Intel's 14nm is not the same as Samsung's 14nm for a host of reasons, including the fact they are delivering a full node's worth of a real shrinking over that of their existing 22nm node.
Part of the reason no one else is able to deliver reduced xtor expense per node is because no one but Intel has invested in the tools necessary to enable a full shrink for the BEOL for the 14nm node.
Does Intel has the balls to publish the process/device specs of its 14nm node?
It isn't a bold statement.
We have Intel 32 and TSMC/Samsung 28.
Intel increases density and adds FinFET and calls it 22.
TSMC/Samsung increases density and then adds FinFETs afterwards, calls it 20 and then 16/14.
See what they did? 20nm should have been called 20LP and the FinFET version should have been called 20HP, just like they did with 28nm.
You links shows no such thing as 14nm or 16nm for AMD.
You make it sound like Samsungs 14nm is some big secret.
Lisa Su - SVP and COOHello John, so let me take a step at that. I think when you look at what's important to us, I mean clearly process technology is an important element but we have invested quite a bit in architecture, design techniques, new IP software. So, I wouldnt say that process technology is the first and primary determinants for us. It is important that we are on competitive technology, so we have said before and I will say again that 20 nanometer is an important note for us. We will be shipping products in 20 nanometer next year and as we move forward obviously a FinFET is also important. So, if you look at our business, it is quite a bit more balanced between the semi-custom, embedded sort of commercial PRO Graphics growth portions as well as the more traditional sort of client and graphics pieces of our business. So technology plays in all of those businesses.
But the rules are the same, right?
Does Intel has the balls to publish the process/device specs of its 14nm node?
Do you have any specs on Intel's 14nm FINFET or are you just arguing for the sake of it. Intel's 22nm FINFET has comparable density to foundries 28nm. The truth is Intel's products like Baytrail have not shown any significant die size advantage over foundries products like Apple A7 (Samsung 28nm) or AMD Beema( GF 28nm). All these chips are roughly a billion transistors and 100 sq mm in die size.
Samsung 14LPE gate last process brings a 1.9x density increase over their 28nm hkmg gate first process. It also brings a 51% power reduction at same transistor performance. 14LPP which comes later brings close to 20% power reduction over 14LPE and 60% power reduction over 28nm.
https://www.semiwiki.com/forum/content/3688-samsung-foundry-explained.html
TSMC has stated their 16FF+ matches Intel 14nm in transistor performance while being marginally behind on transistor density.
http://electronics360.globalspec.com/article/3974/tsmc-tweaks-16nm-finfet-to-match-intel
Samsung 14LPP and TSMC 16FF+ are comparable on both transistor density and performance. Both are competitive with Intel 14nm.
Intel, Samsung and TSMC are all using dual pattern immersion lithography and FINFET at 14/16 nm. Intel's lead which you talk about is only time to market by roughly 9 - 12 months.
FWIW i'm posting research summary from a wall st firm that took Bill Holt and Jason Waxman around for two weeks to meet with investors....
1) Intel cost lead likely extends at 10nm – due to Intel’s “Inverse Optical Lithography” capability
a. Intel’s 14nm MPUs are 6 months behind schedule, making smaller transistors is a real challenge due to the delay in EUV lithography tools
b. To make the smaller transistor using DUV, they had to develop an “Inverse Optical Lithography” capability
i. This means they now have to use “diffracted” light to print transistors on wafers, instead of direct light
ii. The need to use diffracted light created an optical engineering challenge that “exploded” at Intel at 14nm
c. Intel has now developed this this “Inverse Optical Lithography” capability in house, it is a huge competitive advantage
d. TSMC and Samsung will run into this problem at 10nm, because their “10nm” features will be the same size as Intel’s “14nm” features
Bottom line: Intel will gain a transistor cost lead for the first time ever at 14nm, and will extend that at 10nm as TSMC and Samsung struggle to solve this Inverse Optical Lithography challenge
2) Low Power ARM Threat in Servers Appears Overstated
a. One consensus bear thesis on Intel is that ARM will disrupt Intel in Servers like it did to PCs
b. The argument goes like this: “Power costs for servers are high, so low power ARM is the solution…and Cloud players are moving to commodity servers, so that puts downward pressure on ASPs and margins for Intel”
c. Intel showed that Cloud plays (Google, Amazon, Facebook) were actually migrating up the MPU stack to higher end server MPUs
d. Intel server MPU ASPs have increased in 7 of the last 8 quarters, and from $450 to $600 over the past several years, even as Cloud plays have grown in size
e. The premise of the bear argument is misplaced. The issue is TCO (Total Cost of Ownership), or Performance/Watt/Dollar, not pure power, and Intel’s transistor cost leadership keeps them way ahead on this metric
Bottom Line: the data throws a harpoon into the side of the “Low Power ARM disrupts Intel in Servers” bear case
Other Tidbits:
1) Intel is making semi-custom server chips for high-volume customers, creating another barrier to ARM in servers
2) Intel views the Grantley server upgrade cycle as a 7-to-8 out of 10, with Nehalem being a 9/10 and Romley being a 5/10
3) Rockchip is an example of how Intel is getting more creative/aggressive in prosecuting the low-power tablet/mobile market
4) Don’t be surprised to see an announcement from Intel on a III-V compound materials advancement this year, that would put them another leg up on its competitors
5) The Public Cloud players (Google, Amazon, Facebook) are not cannibalizing Enterprise. At least 75% of their workloads are consumer based
a. Search
b. Social Media
c. Shopping
d. Games
Interesting how you take TSMC's word on their transistor performances/densities being "comparable" when Intel hasn't published a thing about them. However, there was a post from liahos1 on this very forum that you may want to read on this topic. He got ahold of what seems to have been a very interesting research note from a sell-side shop that did an investor tour with Intel's Bill Holt.
http://forums.anandtech.com/showpost.php?p=36515598&postcount=10
Intel will publish the specifications for all to see in open forum in 1 month and 7 days, though.
Why has a thread about AMD, and its possible future new FX etc cpus, suddenly apparently turned into an Intel process tech vs Other IC manufacturers discussion ?
good question. the same people always end up in an AMD thread to take the thread off topic and turn it into an Intel worshipping exercise.
No, and TSMC doesn't have any specs either, so TSMC's statement's trustworthiness is zero.Do you have any specs on Intel's 14nm FINFET
TSMC has stated their 16FF+ matches Intel 14nm in transistor performance while being marginally behind on transistor density.
We don't know any launch dates of 20FF, even less about FF+ or LPP, nor do we know how the transistor characteristics compare to Intel (but I suppose it's comparable to 22nm), but we know density will be much less and still quite a bit (1.3x) less (14nm compared to FinFET+). But note that density is mainly important for cost/transistor, which we already know will be extremely bad for the foundries' FinFETs, while Intel's 14nm excels.Intel's lead which you talk about is only time to market by roughly 9 - 12 months.
Worshipping a company is silliness. However, AMD's future is pretty directly impacted by the existence of Intel, so any discussion of AMD's future/competitiveness without understanding where its principal competition will be is pretty worthless.
For example, if the thread were "Best Possible Scenario for Post-22nm Intel Mobile Processors", any discussion that did not include Qualcomm, as well as the various in-house efforts at the relevant OEMs, would be just as worthless. Of course, I am sure in that case, the tables would turn and such a thread would seem a Qualcomm-worshipping exercise because Qualcomm is so far ahead of every other competitor
No they are not, 20nm is planar when 16/14nm is FinFet. You cannot directly port 20nm planar IC design to 16/14nm FinFet.
No, and TSMC doesn't have any specs either, so TSMC's statement's trustworthiness is zero.
But note that density is mainly important for cost/transistor, which we already know will be extremely bad for the foundries' FinFETs, while Intel's 14nm excels.
No, and TSMC doesn't have any specs either, so TSMC's statement's trustworthiness is zero.
Samsung and TSMC have provided quite a bit of information about their process while Intel seems to be waiting for IDF 2014 to reveal details about their process. As for the claims I believe final products will settle the discussion once and for all.We don't know any launch dates of 20FF, even less about FF+ or LPP, nor do we know how the transistor characteristics compare to Intel (but I suppose it's comparable to 22nm), but we know density will be much less and still quite a bit (1.3x) less (14nm compared to FinFET+). But note that density is mainly important for cost/transistor, which we already know will be extremely bad for the foundries' FinFETs, while Intel's 14nm excels.
Why has a thread about AMD, and its possible future new FX etc cpus, suddenly apparently turned into an Intel process tech vs Other IC manufacturers discussion ?
AMD's future depends on 2 things -
1.) the design and architecture of their products which directly affects perf and efficiency (perf/watt).
2.) Their foundry process tech which in this case is Samsung tech licensed by GF (since GF is the major foundry for AMD apu/cpu ).
AMD did well when they had a good design like K8 Athlon 64 and failed badly when they had a bad design (Bulldozer). AMD also had a lesser process gap (6 - 12 months) back in the old days upto 90nm. From 65nm onwards Intel raced ahead and AMD fell behind by roughly a full node (24 months). At Samsung/GF 14nm FINFET AMD will be able to close the process gap slightly.
That's actually not true. TSMC published a paper at IEDM 2013 detailing its 16 FinFET process. From a transistor performance perspective, TSMC's 16 FinFET process was extremely impressive. Where the process wasn't so impressive was SRAM density, which came in at 0.07um^2 for its high density cell.
Because that's how discussions work: they digress to other, related, subjects when necessary. I hope that answered your question.