Originally Posted by DarkIdeals
If Vega does use TSMC then that means they are gonna be on 16nm FinFET most likely as 14nm FinFET isn't really ready yet. So that would actually be a very smart move for AMD to make personally. You don't seem to think 14nm will make a difference compared to 16nm but i guarantee you it DOES make a difference. When you get this small, EVERY single nanometer makes a much larger difference than it did in the past.
This is why you are seeing incremental increases in each node jump. For example, in the past we had 65nm, then went to 45nm, then 28nm etc.. which were large jumps. But now we have 28nm that moves only to 20nm (only 8nm jump) and then from 20nm we only go to 16nm (4nm jump) and 14nm (6nm jump). Next is going to be 10nm which is only a 4 or 6nm increase from 16/14 as well
. The reason for this is that we are literally hitting the long predicted point where silicon just can't get places any closer to each other without significant issues. Each new node size is taking EXPONENTIALLY longer to mature and master than it has in the past.
I mean think about it. Think about how INSANELY hard it is (even for a machine) to place 12 BILLION transistors with each transistor only 10 or 14 or 16nm apart from each other? Do you realize how small a nanometer is? A human hair is 100,000 NANOMETERS THICK!!! A SINGLE STRAND OF DNA IS 2.5 NANOMETER IN SIZE! The current 14/16nm chips are literally requiring Nvidia all companies to have electron microscopes SO powerful that they can see INDIVIDUAL ATOMS!! ( https://www.youtube.com/watch?v=pRz_CG3DZb4 )
So yes, even 2nm difference between 16nm and 14nm does make a decent amount of difference in yields.
Not a WHOLE lot, but it's still somewhat significant.
I guess you didn't get the memo. Nanometers are a constant, process nodes aren't!
I has this conversation before when AMD/Nvidia rumours regarding 14/16nm started popping up and people tried to calculate performance JUST using 14 and 16nm numbers. (Seriously, this is an insult to lithography)
Source #1 indicates that TSMC 16nm has better power characteristics (and thus performance) (However 14nm process has changed a bit)
Source #2 are the different nodes used by companies.
Yes! I used my eyes to look up this information! Did you know that the cones in my eyes can detect a single photon! Yes, its my brain thats the limiting factor here!
Companies give out this kind of information in the hope you will be amazed and hyped up, just like the whole nanosecond gddr5 slideshow that was released some time ago.Edited by Yttrium - 7/28/16 at 11:31am