Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
They mean more then u saying they are meaningless considering it's being stated by them.. the companies making them...
If the company stated a real-world gain without citing the testbed it would just be marketing puff.

Gains realized by a process change on one design, i.e. Cortex-A53, won't be the same on another design.

You didn't cite the platforms used by the companies reporting results, but that doesn't mean they didn't.
 
How is their comparison between their 20nm SoCs and 16nm Finfets applicable to 14nm Samsung Finfets?

Asking the wrong person my man. I just stated an article that was written. Which stated what these companies stated..

Samsung and TSMC are both using Finfet designs, That TSMC called 16nm and Samsung called 14nm. And.. Wait now I'm repeating myself...

Are you confused? Or am I? Because based off what the chip makers are stating.. The article holds true.
 
Asking the wrong person my man. I just stated an article that was written. Which stated what these companies stated..

Samsung and TSMC are both using Finfet designs, That TSMC called 16nm and Samsung called 14nm. And.. Wait now I'm repeating myself...
That much is obvious. You didn't cite anything applicable to the discussion at hand. Saying a 16nm TSMC Finfet is 40% faster - by whatever measure TSMC marketing picked - than a 20nm TSMC SoC is wholly irrelevant when comparing 16nm TSMC Finfets to 14nm Samsung Finfets.

Are you confused? Or am I? Because based off what the chip makers are stating.. The article holds true.
You can cite as many unrelated statistics from their webpages as you want, but that doesn't make them relevant. Can't wait for the real world benchmarks!

"TSMC chose 16nm, which uses smaller transistors but has the same back-end metal layers, as a sort of stepping stone to ease the transition to FinFETs. It won’t provide the area scaling benefits of a typical shrink, but it does boost performance and power."
http://www.zdnet.com/article/the-race-to-the-finfets/
 
Last edited:
That much is obvious. You didn't cite anything applicable to the discussion at hand. Saying a 16nm TSMC Finfet is 40% faster - by whatever measure TSMC marketing picked - than a 20nm TSMC SoC is wholly irrelevant when comparing 16nm TSMC Finfets to 14nm Samsung Finfets.


You can cite as many unrelated statistics from their webpages as you want, but that doesn't make them relevant. Can't wait for the real world benchmarks!

"TSMC chose 16nm, which uses smaller transistors but has the same back-end metal layers, as a sort of stepping stone to ease the transition to FinFETs. It won’t provide the area scaling benefits of a typical shrink, but it does boost performance and power."
http://www.zdnet.com/article/the-race-to-the-finfets/

Ok. I think ur confused.

I have no idea what ur question, or is it a statement? Is....

Basically what the article stated was all major companies still use 20nm planar transistors. Basically combining a 16nm/14nm Finfet transistor with a 20nm planar backend to cut costs.

It gave the stats both companies touted off. TSMC also tooted off their Finfet+ design. The end.

If you don't want to trust there own stats don't. Has nothing to do with me.
 
And I'm sure majority of people, including myself, would like to know they received the version with the Samsung chip...

Any way to check what version you have without doing a teardown?

Don't you want to rip your phone apart just to find out whether the chip in your phone is Xmm smaller? I mean, it has such a great effect on people's lives that people need to trash perfectly good devices to find out. It's *that* importantant!
 
  • Like
Reactions: Marsontherocks
I laugh at anyone who trusts chinese (lol taiwanese) before samsung.
They are constantly lying and crying about how samsung is worse and steals from them yet they are unable to compete.

Idiots.
 
The Samsung die is 10% smaller with a 14nm manufacturing process, which would inherently have better power efficiency than a 16nm chip using the same design.
Not necessarily. All else equal, yes. But there is likely a reason TSMC went with a 16nm process vs a 14nm process. Perhaps they were getting better yields and producing more chips within spec than with a 14nm process.
 
If it really is the same process with different "branding," why is the Samsung chip 10% smaller as expected?
Actually, we'd expect more than a 10% difference between a true 16nm design and a true 14nm design. Both are hybrids, as Intel likes to point out. 14x14 is 23% smaller than 16x16. The fact that the area difference between TSMC and Samsung is only 10% suggests the two processes are much closer.
 
I love the split on this thread: those that are outraged that others might think it's an issue; those that think it is an issue and those who are concerned it might be an issue and are reassuring themselves by posting positive comments.

What do we know? The chips use different die sizes.

What we don't know: whether it makes any difference to the user experience.

If we find that it doesn't, then there is no need to have an issue, if we find that there is a negligible difference, then some will feel justified in complaining, but most of us will get on with our lives. If there is a tangible difference, then we will all be peeved, but most of us won't do anything about it, but the media will make a thing of it.

But, at present, 11 pages of comments in response to a fact that raises questions for which we don't yet have answers is incredible, especially given the comments are of a type that one would expect to the answers that we don't have.
This.
But people are people, and given the choice they will go for what they believe is the best choice for X money.
You could buy two almost identical devices and even if the difference is akin to a placebo but objectively makes no difference, the one that makes YOU feel better is money well/better spent?
 
  • Like
Reactions: jacobj
Not necessarily. All else equal, yes. But there is likely a reason TSMC went with a 16nm process vs a 14nm process. Perhaps they were getting better yields and producing more chips within spec than with a 14nm process.
TSMC's reasoning for using inferior technology doesn't change the fact that their 16nm process is less efficient.

16nm was cheaper and easier for TSMC migrate to, with yields nearing their 20SoC process this year. Samsung and Intel went with 14nm this generation instead and will reap the benefits.
 
I laugh at anyone who trusts chinese (lol taiwanese) before samsung.
They are constantly lying and crying about how samsung is worse and steals from them yet they are unable to compete.

Idiots.

In the end Apple doesn't trust their suppliers either, this is why they all products are heavily tested at all levels of production. If there a any differences between units they are within the tolerances that Apple has set for them and will repair them if they fall out of those tolerances. Apple uses multiple suppliers not only to bring up their production quantity but to avoid being backed into a corner by any one company and to be in a better bargaining position, especially when one is a direct competitor.
 
I have the TSMC chip in my 6S Plus. Happy with it as they really know their stuff. If you all want some benchmarks, let me know.
 
Actually, we'd expect more than a 10% difference between a true 16nm design and a true 14nm design. Both are hybrids, as Intel likes to point out. 14x14 is 23% smaller than 16x16. The fact that the area difference between TSMC and Samsung is only 10% suggests the two processes are much closer.
Splitting hairs: assuming 100% averagegains (not realistic anyway) from 104.5mm2 at 16nm, the 14nm shrink would be 91.5mm2 - not exactly 10% but within expectations.

TSMC's 16nm features are additionally known to be far larger than the 14nm competition:
 
If there a any differences between units they are within the tolerances that Apple has set for them and will repair them if they fall out of those tolerances.
Only if the consumer notices and complains! They're happy to ship inferior components to meet their quarterly goals for Wall St.
 
It was rumored yesterday that AMD will be using TSMCs 16nm fabs instead of Global Foundry's 14nm fabs for their next gen CPUs next year. GFs 14nm should have the same specs as Samsungs 14nm*, so that would suggest that TSMCs 16nm process is quite mature and a suitable alternative to 14nm.

*(Samsung and Global Foundries developed their 14nm technology together, but samsung got their fabs operational much more quickly than GF has been able to).

Absolutely false Samsung licensed there tech to global.
 
image.png


That TSMC chip son! :)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.