Yeah, in a smartphone the SIZE of die could make difference, allowing thinner phone . But no other real benefit there - only fashion.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
antiterror13
Brigadier
Very good point, very interesting if Ryzen 9 5900x (I know it is not possible) were made using 45nm node, how less performance ... my guess is perhaps ~50% less powerful and need 2x of power
antiterror13
Brigadier
Well, neither can the Americans.
That is why this whole fight that the Americans started with China over chips, is a little odd.
well, basically no country could do it ...
Intel uses a lot of foreign products for chips manufacturing, i.e ASML
I have 7nm Ryzen 3600 (AMD w/ TSMC) in my gaming computer and 8nm Snapdragon 735 (Qually w/ Samsung) in my Xiaomi Mi 9T phone.
They are already very fast for my purposes.
China is going be fine "stuck" at 7nm. Unless you are a crazy gamer, or professional video editor, it doesn't matter.
I'd vote for full self-sufficiencty at lesser nodes like 28nm if it satisfied Beidou and 5G and enterprise servers. Of course, SMEE needs to be nutured, and advanced nodes are nice, but not essential.
They are already very fast for my purposes.
China is going be fine "stuck" at 7nm. Unless you are a crazy gamer, or professional video editor, it doesn't matter.
I'd vote for full self-sufficiencty at lesser nodes like 28nm if it satisfied Beidou and 5G and enterprise servers. Of course, SMEE needs to be nutured, and advanced nodes are nice, but not essential.
antiterror13
Brigadier
yes, architecture and the codes are very important too, perhaps even more important than the nodes
In huaweis case I would say they can probably get along with 14nm for a year maybe till the p60. As long as they can keep their computational photography progress evolving. The p and mate series are pretty much photocameras that can call and use apps on.
I wasn't writing a long essay. I was assuming some basic understanding of the context and industry. Chinese semiconductor industry is a very large although fragmented one. I obviously couldn't possibly suggest "Chinese tech company had to sit and wait for two years twiddling its thumbs." This is very silly argument.
The majority of Chinese semiconductor industry is below 28nm, so are the demands. So clearly the construction of fabs of mature nodes have relatively lower risk and much lower cost. They should obviously continue to build, and utilize domestic equipment and materials as much as possible. It's the construction of higher end nodes that are risky and costly.
Hyper-focused narrow goal? What are you even talking about? It's a straw-man's argument.
First of all, having SMEE's 28nm lithography machine ready by 2021/2022 (the often quoted goal by SMEE) doesn't mean a fab will be ready to produce 28nm chips immediately. It will take time to tune and improve yields. Secondly, lithography machine is just one piece of the puzzle. A fab would need thousands of pieces of equipment. Majority of them, if not all of them need to be domestic. That would be a daunting task and it'll take time. 2-3 years to build a 28nm fab with indigenous equipment and materials? I'm being optimistic. It would be a great achievement.
A EUV solution emerging in 2-3 years? I don't understand what's your definition of "emerging." You could argue an EUV solution is emerging today, given the news/rumors about the progress of some of the key subsystems. But if the criteria are mature and stable production of 7nm and below chips (or even trial production), they're at least 5-10 years away.
The demands for 7nm nodes and above are very small part of the market. The customers are also mostly American ones. These customers are unlikely to outsource their chip production to SMIC or other Chinese fabs, given the relationship between the two countries. QUALCOMM had given SMIC some orders at 28nm nodes a few years ago. That was more of a political gesture, coming after Chinese government fined QUALCOMM over a billion USD for monopolistic behaviors.
Among the Chinese companies, Huawei is probably the only one with urgent demand for 7/5/3nm technologies. Most people were thinking of Huawei when they learned about China's semiconductor industry is being sanctioned by the US. It's a shame, but Huawei is just way ahead of everybody else in China. They have to step back and develop together with the rest of the supply chain. At this point, only TSMC, Samsung and Intel are still developing 7nm and below, all other semiconductor companies have stopped pursuing 7nm and below.
If China becomes self-sufficient at 28nm in 2-3 years, and then reach 7nm in 5 years using SMEE's DUV lithographic machine and other domestic equipment and materials, China should be fine.
I mean the narrow goal of totally indigenizing the Chinese semiconductor supply chain. That's a long term goal with many intermediate stages; I think the discussion here often focuses exclusively on the narrow (although grand) goal, without sufficient discussion of intermediate options.
Maybe. The 28nm lithograph is the keystone to this whole enterprise - for the other components China already has world-class producers (AMEC's etchers), has made recent breakthroughs (the ArF photoresists), or has options outside the US. The lithograph is already in SMIC's hands and the verification work (tuning and yield improvement, as you put it) is being done now. 2021/2022 is when mass production and fab installation is expected to roll out.
Besides, I think it's <1-2 years, you think its 2-3 years - not a huge difference and certainly not worth getting into it over.
Sorry, I should have been clearer. I mean the first complete prototype machine capable of producing at the sub-7nm node. About the state of the SMEE DUV when it was first delivered to SMIC.
That's the state of the market today. It will likely look different in 2025 when <=7nm matures.
- Status