMC low yield issues.

   01/13/10 -- DigitTimes reports "TSMC has been struggling to increase
               yields on 40nm to over 70%, according to industry sources"
               and Xilinx was moving to UMC to minimize risks.

   01/20/10 -- DigiTimes reports that Mark Liu, senior VP at TSMC had stated
               "that the chamber matching problems that had impacted yield
               rates for the company's 40 nm node have been resolved" and
               that 40 nm yield is "about the same level as its 65 nm node."

   We don't know if we should double order TSMC and pray this is mostly true
   or to focus our efforts on making UMC or GF as our primary.  Suggestions?

      - from http://www.deepchip.com/items/0484-05.html

Since that Feb 2010 post, 40 nm TSMC yield problems had been plaguing both
Nvidia and AMD such that it was hurting both their chip sales.


In stark contrast, today's TSMC 28 nm press release reports that "TSMC 28HP,
28HPL and 28LP are all in volume production and 28HPM will be ready for
production by the end of this year."

Altera, AMD, Nvidia, Qualcomm, and Xilinx were all happy 28 nm customers
cited in this press release.  Rumor has it Texas Instruments also has
some ARM Cortex chips at TSMC 28 nm, too.  And some industry observers were
surprised to find Apple missing in this TSMC 28 nm press release.

Along with implied good yields, TSMC claims its 28 nm chips are 45% faster
than its 40 nm chips.

         ----    ----    ----    ----    ----    ----   ----

STILL A WARNING ABOUT TSMC 20 NM

   "20 nm will be a challenging node.  Double patterning and finfets
    will make it take some time."

        - Moshe Gavrielov, CEO of Xilinx, (EW 10/24/11)

Which is no big surprise because everyone's still having trouble at 20 nm.
      An archive of prior intercepts       Next intercept       To reply or send a story to John

 Sign up for the DeepChip newsletter.
Email
 Read what EDA tool users really think.


Feedback About Wiretaps ESNUGs SIGN UP! Downloads Trip Reports Advertise

"Relax. This is a discussion. Anything said here is just one engineer's opinion. Email in your dissenting letter and it'll be published, too."
This Web Site Is Modified Every 2-3 Days
Copyright 1991-2024 John Cooley.  All Rights Reserved.
| Contact John Cooley | Webmaster | Legal | Feedback Form |

   !!!     "It's not a BUG,
  /o o\  /  it's a FEATURE!"
 (  >  )
  \ - / 
  _] [_     (jcooley 1991)