Finally it seems someone can catch or at least come closer Intel fabs. I bet Samsung 14nm FinFET will come later than Intel 14nm, but not too late as many fabs will be. Seems like FinFet and 14 nanometers are both a big challenge for all foundries(include Intel), so Samsung is really doing the things very well. If they can still give continuity to Moore's Law on processors.
Yes... they hired Brian Klug en Anand Shimpi, basically halving AnandTech's number of great reviewers. Too bad they don't have the right fabs to get the full potential out of their chips.
I don't know if you've noticed, but Apple seem to be doing just fine with their chips. 20nm A8(X) is kicking ass.
Because Intel is still catching up. It doesn't matter if Intel has 5nm, if every other company if using 130nm and Apple has 90, sure it's going to perform well, but as great as it could. Never mind throttling.
The ones who are "catching up" are the foundries, not Intel!
Per Ultratech's October conference call:
As we have discussed on past conference calls, the difficult implementation of 3D FinFET microprocessors to high production manufacturing. Once again a major logic manufacturer delayed their FinFET ramp. We had then requested to prepare LSA tools for shipment for the end of the third quarter which was delayed. These LSA shipments for the most part caused our third quarter revenue to be less than projected. These LSA systems have been rescheduled for shipment in the fourth quarter.
There is anticipation of some minor ramp that were aware of in the fourth quarter. There is capacity in place currently. We have equipment that has been relegated to the 14 and 16 nanometer note, primarily 14. And so at this time we have capacity in place to take care of that need as they begin to ramp slowly. So we dont see a significant ramp in Q4 in FinFETs. We see that occurring later or in 2015 and its really hard to project when. The current anticipation is theyre all being overly optimistic as to when theyre going to solve their problems. But the yields on the major companies right now is in the 10% to 20%. And so it's not giving them much indication as to when theyll grow that problem, that area. So it appears to me right now and the problems arent really consistent in one area. They vary in the processing and possibly design.
I am reasonably confident that this "major logic manufacturer" is Samsung.
A$35-$40 apps processor would hurt Apple's bottom line quite a bit (i.e. impact measured in billions of dollars). I think Samsung is going to take the margin hit in order to win back Apple.
Also, re: the legality of Samsung's statements, the phrase "mass production" is oh-so beautifully vague
Samsung "14nm" and Intel "14nm" are hardly identical.
Samsung's 14nm offers a gate pitch of 78nm and minimum metal pitch of 64nm; Intel's 14nm offers a 70nm gate pitch and 52nm minmum metal.
You can see this delta play out in the SRAM cell sizes both companies have published. Intel's high performance SRAM cell size is 0.0588um^2 while Samsung's is 0.080um^2; Intel's high density SRAM cell at 14nm is 0.05um^2, compared to 0.064um^2 for Samsung.
They might both be labeled "14nm" but Intel's is quite a bit denser. To quote Charlie Demerjian of SemiAccurate:
Sram density is not logic and there is other metrics pointing to other direction. Yes Intel i probably denser but my guess when silicon hits its far less than 20%. Real difference is yield and performance. Lets see it might even be same density.
A8x walks over everything else in its market and we have a a9 on 14nm finfet comming. My guess is the difference will just get far bigger. 20nm is not a performance node.
It walks all over its competition for a couple of important reasons: they have an L3 cache, and their cores are much larger. They have the best performer simply because they are taking the profit hit of doing so. It's much easier for them to do this, as they can hide the cost over the entire device, and they also are getting their wafers for dirt cheap, relatively speaking, of course.A8x walks over everything else in its market and we have a a9 on 14nm finfet comming. My guess is the difference will just get far bigger. 20nm is not a performance node.
Yes... they hired Brian Klug en Anand Shimpi, basically halving AnandTech's number of great reviewers. Too bad they don't have the right fabs to get the full potential out of their chips.
It walks all over its competition for a couple of important reasons: they have an L3 cache, and their cores are much larger. They have the best performer simply because they are taking the profit hit of doing so. It's much easier for them to do this, as they can hide the cost over the entire device, and they also are getting their wafers for dirt cheap, relatively speaking, of course.
The SRAM cell size comparison is useful because it's as close to "apples to apples" on two different processes as it gets.
I don't disagree, and I'm actually very much looking forward to the A9, and other phones, making it to market with FinFETs. I'd really prefer them being on the market right now, but sadly Intel's dropped the ball on that.So what? The same could be said of Intel server solutions. A9 will just build on that strenth even more when it aquire even more performant and expensive finfet.
I don't disagree, and I'm actually very much looking forward to the A9, and other phones, making it to market with FinFETs. I'd really prefer them being on the market right now, but sadly Intel's dropped the ball on that.
However, the server market doesn't apply here. Intel can't hide their costs, since they're supplying the components directly. They also don't have two foundries fighting over it, like a divorced couple fighting for custody of their kid.
They kind of do. Apple's goal is to be the best in every category possible (except price, lol).Intel can spread process development cost between the entire portfolio. I dont know if you can do that a correct way because what is the correct distribution of cost from a business perspective?
Apple shares with other customers like qcom and microwaves so to speak. They do that on an open field. They pay what should be paied. Its not like samsung or tsmc are small suppliers - they got the cash to say no. But the one with the strongest brand always wins most because apple, as crazy as it is, doesnt need a super fast cpu.
They kind of do. Apple's goal is to be the best in every category possible (except price, lol).
And no, Apple gets special pricing. I'm sure Qualcomm does as well, to some extent. They have the kind of volume to negotiate such deals. And they should, given how economies of scale work.
They kind of do. Apple's goal is to be the best in every category possible (except price, lol).
And no, Apple gets special pricing. I'm sure Qualcomm does as well, to some extent. They have the kind of volume to negotiate such deals. And they should, given how economies of scale work.
I think the reason Apple can get better deals is not only the crazy volume but also that you know they are going to need that volume, and in no way will not be apple to pay. Its the perfect customer in that way. And ofcource they rip what they can
Apple not only needs volume, but it needs leading edge volume.
Is this actually a fair comparison since the compiler can have a lot of influence: http://seekingalpha.com/article/1848061-intel-vindicated-very-competitive-with-apples-a7.
Not necessarily apples to apples speed.So what? Speed is speed.
Is this actually a fair comparison since the compiler can have a lot of influence: http://seekingalpha.com/article/1848061-intel-vindicated-very-competitive-with-apples-a7.
the Intel CPU dispatcher does not only check which instruction set is supported by the CPU, it also checks the vendor ID string. If the vendor string is "GenuineIntel" then it uses the optimal code path. If the CPU is not from Intel then, in most cases, it will run the slowest possible version of the code, even if the CPU is fully compatible with a better version.