Home > EDA > TSMC iPDK Debate: Lets Play Monopoly!

TSMC iPDK Debate: Lets Play Monopoly!

Talk about a giant game of monopoly, probably one of the most entertaining panels of the 46th Design Automation Conference was: Will Interoperable PDKs Fly in a Stodgy Analog World? The real question here is will TSMC be able to break EDA tool strong holds for the greater good of semiconductor design enablement?

Monopoly

A Process Design Kit (PDK) is the interface between semiconductor design tools and the semiconductor manufacturing process. For every process node there will be multiple PDKs to support the different tool vendors. Multiply that by dozens of process variations and you get millions of dollars in foundry overhead expense. Closed PDKs also allow an EDA vendor to dominate a given market segment. For example: the analog design world is dominated by the Cadence Virtuoso franchise which maintains an 80%+ market share by locking customers into a Cadence proprietary PDK technology, specifically, the Cadence PCell (parameterized cell) and the Skill scripting language.

Like any other industry, EDA standards initiatives have been failing since the beginning of time. Vendors need to lock customers into a set of products to keep the ASP (average selling price) high, which of course suppresses innovation by locking other companies out. So the trick is for the top EDA companies to participate in standards initiatives when it helps them, to block them when it does not, but most importantly pretend you are not trying to monopolize when of course you are.

catbox

Accellera was a token EDA standards organization that supported a mix of user and vendor standards. It is less constrained than a real standards organization like IEEE, so it is a veritable sandbox for the big EDA standards cats. In 2003 Accellera announced an Open Kit Initiative to standardize PDKs, members included the top EDA, IP, and semiconductor companies, plus TSMC and other foundries. The only thing accomplished by this initiative was the website. Accellera recently merged with the Spirit Consortium, which to me is just putting more sand in the box.

Enter TSMC:

TSMC Launches First Advanced Technology Interoperable Process Design Kit Industry-Wide Collaborative Development Effort Delivers First Unified iPDK for 65nm Process Hsinchu, Taiwan, R.O.C. – July 21, 2009 – Taiwan Semiconductor Manufacturing Company, Ltd. (TWSE: 2330, NYSE: TSM) today announced the release of the semiconductor industry’s first interoperable process design kit (iPDK) for advanced technology…

Other than removing millions of dollars of expenses from the TSMC financials, iPDKs will increase competition between EDA tool vendors spurring innovation. It will also strengthen TSMC’s position as the leading  innovator in the merchant semiconductor manufacturing business. Of course TSMC needs to keep their ASPs above industry average so this is also a monopolistic move.

800lb Gorilla

“The iPDK is a long-awaited solution to many of the design problems that have delayed new technology and tool adoption. Multiple and/or incompatible PDKs on proprietary custom design databases have limited design reuse and portability, and has resulted in high development, maintenance and support costs for all ecosystem partners including TSMC,” explains ST Juang, senior director of design infrastructure marketing at TSMC. “The interoperable design kit will enable higher level of innovation and differentiation for our customers in full custom, analog, and mixed-signal designs.”

If not for TSMC, open PDKs would never happen, the EDA industry would continue business as usual, whining about standards, but doing nothing about it. Other foundries are certainly at risk if they do not adopt the iPDK framework, believe me, there is no stopping TSMC on this one. The only risk I see with iPDK is Apple taking issue with the name.

  1. August 27, 2009 at 5:17 PM | #1

    Dan,

    Nice article.

    iPDK was a buzz word in the latest DAC.
    Is iPDK talking about iCDK (OA Pcell) only,
    also regarding entire PDK (Model, DRC, LVS,
    DFM. Parasitic extraction deck)?

    In one DAC “pavilion session”- the panel was talking about
    iDRC, iLVS.

    Any good link/ doc on iPDK, iDRC, iModel…?

    Thanks,
    Faizul

  2. new reader
    August 27, 2009 at 8:29 PM | #2

    I agree tsmc releaseing an open pdk is a great push towards having standard pdks across the industry (good news for all).
    But unless i’m missing somethign this is based on the open access database which is an accellera standard isn’t it? So it seems to me Cadence took a big chance by opening this db as a industry standard and it’s hard to imagine how they get anything out of this other than a reduction in their market share. (our best guess is that it will make it easier for them to aquire and intergrate new innovative companies)

    Sure it took a long time, but without this donation from cadence to a standards body the ipdk wouldnt’ be possibible. Like most everythign in semiconductors it takes collaboration.

  3. new reader
    August 28, 2009 at 2:56 PM | #3

    if this is all about pcells why not pcells come built in.., the tools should be able to generate any pcells given certain set of rules…

    • dothead
      September 7, 2009 at 3:23 PM | #4

      Why don’t PCells come built in? Built into WHAT? They are an integral part of a PDK, and I think you severely underestimate the effort and skill (no pun intended) needed to create high-performance PCells. They are based in part on the specific design rules of a given process, but it may not be clear which specific rule governs a given structure.

  4. Bill
    August 31, 2009 at 4:56 PM | #5

    In reply to new reader, OpenAccess is a Silicon Integration Initiative (Si2) standard, not Accellera

  5. Tim
    September 2, 2009 at 10:38 PM | #6

    … and SKILL is not an unlicensed standard – Cadence still owns it. However, there are equivalent representations available to plug into OA, and Ciranova has a strategic PyCell partnership with TSMC.

  1. September 10, 2009 at 2:02 PM | #1
  2. November 2, 2009 at 4:05 AM | #2