Samsung always is a cheater. They cheat on benchmarks, and they cheat on what their process is called.
:) Yeah, they cheated on benchmarks, but Intel isn't exactly the model citizen here either...
themselves and the idiots like monk who post that nonsense on this board.
:) If you say so. While it's true there is a whole set of marketing nonsense around node designations, you have to look deeper to get the full picture.
The reality is that Samsung has dropped QCOM for it's S6 launch. Why? The suggestion is that it's powered by a 14nm FinFet SoC and if this is true (which it may not), it goes to explain some of the impressive performance numbers. This formal announcement also adds weight to Apple going back to Samsung for the A9.
[Is English your second language? Can you translate that out of Gibberish? ]
:) When will Intel's product (that goes up against the 7) arrive?
Wallis, so when do you expect Intel's "real" 14nm Finfet product that will go up against the 7, arrive?
"As the most advanced technology available today, 14nm FinFET process is able to achieve the highest levels of efficiency, performance and productivity. When compared to Samsung’s 20nm process technology, this newest process enables up to 20 percent faster speed, 35 percent less power consumption and 30 percent productivity gain."
"Samsung’s leading-edge14nm FinFET process will be adopted by its Exynos 7 Octa, then expanded to other products throughout the year."
How can you design "something" without understanding the manufacturing aspect?
In particular as geometries shrink and tolerances get smaller and smaller - the fabless model lacks the close interface/feedback between manufacturing and design - regardless of what Mr. nenni says the IDM model is not dead quite the opposite - the strength of the IDM model is becoming more apparent
Er, no they don't - the prime IP suppliers work very closely with the fabs. This is the point of ARM's POP IP - hard cores for a specific foundries process. ARM bought Artisan a few years ago to bring them closer to the manufacturing details.
This simply doesn't sound right on the face of the article. Sale have been good for TSMC so now they ave decided to add 16 billion in upgrades assuming that it will be ready fast enough to extend the ride they've caught recently?
Well, they did mention new green field sites, so some of this cash would be on new plants. What do you think will happen to all that Apple capacity once Apple move over to Samsung later on in the year?
And no bunch of stinking scientists is going to convince me otherwise!!!
Open your eye's man...
"This website is dedicated to unraveling the true mysteries of the universe and demonstrating that the earth is flat and that Round Earth doctrine is little more than an elaborate hoax.
The Flat Earth Society has dedicated itself to starting science afresh from the ground up, to begin to see the world without bias and assumption. Experiment and experience has shown that the earth is decidedly flat. Time and time again through test, trial, and experiment, it has been shown that the earth is not a whirling globe of popular credulity, but an extended plane of times immaterial."
he still drives on the wrong side of the road and will continue so... for NO reason
he googles TSMC FinFet ramp.... but has no clue about AMAT
Monk is used to the fabless model IOW completely clueless about the manufacturing aspect .
Brits are brits .... and so is ARMH
Er, 35% of the worlds population drive on the 'wrong' side:)
I admit that I know more about the IP side of the business than manufacturing and as such I could be completely wrong on the ramps that I have suggested. But you only have to open your eyes to see what is coming (such as the 810 and the 60 odd device wins), the new Exynos's and Apples next gen. Clearly, they aren't made out of thin air.
(1) What is their yield on 20nm?
(2) Can they execute on Finfet at the same speed?
(3) What yield can they get initially and what timeframe will they get to decent yields?
1) No idea
2) That's my understanding, but that could just be spin. I do know that ARM (the company) does support this node (16nm) with it's low level libraries and POP solutions (which it didn't at 20nm), so at some point the bulk of the ARMy will be hitting this node with lots of different designs.
3) Don't know, but I have seen the suggestion that as16nm has the same metal backend then yields should be a less of an issue. No idea if this is true in practise.
By the way, Monk, Ash was totally and completely wrong in his assessments of Intel needing to do mobility on the time frame he suggested in order to be successful. He was so horribly negative about Intel he convinced one of the guys on iHub to sell half his shares just before Intel moved up very significantly.
Was this person who sold flumuxed?
I do agree with you on this, Ash was wrong about the timescales, but I completely disagree about him being negative. Yes he was critical of Intel's decisions and roadmaps, but given Intel's history with mobile, do you blame him?
So, are you going to be the champion of every ARM public figure who exploded their views on the internet but was wrong?
Not at all, I was merely pointing out your behaviour, is, well, a bit odd.
So, why don't you put a sock in your made up fiction about me chasing Ash all over the internet. My god, with all the horrible, horrible, awful, horrible stuff I've done and said you would think you could find something factual inside of manufacturing commentary that is blatantly false
Those 2 posts where in the first 5 I looked at from you. Perhaps I should have looked at more to get a better sample? What do you think?
TSMC came out their 20nm late 2014 (only for Apple as far as we can tell and that too, with rumors of very low yields). Now they are going to do 16nm the within 12-months of coming out with their 20nm.....?
(Keep in mind, SE_Junkie is very emphatic that based on conference call transcripts of equipment companies, most of this $16bn investment is going into 28nm, not 20nm or 16nm finfets.)
So never mind what the Google search says. Putting your credibility on the line, BRM, how would you call it? Is TSMC going to come out 16nm Finfets in the second half of 2015?
I don't have any inside information regarding SoC delivery for 2015, but I will attempt to answer your questions.
As has been pointed out by others, TSMC's backend for 16nm is the same as 20nm, and as such the move to the new node is straightforward for TSMC (think about all that Apple capacity that won't be needed soon...). If you think this move is aggressive in the timescales, you don't want to know about 10nm:)
Regarding SE_Junkie, I don't doubt what he says - or others like him - however, you only have to look at the TSMC ramp of 20nm of Apples SoC to see how people can be wrong.
Given what I've read, this is what I expect to see from the ARMy in 2015:
Samsungs 14nm finfet in the S6 and derivatives
QCOM's 810 on TSMC's 20nm
Apples A9 on Samsungs 14nm finfet
And that's basically it. I've seen some noise about the 820 which would be on 16nm finfet for later on in the year, but this would be very aggressive. I expect most TSMC's ramp for the ARMy to be in 2016.
Yields are so low that only apple can afford them it seems.
Why do you think yields are low? QCOM's 810 will be here soon enough in huge numbers...
(1) Do we know which node the $16bn build-out is for? 28nm or 20nm or 16nm....or a combination? If it is combined investment, how much would TSMC be investing in each of these nodes? Of course, the reason for the question is: Intel's 14nm has been up and running since last year. When will TSMC's 16nm start delivery?
I did a quick google from the original article. Frankly it's not clear, part for the 16nm ramp, but they are also talking about new factories (with the suggestion it's for 10nm so they can can catch up to Samsung).
As for delivery of 16nm, a google says second half 2015 for volume production.
(2) Assuming they buy the $16bn worth of equipment throughout 2015, when will the equipment be delivered, installed, tested, and be ready for production?
If this is for a new plant, it's years away but it looks like this 16 number is split.
(3) Would the production-ready dates be different for 28nm vs 20nm vs 16nm? Would you care to hazard a guess for each of them?
No idea. But it's not a clear cut question, for 20nm it's just one process (I believe), with, what 4 for 28nm? Then there is a couple at 16nm as well. From an IP point of view, ARM is talking 2016 for it's latest gear (from the likes of Rockchip and MediaTek) at TSMC's 16nm. So you would expect the likes of Apple and QCOM to be earlier than these guys.
[This is exactly what I mean about you making things up. I haven't posted on iHub in just about forever.]
"wallisweaver, you really need to stop trolling me. I am simply asking if they've used fairly recent models; the improvement in 2013 phones over 2012 ones is pretty staggering."
"Now, wallisweaver, I am done with you. I am putting you on my ignore list, so please don't bother responding to anything else. You, unfortunately, have made it clear that you wish to attack me."
Both of the above where from Ashraf, there is plenty more as well.
[This is unmitigated BS and you know it. I've documented the reasons for my opinions and backed them up extensively. The idea that I've followed Ash or anyone around the net is totally and completely false. The people I've done battle with have all been here. If you have anything to the contrary please post it up. You can't because there isn't anything. If you have honest criticism post it up. But try to avoid the sheer making up of things that don't exist. It makes you sound just like Twink, who is as dishonest as the day is long. ]
Wallis, I regularly read the investorshub intc forum. However, as I don't have an account, I can't search for your posts to show the 'battle' you had over there. I am sure you can though...
Actually they are adding 28nm capacity which is going to last very long - but won't cut it for for ultra low power.
TSMC always "talks big" opposite to Intel - Intel was very open when reporting yield problems.
compared to Intel i find TSMC, Samsung, GF credibility very low - Nenni reflects perfectly TSMC's "PR"
Yeah, 28nm will last a long time, as will other older node such as 65%. As for yields, what numbers would you like to see? Remember the conversations with the 'yield' issues with TSMC at 28nm? One design can have a different yield rate than another and as TSMC has lots of different designs to fab...
As for low power, I assume you are talking about mobility? Given that 20% of TSMC's revenues are on 20nm (Apple and QCOM's ramp) and given what we know for the designs of TSMC 16nm, I think we can safely say that TSMC will do lots of wafers at 16nm.