So it seems both TSMC and Intel will have 10 nm in mass production in 2016 according to plan.Taiwan Semiconductor Manufacturing Company (TSMC) will operate at full capacity in the third and fourth quarters, with revenues generated from 20nm processes to account for 10% of total revenues in the third quarter and rise to 20% in the fourth, according company chairman Morris Chang.
[...]
TSMC currently has a total of over 20 tape-out clients for its 16nm FinFET process, the company noted.
[...]
Meanwhile, TSMC plans to set up a special R&D unit, which will be powered by 300-400 technicians, to develop 10nm processes, aiming to push the process into trial production in 2015 and mass production in 2016.
TSMC have been mass producing 20nm since 2012, not to mention 28nm since 2010 and 16nm FF since 2013
TSMCs foundry roadmap cant be used for anything when compared to actual products(Or anything at all really). Intels roadmap is a product roadmap.
And I assume what TSMC calls 10nm is actually the real 14 or 16nm process.
Remember, we are still waiting for 20nm products today.
TSMC have been mass producing 20nm since 2012, not to mention 28nm since 2010 and 16nm FF since 2013
TSMCs foundry roadmap cant be used for anything when compared to actual products(Or anything at all really). Intels roadmap is a product roadmap.
TSMC may not even ship its 20nm with FF, aka what they call 16nm with FF before sometime in 2016 in terms of real products. And I assume what TSMC calls 10nm is actually the real 14 or 16nm process.
Remember, we are still waiting for 20nm products today.
TSMC have been mass producing 20nm since 2012, not to mention 28nm since 2010 and 16nm FF since 2013
TSMCs foundry roadmap cant be used for anything when compared to actual products(Or anything at all really). Intels roadmap is a product roadmap.
TSMC may not even ship its 20nm with FF, aka what they call 16nm with FF before sometime in 2016 in terms of real products. And I assume what TSMC calls 10nm is actually the real 14 or 16nm process.
Remember, we are still waiting for 20nm products today.
TSMC may not even ship its 20nm with FF, aka what they call 16nm with FF before sometime in 2016 in terms of real products. And I assume what TSMC calls 10nm is actually the real 14 or 16nm process.
Intel will launch 10nm products around mid-2016, while TSMC won't have their 14nm Intel equivalent for about 2 years.
Sure, but how are the yields?Intel is ramping Broadwell at 14FF just as TSMC has been in volume production of 20nm from January.
Source? Today I read a recent article predicting Samsung is the supplier of the A8.The lion's share of 20nm wafer allocation has gone to Apple.
That isn't really steep. In fact, most of TSMC's revenue comes from old nodes.The production ramp at TSMC 20nm is steep from Q3 2014. 10% of wafer revenue in Q3 and 20% of wafer revenue in Q4 from 20nm.
The world doesn't only consist of Apple. The rest of the world, ~85% or so, will have 20nm only in H1 2015 with the SGS6 etc.The truth is by the time Intel has Broadwell available for purchase, Apple will be selling millions of iPhone 6 and iPad Air 2 with the A8 chip which is manufactured at TSMC 20nm.
We'll see.TSMC 16FF/16FF+ comes a year later so we can expect the Apple A9 chip to power the iPhone 6s and iPad Air 3 by Q4 2015. Qualcomm's 16 FINFET modems should start shipping in mid - 2015. Depending on how successful Samsung is in ramping 14 LPE we could see Apple multi source A9 from multiple foundries.
Not true. Intel's lead is bigger than ever. Intel has 22nm products widely available since mid-2012. 20nm? H1'15. FinFETs? H1'16 or later, 3 years after Intel.Intel's lead at 14nm FINFET is roughly a year as opposed to the multi year lead which they had till the 22 FINFET node.
Intel is in a unique position. Intel with its ever growing manufacturing leadership will take away all cutting edge revenue by gaining massive market share in all important areas of computing + shrinking the gap between IGPs and dGPUs. While Intel's income will grow, TSMC, Samsung and GlobalFoundries will each fight for the remaining dies. The leading edge market will further consolidate as necessary.The future of the foundries lies in enabling Apple, Qualcomm, AMD, Nvidia and other fabless companies to compete and take market share from Intel in the various segments that they compete. The only way for the foundries to survive is to take semiconductor chip business from Intel as the cost of fabs is getting too high (USD 5 - 7 billion). TSMC does face competition from Samsung /GF at 14 FINFET and future nodes. Almost every fabless company wants multiple sources to avoid supply shortages and having to wait in a queue behind Apple.
Look at my post above. TSMC is doing some 'magic tricks' to make their ever growing gap with Intel look better. Node names are not more than marketing names. Instead of naming their node like traditionally, where it indicates a new step of Moore's Law, TSMC decided it was okay to take a new node for just releasing a new transistor technology without any increases in density, which is exactly what they will do at 16nm. So now we will get the better transistor sooner than before (half a year), but we'll have to wait 2 years after FinFETs for a new node with improved density. In the meantime, TSMC will release a 'FinFET Plus' with 15% scaling.So, around 18-20 months after they release their first 14nm products? I doubt it. They've been trending a little over 24 months for the past few nodes (like around 26-27 months), I don't see them vastly undercutting this now. I expect first 10nm products from Intel early 2017 or late 2016 at the earliest.
Except then that TSMC's comments have been proven to be not factual (it looks more like a 1.65x improvement).Also, TSMC is saying that the density increase from 16nm to 10nm will be more than a typical node shift, so I think they'll be pulling in some of the slack introduced by not or barely increasing density with 16nm over 20nm. Then again, Intel is saying their 14nm will increase density more than usual too, so that could cancel it out.
Good to see that TSMC is adequately concerned about their competition to keep up their roadmap fantasies. Though I'd hope that final statement is merely a misinterpretation? If they're just now setting up a special R&D unit to develop their 10nm process how do they expect it to be anywhere near ready in 2016? I'd guess what they're intending to convey is that they're setting up something similar to Intel's D1D development fab and it's merely taking over development.
It definitely is a nice, concise article though. It also hits on the important points regarding capacity ramp which many prefer to ignore in their analysis of the process node race. TSMC wants to claim that they've been in mass production of 20nm since the beginning of this year... and yet it's only going to account for a measly 20% of their revenue in Q4. By comparison 22nm accounted for a quarter of Intel's revenue in Q2 2012 (the quarter when it was released) and, according to those articles, was projected to rise to half of Intel's revenue in Q3 2012. If Intel released products at the same level of process maturity as TSMC they'd be anywhere from half to a full year further 'ahead'. (Of course Intel doesn't do that because, unlike TSMC, they don't sell wafers, so defect density actually affects their margins.)
Anyway, I know that I'm certainly looking forward to seeing what actually comes of TSMC's roadmap. We've been stuck on 28nm for far too long - 2.5 years and counting. Which is another reason why their claims of such an accelerated schedule are so bizarre - how precisely do they plan on recouping their investment for these interim processes that are supposedly only going to be around for roughly a year? Sure they're a necessary stepping stone for the next one, but it still seems quite peculiar. Especially when well over half of their revenue continues to come from 28nm and larger legacy processes.
OK now I am confused. What Intel node will TSMC's 10nm node be most similar to?
I generally agree, but 1H 2018 is too optimistic for 10nm. The only company you can expect a 2 year cadence from is Intel.Look, the foundries are just doing what they're best at -- fabrication. Period. Now that Samsung is thumping its chest loudly, TSMC needs to thump louder.
Realistically, expect the first 20nm non-Apple SoCs (i.e. from vendors that can't afford to pay for crap yields) to come in 1H 2015 (per Qualcomm's public roadmap), then the first 16 FinFET mobile chips to come in 1H 2016. Then expect 16 FinFET+ in 1H 2017, and then finally 10nm in 1H 2018.
I generally agree, but 1H 2018 is too optimistic for 10nm. The only company you can expect a 2 year cadence from is Intel.
I've seen you say this in another thread too. Why do you think there's a chance TSMC will cancel their 16nm process? They've spent most of the money developing it and I've heard a lot of rumors of several vendors waiting on it instead of going with their 20nm.
The only company you can expect a 4 year cadence from is Intel.
Intel's tick-tock cycle refers to manufacturing node and architecture, not specific SKUs.
Long term it's effectively the same.
I just realized, however, that TSMC might use EUV for their 10nm node, so they might outperform current predictions and make 10nm look a bit more like Intel's 10nm, density wise.
If you make the ridiculous mistake of picking the worst case scenario...Fixed that for you.
Haswell-K -> Broadwell-K is 2 years. Multiply that by 2x to get a complete tick-tock cycle and you get 4 years.
That article says that the EUV unit fried. The implications are the opposite of the way you took it. A $150 million dollar machine went "kablooey."Huh, I guess that TSMC actually has stated that they're still planning to use EUV at '10nm' - http://www.theregister.co.uk/2014/02/25/asml_euv_scanner_fails_at_tsmc_and_intel_investigates_dsa/ - which, well, good luck to them. Granted I can understand why they'd take that route as using 193nm lithography for their '10nm' node may well not be economically feasible for their model. But given EUV's track record relying on it for their '10nm' node could easily result in a delay of multiple years, and that's assuming that they do eventually get it to work.
(As a side note, I also wonder how much that approach on TSMC's part is a result of Intel's silence regarding their 14nm process node?)
Skylake in summer next year. So obviously not.