Reality Check: Intel's Process Lead Examined

There's been a lot of fear, uncertainty, and doubt going around lately with respect to the progress that the third-party semiconductor foundries have been making on next-generation process technologies. While Taiwan Semiconductor (NYSE: TSM  ) , Samsung (NASDAQOTH: SSNLF  ) , and others have been claiming that everything is going swimmingly for production during 2014, the reality from the semiconductor equipment vendors' point of view is a different one that should allow Intel (NASDAQ: INTC  ) investors to breathe a sigh of relief. 

KLA-Tencor and ASML spill the beans
While TSMC, Samsung, and others interested in promoting the foundries claim that TSMC and Samsung are seeing "great" yields for their FinFET devices, KLA-Tencor (NASDAQ: KLAC  ) made the following remarks on its most recent earnings call when explaining why its logic customers are pushing out FinFET equipment purchases (emphasis mine):

Now in logic and foundry with the introduction of the new 3-D gate architectures, the yield issues our customers are grappling with today are proving to be the most challenging that the industry has ever faced and even the smallest variation in process margin can cause significant yield losses for these devices.

So, according to a leading equipment vendor, the foundries are facing the most challenging yield issues that the industry has ever faced, yet according to TSMC, everything is going swimmingly with its 16-nanometer FinFET process. Now, KLA-Tencor isn't the only one to claim significant uncertainty surrounding FinFET production at the foundry; leading lithography vendor ASML (NASDAQ: ASML  ) chimed in on its most recent earnings call echoing similar thoughts (emphasis mine):

With that, I would like to turn to our expectation for Q2 and Q3. Let me start with a view at the markets we serve. We are experiencing solid demand in memory with some uncertainty on the ramp of 3-D versus planar NAND, which is changing our product mix slightly. IDM is experiencing strong year-over-year growth as expected. And in foundry, we see strong demand for the 28-nanometer node, but expect some adjustments to shipments for the continued ramp to volume production for both 20-nanometer planar transistor and 16-namometer plus 14-nanometer FinFET nodes. These adjustments are impacting our revenue forecast for Q2 and Q3.

In short, while TSMC and Samsung talk a big game with respect to their FinFET nodes, the truth is that the foundries are having a difficult time getting the yields to be passable and seem to be quite a way from production. The question, then, is how far from volume production are the foundries?

Expect production to begin in mid-2015
On ASML's most recent call, CEO Peter Wennink had the following to say with respect to the lead times in equipment orders and ramp to production:

Now, if they have some breakthroughs, they will pull things in, because they know on my earlier question ... what's the time that they need? They need about nine months to six months of our production time and then about one quarter of installation and ramp-up time.

So, once the yields on the manufacturing process are ready to go, the equipment vendors will need about six months to actually build the equipment and then an additional quarter (about three months) to get that equipment up, running, and ramped. If ASML believes that both Q2 (current quarter) and Q3 will be weak (implying orders coming in during Q4), this would imply the beginning of the ramp in Q2 2015.

Given the time it takes to actually produce product -- and assuming that the designs from TSMC's and Samsung's customers are ready -- this would imply product in the market in late 2015 or early 2016 if everything goes as planned. This is roughly in line with TSMC's claims that 16 FinFET would be in volume production roughly one year from the beginning of volume production of 20-nanometer.

Circling back to Intel
On Intel's most recent earnings call, CEO Brian Krzanich confirmed that Broadwell -- Intel's first 14-nanometer product -- began production in Q1 2014. If TSMC and Samsung hit their goals, then on a named node-to-named node basis (that is, Intel 14-nanometer relative to TSMC/Samsung 16/14-nanometer), Intel is about one and a quarter to one and a half years ahead of TSMC. However, there is some subtlety required here.

TSMC and Samsung are essentially grafting their first-generation FinFET transistors onto the same metal stack (which is a key determinant of process density) as their 20-nanometer processes, implying that there will be a minimal density improvement in going from 20-nanometer to 14/16-FinFET. Intel, on the other hand, plans to do a "true" 14-nanometer process. In particular, Intel should have tighter gate and minimum metal pitches than TSMC/Samsung so its 14-nanometer will be "ahead" of its competition.

Foolish bottom line
While you will see some screaming about Intel's losing its manufacturing lead, the bottom line is that Intel will have a denser and probably higher-performing 14-nanometer process, probably with high yields (thanks to learnings from the 22-nanometer FinFET generation) and with multiple flavors of products on the shelves, before TSMC and Samsung ship their first-generation FinFET products.

Now all Intel needs to do is release mobile chips that people actually want, something that maybe will happen during the course of 2015.

The biggest thing to come out of Silicon Valley in years
If you thought the iPod, the iPhone, and the iPad were amazing, just wait until you see this. One hundred of Apple's top engineers are busy building one in a secret lab. And an ABI Research report predicts 485 million of them could be sold over the next decade. But you can invest in it right now, for just a fraction of the price of Apple stock. Click here to get the full story in this eye-opening new report.


Read/Post Comments (17) | Recommend This Article (5)

Comments from our Foolish Readers

Help us keep this a respectfully Foolish area! This is a place for our readers to discuss, debate, and learn more about the Foolish investing topic you read about above. Help us keep it clean and safe. If you believe a comment is abusive or otherwise violates our Fool's Rules, please report it via the Report this Comment Report this Comment icon found on every comment.

  • Report this Comment On May 20, 2014, at 12:50 AM, DaMuncher wrote:

    Broadwell products in 14nm will hit the market after Sept. but before Christmas.

    http://www.pcworld.com/article/2156741/intel-guarantees-dela...

    Apple A8 products in 20nm will presumably be launched in Sept. for Christmas delivery as well. That will be the next interesting comparison point for process technology.

    Until then, the real numbers to watch will be unit and margin numbers for Intel's 40M tablet program.

  • Report this Comment On May 20, 2014, at 2:30 AM, raghu78 wrote:

    Ashraf

    Intel's 14 nm started production in late Q1 2014. By that it means production wafers in towards the end of March. Since it takes atleast 3 months from wafers in to testing and packaging those chips are out in late Q2 2014. Intel requires atleast 3 months of production to build enough volume for a launch. So late Q3 to early Q4 2014 for Broadwell release. thats eactly what brian krzanich confirms by saying Broadwell on shelves for holiday

    http://www.dailytech.com/Intels+Broadwell+Will+Likely+Slip+P...

    http://semiaccurate.com/2014/02/19/sky-falling-intels-14nm-b...

    So if TSMC/Samsung start production by late Q1 2015 they are exactly a year behind Intel. If they start production in late Q2 2015 they are a year and quarter behind.

    Also you are again spreading FUD that Intel has significant transistor density and performance lead. Its density lead if anything is negligible (most likely 10%) as all the companies are using dual pattern immersion litho with a M1 metal pitch of 64nm. TSMC has already stated that their 16FF+ matches Intel 14 FINFET in performance.

    https://markets.jpmorgan.com/research/email/eopsgtl5/GPS-136...

    https://markets.jpmorgan.com/research/email/-kjegkq4/GPS-133...

    http://www.tsmc.com/uploadfile/ir/quarterly/2014/1aT1b/E/TSM...

    The ASML question was raised in TSMC Q1 2014 earnings call. see the answer on page 6. btw the majority of designs ramping in 2015 will be on 16FF+.

  • Report this Comment On May 20, 2014, at 10:22 AM, TMFAeassa wrote:

    raghu78

    How do you know Intel is using an M1 pitch of 64nm? Are you just assuming that, or have you actually checked with somebody who knows?

  • Report this Comment On May 20, 2014, at 12:54 PM, raghu78 wrote:

    Ashraf

    The information from people who track or follow these companies

    http://www.bnppresearch.com/ResearchFiles/31175/Semiconducto...

    page 11

    Intel's M1 seems to be 63nm. so minor density advantage over Samsung 14 FINFET or TSMC 16FF+. TSMC 16FF will be at a die size disadvantage but as TSMC highlighted the majority of their customer designs ramping in 2015 will be 16FF+ .

  • Report this Comment On May 20, 2014, at 1:05 PM, TMFAeassa wrote:

    raghu78

    In other words, you have not verified this information yourself.

    Tell me: why would Intel explicitly highlight aggressive BEOL scaling at its Investor Meeting and claim a 35% density advantage (knowing full-well TSMC's 64nm M1 for 20nm), if it had only a minimal BEOL scaling advantage over the foundries?

    At any rate, I assume Intel will reveal the details of its 14nm process soon enough.

    p.s. on TSMC's Q1 CC they didn't even refute ASML's claim, just claimed that 16 FinFET would begin to ramp in 2015, which is exactly what I am claiming in my article.

  • Report this Comment On May 20, 2014, at 1:19 PM, TMFAeassa wrote:

    @ raghu78

    I should also add that I looked through that research report and found no actual basis for the claim that Intel's 14nm M1 is 63nm. It's just thrown into the chart with no justification.

    My guess is that this number comes from 0.7*90nm (M1 pitch for 22nm Intel process). However, Bill Holt explicitly claimed a greater than 0.7x shrink in M1 in his speech at Investor Meeting.

    I would strongly suspect that it is more like 0.6 * 90 = 54nm.

    If it is 54nm then this would give us roughly the 30-35% advantage that Intel is claiming, and makes a lot more sense than the 63nm number in that report.

    As always, we'll know soon enough :-)

  • Report this Comment On May 20, 2014, at 2:37 PM, raghu78 wrote:

    Ashraf

    Intel's so called process advantage is yet to materialize at least in terms of density. In terms of transistor performance they do have the lead over 28nm hkmg planar. but that is due to FINFET.

    Explain this. Intel Baytrail built on a 22nm FINFET process with a contacted poly pitch of 90nm had no significant density advantage over Apple A7 built on a Samsung 28nm hkmg process with a contacted poly pitch of 120nm (which is equal to TSMC 28nm contacted poly pitch). Both Apple A7 and Baytrail have roughly a billion transistors and measure around 100 sq mm (thats a rough guess given how secretive Intel has been on actual die size and transistor counts for Baytrail).

    http://www.chipworks.com/en/technical-competitive-analysis/r...

    http://www.chipworks.com/en/technical-competitive-analysis/r...

    http://www.chipworks.com/en/technical-competitive-analysis/r...

    According to the BNP report , Samsung is touting 77 - 78nm contacted poly pitch at 14 FINFET. TSMC 16FF will be larger with a 90nm contacted poly pitch while TSMC 16FF+ brings 15% density scaling over TSMC 16FF, so around 79 - 80nm.

    Anyway I am waiting for a full disclosure by all the companies at conferences like IEDM 2014

    http://www.his.com/~iedm/general/

  • Report this Comment On May 20, 2014, at 2:52 PM, TMFAeassa wrote:

    raghu78

    raghu78

    "Intel's so called process advantage is yet to materialize at least in terms of density. In terms of transistor performance they do have the lead over 28nm hkmg planar. but that is due to FINFET."

    Sure, that's fair, but from what I have heard, Intel's density issues at 22nm were due to problems on the design side of the house moreso than any process deficiency. That said, in terms of shipping product I don't disagree with you.

    To illustrate my point on how important design is to density look at Apple A6 v.s. A7. Did 28nm bring ~2x the density in terms of contacted poly pitch or metal? Of course not, and yet the A6 offered ~2x the density of the A7.

    "According to the BNP report , Samsung is touting 77 - 78nm contacted poly pitch at 14 FINFET. TSMC 16FF will be larger with a 90nm contacted poly pitch while TSMC 16FF+ brings 15% density scaling over TSMC 16FF, so around 79 - 80nm."

    OK, so if Intel 22nm offers a 90nm contacted poly pitch, and if Intel has already made clear they will get a greater-than-normal shrink, then we can assume Intel's 14nm contacted poly pitch will be somewhere south of 63nm -- again implying a pretty healthy density lead.

    We will have Broadwell in the market probably in the late October/November timeframe and I am certain Chipworks will tear one of those puppies down to give us the details :-)

  • Report this Comment On May 20, 2014, at 2:53 PM, TMFAeassa wrote:

    ** A7 offered 2x the density of A6

  • Report this Comment On May 20, 2014, at 3:35 PM, raghu78 wrote:

    Ashraf

    yeah a chipworks teardown of broadwell should provide a lot of information. I am of the opinion that Intel 14nm will have 10% density advantage over Samsung 14FINFET and TSMC 16FF+ . anyway we will see how this unfolds over the next 12 - 18 months. btw the bigger problem for Intel is the transistor performance gain at these foundries due to FINFET

    TSMC data from Q1 2014 earnings call pdf

    TSMC 16FF+ - 15% faster than TSMC 16FF, 38% faster than TSMC 20 SOC, almost 60% faster than 28HPM

    TSMC 16FF - 20% faster than TSMC 20 SOC, 38% faster than TSMC 28HPM

    So at same power the speed is

    TSMC 28 HPM - 1x

    TSMC 20 SOC - 1.15x

    TSMC 16FF - 1.38x

    TSMC 16FF+ - 1.587x

    Power at same speed

    TSMC 28 HPM - 1x

    TSMC 20 SOC - 0.7x (30% lower than 28HPM)

    TSMC 16FF - 0.45x (55% lower than 28HPM)

    TSMC 16FF+ - 0.32x (30% lower than 16FF)

    Samsung 14LPE matches TSMC 16FF and Samsung 14LPP matches TSMC 16FF+ on performance

    So Intel's headaches get even worse starting in late 2015.

  • Report this Comment On May 20, 2014, at 4:44 PM, H2323 wrote:

    I bet you A8 is 28nm......Intels design fab advantage is meaningless due to them staying with x86 in mobile....useless.

  • Report this Comment On May 21, 2014, at 12:07 AM, SemiWiki wrote:

    Ashraf,

    You are betting on supplier conference call revenue guidance against TSMC executives committing to specific dates to their top customers?

    In other words, you have not verified this information yourself.

    Good luck with that.

    Dan

  • Report this Comment On May 21, 2014, at 1:41 AM, TMFAeassa wrote:

    SemiWiki

    The semiconductor equipment suppliers can't mask the fact that the orders haven't come in. TSMC and the like can make all of the claims that they would like, but at the end of the day, the semiconductor equipment vendors have to explain to their shareholders why they haven't booked the revenues.

    Do remember that you were telling us on SemiWiki that Intel was delaying their 14nm ramp because equipment move-in had been delayed and, not surprisingly, knowing this information allowed you to make a bold (but correct) call that Intel's 14nm ramp would be delayed.

    Follow the money trail...it's rarely wrong :-)

  • Report this Comment On May 21, 2014, at 9:39 AM, SemiWiki wrote:

    The 14nm delay info came from people who worked in the field, not from a carefully scripted analyst conference call, big difference.

    I will blog this today but you have been duped once again.

  • Report this Comment On May 21, 2014, at 11:20 AM, iblishsakora wrote:

    FYI...All I can say good luck to you all. Do not let Intel mirage of scaling down fool re_moltely fool you. n-3 TSMC HPM is still better than nth gen Intel. Much of the innovation will come from architecture in design yet still using. There is paradigm shift in process vs . design. $/transistor curve or transistor density not measure against complexity of devices.

  • Report this Comment On May 21, 2014, at 6:46 PM, TMFAeassa wrote:

    @ SemiWiki

    "The 14nm delay info came from people who worked in the field, not from a carefully scripted analyst conference call, big difference."

    No, it's not a "carefully scripted analyst call" -- it's the fact that Q2 and Q3 results are going to be pretty materially impacted as a result of this delay.

    You really can't hide the money not showing up, and it is not in the semiconductor equip companies' best interest to throw TSMC/Samsung under the bus unless they have to in order to explain why the dollars aren't showing up.

  • Report this Comment On May 22, 2014, at 6:11 PM, PDM wrote:

    I tend to agree that there was some scripting of KLA-Tencor and ASML commentary and answers, but only insofar as this is what was agreed between them and the customer(s) who are experiencing the difficulties that are leading to push-outs of equipment purchases.

Add your comment.

Sponsored Links

Leaked: Apple's Next Smart Device
(Warning, it may shock you)
The secret is out... experts are predicting 458 million of these types of devices will be sold per year. 1 hyper-growth company stands to rake in maximum profit - and it's NOT Apple. Show me Apple's new smart gizmo!

DocumentId: 2962137, ~/Articles/ArticleHandler.aspx, 9/18/2014 2:14:48 AM

Report This Comment

Use this area to report a comment that you believe is in violation of the community guidelines. Our team will review the entry and take any appropriate action.

Sending report...

Today's Market

updated 4 hours ago Sponsored by:
DOW 17,156.85 24.88 0.15%
S&P 500 2,001.57 2.59 0.13%
NASD 4,562.19 9.43 0.21%

Create My Watchlist

Go to My Watchlist

You don't seem to be following any stocks yet!

Better investing starts with a watchlist. Now you can create a personalized watchlist and get immediate access to the personalized information you need to make successful investing decisions.

Data delayed up to 5 minutes

Related Tickers

9/17/2014 4:00 PM
ASML $100.44 Down -0.79 -0.78%
ASML Holding N.V.… CAPS Rating: ***
INTC $34.98 Up +0.05 +0.14%
Intel CAPS Rating: ****
KLAC $79.57 Up +1.29 +1.65%
KLA-Tencor Corp CAPS Rating: ***
SSNLF $1175.00 Down -1175.00 -100.00%
Samsung CAPS Rating: ***
TSM $20.62 Up +0.14 +0.68%
Taiwan Semiconduct… CAPS Rating: ****

Advertisement