Taiwan Semiconductor Manufacturing Company (TSMC) has obtained a bunch of new orders, mainly advanced AI solutions, however requiring 7nm and 5nm process capacities in 2019, according to industry sour... TSMC going strong on 7nm and 5nm
It's refreshing to see someone is actually making an effort to push technology forward. Not like a certain company stuck at 14nm since 2014.
Global Foundries and Samsung are also making great strides. It seems GloFo seems they may be skipping 5nm. So it's great to see so much competition.
I just want to point out that Intel is shipping processors on the 10nm process in low volume and their 10nm process is slightly better overall compared to TSMC's 7nm process. Like yeah, Intel is losing its lead but it's still ahead and they clearly spent a metric ton of money in the last 3 years trying to make 10nm work. So I wouldn't say Intel isn't pushing forward, they just ran into some unknown roadblocks that these other companies are avoiding - probably because they all have knowledge of where Intel got tripped up.
They cooperate to some degree because that's way forward. There's enough orders for everyone if technology does work. Or there are orders for nobody if they try to gain advantage over everyone and fail.
Its all ASML's fault! https://www.extremetech.com/computing/265064-euv-integration-5nm-still-risky-major-problems-solve
there are already 5 o.e.m.'s that have windows 10 notebooks and tablets on the new ARM chips. all day battery life on a windows 10 machine is a slam dunk for anyone who travels. i can leave the gaming laptop in the luggage and carry around an iPad with balls.
I thought that 12nm was around the corner. September 20th, wasn't it? WOW 7nm/5nm around the corner... are we talking in geological terms?
Not that I understand the logic behing the chip making but what happens when we get too 1nm? Were do we go from there? Intel's doing just fine I think Amd sorta pocked em in the ass a little bit to wake up.
We’re getting very close to the limit of how small we can make a transistor. 1nm chips may not even be possible.
5nm was said to be near impossible but we'll see. Those poor electrons need some room to move in 5nm will have some leakage i bet. When i say around the corner i mean a year or two. Time travels fast where tech is involved. ^^
1nm is quite unrealistic with current selection of materials. 3nm is already crazy. I did calculation of atomic matrix which make transistor and IIRC at 1nm shortest material used there has 13 atoms. Atoms are big, Man. They'll have to go back to drawing board and find something that behaves in similar way, but can tick much faster without melting and it can be on bigger node too...
https://semiengineering.com/quantum-effects-at-7-5nm/ Here is a great article about this kind of stuff.
In theory you can have a transistor just from the 2 atoms (aobut 0,5 nm), but the problem we will face very soon are the quantum effects. So I don't think we will reach that limit at all. It's as you've said, we need to use different material that can clock higher, like GaAs. And with the physical limit of silica aproaching, we might finaly move there soon.
I read somewhere a while ago that 3 nm was probably rock bottom. I have no clue if this is still the case.
You have leakage on all nms 5nm is about the limit of silica where you start losing so much it effects wattage and energy needs. 3nm might be possible but is it worth it because you start flogging a dead horse at this point. When an electron is sent through the system is has a small % chance not to reach it's destination and this is when error correction steps in and fire another one. All these electron paths degrade over time. With a smaller surface you'll get leakage faster too. Which is basically what that study is saying. Electrons can sometimes tunnel if they want and fly out of the traces, that could be motherboard cpu memory or any of the components in a pc. We have minimal space left over at 5nm so electrons see a weak point and fly off or they start ignoring the gates in the cpu and either just go under or over the gate = lots of problems to solve.
electron migration will make anything beyond 5nm unsustainable, & i do mean completely unstable. theres literally no way to control it barring a physics-based miracle in silicon. thats why alternate materials like graphene, et al, have been researched so heavily in the past decade, & lack of progress in that area has focused more practical approaches ie heavy parallelization which eventually brought us threadripper luckily for us! as others here have already touched on, wat everyone calls 12nm, 10nm, 7nm...isnt. its marketing larger nodes marginal restructuring & optimizations. theres a respectable amount of leg room & design advancements to be made before we really well & truly touch upon a legitimate 5nm semiconductor layout. also lucky for us TSMC seems to be barreling forward on their optimizations & glofo/samsung are just as motivated to keep pace. i find the recent developments fascinating from an engineering standpoint & exciting from a gamers perspective as well
