Anything between 0 and a few 10 billions, depending on how much chip area customers are willing to pay for. Did you want to ask for transistor density? That would also depend on the mix of transistor types, but is arguably somewhere between Intels 14 nm and 10 nm.
That is what I heard also - that it more than Intel 14 nm but less than Intel's up and coming 10nm.
Intel is actually very smart with there process where they allow different size chips, for example CPU and GPU can be on 10nm will IO components can be on 14nm or even 22nm. This actually keeps the cost down instead having everything on 10nm.
Intel appears to have been too ambitious with their 10nm node, where they managed to pack 101 million transistors inside a mm^2, which is why it was delayed that much. I doubt the others will manage to go beyond that at 7nm (or even match it, perhaps), not in that node's first generation anyway. Of course all that matters is that Intel ships 10nm CPUs (hopefully in Q1 2018).
TSMC and Samsung 7nm are 20-30% denser than Intel 10nm. TSMC has started production of 7nm and should enter mass production within 6 months.
However I think it is misleading just consider maximum theoretical transistor density. How you use the transistors matters too. See eg. https://www.informaticapremium.com/blog/wp-content... where AMD manages to pack transistors tighter despite using a less advanced process.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
9 Comments
Back to Article
HStewart - Wednesday, November 29, 2017 - link
Does anybody know what Transistor count will be this process and how it compares to existing processes?MrSpadge - Wednesday, November 29, 2017 - link
Anything between 0 and a few 10 billions, depending on how much chip area customers are willing to pay for. Did you want to ask for transistor density? That would also depend on the mix of transistor types, but is arguably somewhere between Intels 14 nm and 10 nm.HStewart - Thursday, November 30, 2017 - link
That is what I heard also - that it more than Intel 14 nm but less than Intel's up and coming 10nm.Intel is actually very smart with there process where they allow different size chips, for example CPU and GPU can be on 10nm will IO components can be on 14nm or even 22nm. This actually keeps the cost down instead having everything on 10nm.
Santoval - Thursday, November 30, 2017 - link
Intel appears to have been too ambitious with their 10nm node, where they managed to pack 101 million transistors inside a mm^2, which is why it was delayed that much. I doubt the others will manage to go beyond that at 7nm (or even match it, perhaps), not in that node's first generation anyway. Of course all that matters is that Intel ships 10nm CPUs (hopefully in Q1 2018).Wilco1 - Thursday, November 30, 2017 - link
TSMC and Samsung 7nm are 20-30% denser than Intel 10nm. TSMC has started production of 7nm and should enter mass production within 6 months.However I think it is misleading just consider maximum theoretical transistor density. How you use the transistors matters too. See eg. https://www.informaticapremium.com/blog/wp-content... where AMD manages to pack transistors tighter despite using a less advanced process.
Lodix - Wednesday, November 29, 2017 - link
It has around 50MTr/mm^2 similar to TSMC's 10nm and 30-40% more dense than Intel's 14nm++.HStewart - Wednesday, November 29, 2017 - link
But Intel's 10nm is suppose to be around 100nm which could explain why it been delay so much.atragorn - Wednesday, November 29, 2017 - link
Samsung yet has to detail the 10LPU,SquarePeg - Thursday, November 30, 2017 - link
So in other words, Snapdragon 845 inbound with Galaxy S9/S9+ early exclusivity.