So apparently Global Foundries has a 7nm process already working, however, Intel is planning to deploy their 10nm process later this year while GF isn't expected to deliver 7nm until mid 2018. In addition, Intel does not have to worry about other companies siphoning their manufacturing yield like GF does so I can imagine it probably wont be until 2019 when we see the first 7nm Ryzens roll off the line while Intel can easily roll out 10nm chips to consumers right away. This still means that Intel will actually be behind in the market come 2019 to 2020 but I can easily see them adopting GFs process. In addition, Intel is pouring R&D into non-silicon based chips in order to get yields as low as 5nm, GF is not, so I can also easily see Intel undercutting GF by skipping straight from 10nm to 5nm come around late 2020
CPU processes
Well, the "Law" is dying anyway.
There's a huge difference between a 48nm and a 24nm chip, but it's starting to become meaningless from 7 to 5, both in percentage and absolute size.
So, don't get too carried away with the downsizing, other areas will get the spotlight in the future, like the very power consumption.
That's not how yield works
Well, in general process shrinks are getting harder for everyone. We saw how many times Intel delayed their 14nm process. We are getting to the very limits of silicon as a viable semiconductor. Even with moving to another material we're reaching the very limits of physics.
Going from 10nm to 7nm may not sound like much but proportionally it's in line with previous process shrinks.
You dont understand how fractions work do you?
dude will we ever have negative nm processes?? like, with absolute zero and shit you know? you know they went below absolute zero? we can do this with fab processes too go -5nm -20nm etc. you know man? does moore's law even apply then?
what would that even be? It sounds like it would be some kind of static electric field suspended in place by supermagnets with multiple planes that are jumped to vie quantum tunneling
semiconductor nodes are the new gigahertz, its a number companies can use to make their chips stand out now.
idk i try opening my 3rd eye to see it but it's hard to wrap the mind around man. what is negative absolute zero? maybe it would be like string dimensions or mirror mass or quantum magic that makes it have negative dimensions. or dude there could be a singularity so once we hit negative processes we start speeding up again on moore's law.
This shits blowing my mind my dudes
Antimatter. If you thought Intel would never make a hotter-running chip than the P4 EE, think again.
That's only in one direction, going from 10nm to 7nm is actually an effective decrease in transistor area (and in turn power consumption) of 51% (0.7^2).
DUDE
What are you talking about, Intel gave up on 10nm and is now trying to start on 7nm because their yields were so terrible. Cannon lake is going to be another 14nm process.
For those more knowledgeable than me, what is the limits of transistors? When are we going to reach a point where we just can't cram any more of them onto a board? After that what then? Will computers just reach a power-manufacturing equilibrium and everyone will have exactly the same computer?
There are two limits you need to take into account, the first is by physical transistor dimensions, and the other is by the manufacturing process, photolithography.
In terms of transistor dimensions, my guess is that once you reach ~100s of atoms scale it will be completely dominated by quantum mechanical effects and stop behaving like a semi-conductor device. As to the question of how it would behave, I have no idea.
Photolithography is limited by wavelength. I recommend you read pic related and if you want to know more, read about diffraction.
Well, the limit for practical silicon transistor nodes is believed to be 5nm. Using non-silicon semiconductors such as gallium-arsenide it might be possible to reach as small as 1nm. I honestly doubt we'd even need nodes smaller than that. After that we outta focus on a new computer architecture that is able to take maximum advantage of a given target die size now that we have transistor node sizes down.
That's right goy wait for Intel don't buy ryzen in one year Intel will have 10nm and they will be the kings again don't worry they will go back to providing superior value like they have the last 5 years!
Intel is making a factory that produces 7nm chips though...
That would take at least a few hundred years; but, who knows, it could be sooner, last year we discovered "Time Crystals" which is a four-dimensional material.
I hear that time cristals are what are needed to make the EM-drive work.
I can't tell if that's a genius idea or the result of the desperate ravings of a lunatic turned into a painting