https://www.techpowerup.com/225095/globalfoundries-to-skip-10-nm-and-jump-straight-to-7-nm
Will this benefit GPUs, I have no idea.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
GlobalFoundries to Skip 10 nm and Jump Straight to 7 nm
- Thread starter Kaapstad
- Start date
https://www.techpowerup.com/225095/globalfoundries-to-skip-10-nm-and-jump-straight-to-7-nm
Will this benefit GPUs, I have no idea.
*Theoretically* it should benefit GPU's since you'd be fitting 50% more transistors in a die of equal size going from 14 - 7nm. Course there's variables that can affect things. If GloFO can get this up and running in the next few years with good yields this could be quite the coup for them.
Soldato
- Joined
- 25 Nov 2011
- Posts
- 20,680
- Location
- The KOP
What happens next? is it possible we could see 1nm?? what happens after that?
Honestly, i've no idea. Smaller transistors should use lesser voltage, whether the voltage reduction will mitigate the transistor density... Who knows. Could be a while before we find out too.
*I believe* 5nm is kind of a hard limit. Below that (or somewhat below) certain principles in quantum physics start to kick in which ruin things. But i'm no expert.
I have said this before, 10nm is a bit of a dud in regards to graphics cards and will be skipped liked 20nm was. This is why Nviida are going to a plan B with Volta and releasing a 2nd generation product on 16/14nm in H1 2017 with a cut-down Volta design. With full Volta or something beyond on 7nm in H2 2018.
Only thing I've heard about 10nm being a "dud" for GPUs is that the setup/design cost for a GPU is like 150 million USD compared to like upto 80 million for stuff based on 20nm (including 16nm FF). TSMC is undergoing massive retooling for 10nm so I'll be surprised if they skip it (though I guess it is easier to transition from there to 7nm than from 28nm to 16nm FF).
Last edited:
It's definitely not anything to do with quantum physics
5nm is still 3.125e+26 times the scale at which quantum physics truly takes over - which is kind of scary trying to comprehend that.
EDIT: You still have to deal with the whole quantum tunnelling malarkey at 5nm though.
Last edited:
5nm/7nm/14nm etc doesn't really mean a whole lot nowadays. IIRC it more refers to the accuracy of the processing node. Many individual parts can be smaller and others much bigger.
In the past it used to refer to the length of the (gate?) but not so much it seems now.
Yeah 10nm is getting slapped on just about anything sub 19nm planar and 7nm is probably more accurately what 10nm would be.
I'm a bit sad you didn't quote the original article instead, as here's something very important to highlight:
As for people talking about expecting it in 2018, that won't happen. It will be many years still until that jump is made. But that's all good news for people on a budget imo. Cheap & powerful cards are here to stay and their longevity is well assured, what's happening is quite a bit like what's going on in the CPU world.
As for people talking about expecting it in 2018, that won't happen. It will be many years still until that jump is made. But that's all good news for people on a budget imo. Cheap & powerful cards are here to stay and their longevity is well assured, what's happening is quite a bit like what's going on in the CPU world.
Last edited:
Yup cannot see this happening anytime soon.
I mean it was back in Dec 2012 that Samsung announced their first 14nm test chips. so it has taken three and a half years before we got the first graphics cards on that node and there has been no test chips from Samsung's 7nm process that have been announced as of yet. (couldn't find any mention of them on google)
I mean it was back in Dec 2012 that Samsung announced their first 14nm test chips. so it has taken three and a half years before we got the first graphics cards on that node and there has been no test chips from Samsung's 7nm process that have been announced as of yet. (couldn't find any mention of them on google)
Interesting isnt it, moving to lower power requirements which is sort of a good thing but I've got to assume it starts to heavily restrict overclocking performance/voltage increases as it shrinks. Though realistically i'm making a bit of a guess there!
One of the biggest delays with going below 28nm was in relation to the lithography which won't be the case going from 20nm downwards until possibly 5nm which would likely require a whole new approach again so I don't see it being as delayed as the move to 14/16nm FF was though you never know what other problems might crop up.
An atom is like 0.1nm long that's just 1/50th the size of a 5nm transistor. That's got me thinking maybe I'm wrong. I always just assumed Quantum is something on another level in terms of size but there does not seem to be defined size of quantum behaviour. In other words at what scale does theory of relativity cross over to quantum physics.
Schroedinger's Cat. Things like photosynthesis happen through a quantum process so the fact that we have macro-sized plants is purely due to quantum phenomena.
Silicon processors already suffer from quantum issues suchas electron tunneling which is why some years back they moved to soecial doped low-k silicon