How’s 28nm?

So what’s happening with 28nm?

Nvidia, which pays for its chips by the wafer, has blamed low yields on 28nm for a decline in profit margins.

“The gross margin decline is contributed almost entirely to the yields of 28nm being lower than expected,” says Nvidia CEO Jen-Hsun Huang.

Because Nvidia has a wafer-based pricing arrangement with TSMC, the fewer die per wafer then, obviously, the higher the die cost.

Last month at IFS2012, Mike Bryant, CTO of Future Horizons, said that several customers were experiencing yield problems on TSMC’s 28nm process.

All that TSMC would say in reply was that the 28nm ramp was 3x faster, and defect density issues 3x better, than the 40nm ramp at the same stage. However 40nm was a notoriously tricky node.

Xilinx recently said it is having the fastest new-node product roll-out in its history having shipped four of its five 28nm product families in 11 months. That is half the time it took to roll out initial devices in two product families at 40nm.

Nvidia, of course, makes very large die which are more likely to have defect densities, which could account for the different experiences of Xilinx and Nvidia.

Also Xilinx uses the HPL version of TSMC’s 28nm process while Nvidia is using the HP version.

Not many can do 28nm. Intel – with its 32nm equivalent obviously –  and Samsung up to a point, are  the only ones. GloFo is said to measuring its 28nm yield in wafers-per-die. ST-Ericsson has still not announced its 28nm fab.

It all goes to show how difficult advanced nodes have become, how crucial they remain to competitiveness and how ineffably silly it was of the financial community to press their view that advanced CMOS manufacturing was a commodity.

Tags: Nvidia, tsmc, wafer, wafers

Related Tech News

7 Comments

  1. David Manners
    February 21, 2012 15:50

    An excellent point, plk, as a SEMICO report last year pointed out, 36 fab-sites representing 41% of worldwide fab capacity, are in the seismically sensitive ‘Pacific Rim of Fire’. So the answer is – go to GloFo for fab because Dresden and New York are not in the Rim. Just check out the yields first!

  2. plk
    February 21, 2012 13:24

    Having fabs geographically close together presents a business continuity risk. A major natural disaster (or less likely, war) in Taiwan could devastate TSMC. Co-location is kind of like the Euro–everyone gains in good times, but lack of mitigation options means bad times can be really bad.

  3. David Manners
    February 21, 2012 11:18

    Yes indeed mgp-1, but it was the financial community who pressed the fab-lite bollox on the IDMs. In the ST press release announcing Carlo Ferro’s appointment to be COO of ST-E yesterday it said he’d closed nine fabs – as if this is some kind of meritorious achievement. The same with NXP in their financial results a couple of years back listing all the manufacturing sites they’d closed as though this was something to be proud of. It is those CEOs who are most vulnerable to the pressures from the financial community who have swallowed the bollox the most enthusiastically.

  4. David Manners
    February 21, 2012 11:14

    Yes [Anonymous] I was reliably informed yesterday that GloFo’s 28nm process has transitioned from yields measured in wafers-per-die to the more conventional die-per-wafer. A step in the right direction. However, with these fiendishly difficult advanced nodes, one wonders if the GloFo strategy of having multiple fab-sites dispersed around the world is going to be as effective as the TSMC strategy of keeping all the fabs in one palce. TSMC say that it’s a big plus for them having all their fabs close together. And of course you’re spot on right about the silliness of the ‘vanilla’ or ‘commodity’ label put on advanced digital CMOS. Utter bollox – like so much that spews from the bankers.

  5. Anonymous
    February 21, 2012 06:31

    It is getting interesting isn’t it? When fab-lite was taking off there was a lot of talk about ‘vanilla’ (aka commodity) CMOS, implying little value in manufacturing technology.
    Success at manufacturing 20/22nm will be immensely difficult. The IDMs/foundries will have needed to have selected the optimum ‘flavour’ or will end up peddling a lemon.
    It will be interesting to see who will succeed. You would not bet against TSMC. Really interested to see how it goes for GloFo – they have a bucket load of money and people to throw at the problems. (Pleasantly surprised to see how many ex-colleague now show in LinkedIn under GloFo).

  6. mgp-1
    February 20, 2012 15:57

    Not just the financial community David … but all the fab-lite CEOs subscribing to the notion that fabs have ‘no strategic value’. We’ve been warning this was a disaster waiting to happen for years … and made many enemies in the process … yet still no-one seems to be worried! If you think 28nm is bad, just wait for 20/22nm. As Bob Marley would have doubtlessly sung “No Wafers No Cry”

  7. February 20, 2012 10:47

    Good article, and we agree with your final point: “how ineffably silly it was of the financial community to press their view”

Share your knowledge - Leave a comment