Chinese semiconductor thread II

olalavn

Senior Member
Registered Member
good news...

with a total area of 224 acres, a construction area of 250,000 square meters, and a total investment of 5.8 billion yuan. The pile foundation construction began in June last year, and the capping ceremony was held on December 31 of the same year. It took only six and a half months to complete the capping of the 250,000 square meter building in October this year. The main processing equipment of the project has begun to be delivered and debugged. Now the first bar of the monocrystal furnace has been successfully produced and successfully logged out. After the project is completed and reaches production capacity, it will have an annual production capacity of 3.6 million 12-inch polished wafers and is expected to reach a monthly production capacity of 150,000 wafers by the end of next year.

Please, Log in or Register to view URLs content!
 

JPaladin32

New Member
Registered Member
There are some Chinese tech guys who do chip teardowns also. Actually they do a better job than Techinsights in many ways. Some of the analysis should be coming out soon and we will have more evidence on the process.

To be honest I was wishing for miracles but in the end reality is reality. Reality is that mass production of a new node is about 1.5 years behind risk production and the start of design of a new SoC is usually 1.5 to 2 years ahead of release also. There was news about successful prototyping of 5nm earlier this year, then it's likely we won't see it in mass production until late 2025. With this estimate, it's actually perfectly normal that even Mate 80 won't be able to make it to 5nm. If we see 5nm on Mate 80, then it is already ahead of schedule by industry standards.
 

iewgnem

Junior Member
Registered Member
Then it means the design optimizations they made are so much more impressive, to use a 7nm process and have similar or better performance than SD8+ gen1, which is fabbed on TSMC 4nm.
Honestly this has much bigger implications, if they achieved TSMC 4nm performance on 5nm, it's okay but still behind.
But if they achieved 4nm performance on 7nm, that means they're ahead in design, and there's a good chance they can achieve TSMC 3nm performance on 5nm, and so on.
It also means once Chinese EUV comes online it's game over for everyone else.
 

Hyper

Junior Member
Registered Member
There are some Chinese tech guys who do chip teardowns also. Actually they do a better job than Techinsights in many ways. Some of the analysis should be coming out soon and we will have more evidence on the process.

To be honest I was wishing for miracles but in the end reality is reality. Reality is that mass production of a new node is about 1.5 years behind risk production and the start of design of a new SoC is usually 1.5 to 2 years ahead of release also. There was news about successful prototyping of 5nm earlier this year, then it's likely we won't see it in mass production until late 2025. With this estimate, it's actually perfectly normal that even Mate 80 won't be able to make it to 5nm. If we see 5nm on Mate 80, then it is already ahead of schedule by industry standards.
If you are talking about Geekerwan, I think they only do laptop and desktop cou teardown.
 

interestedseal

Junior Member
Registered Member
In the absence of much visible US progress, the market for gallium nitride-based RF power amplifiers used in 5G networks is now a three-player race between NXP, Sumitomo and "whoever Huawei is using over in China through a foundry," according to Lum.
Please, Log in or Register to view URLs content!

Apparently GaN power amplifiers in 5G base stations are really hard to make. HW probably sourced from Sanan or fabbed it in-house.
 

huemens

Junior Member
Registered Member
There are some Chinese tech guys who do chip teardowns also. Actually they do a better job than Techinsights in many ways. Some of the analysis should be coming out soon and we will have more evidence on the process.

To be honest I was wishing for miracles but in the end reality is reality. Reality is that mass production of a new node is about 1.5 years behind risk production and the start of design of a new SoC is usually 1.5 to 2 years ahead of release also. There was news about successful prototyping of 5nm earlier this year, then it's likely we won't see it in mass production until late 2025. With this estimate, it's actually perfectly normal that even Mate 80 won't be able to make it to 5nm. If we see 5nm on Mate 80, then it is already ahead of schedule by industry standards.

Even when they go for 5nm mass-production, it is more likely they would prioritize AI GPUs over smartphone SoCs, because that gets the most bang for the buck. For example Nvidia get nearly 1000% profit from an AI GPU but no one makes that kind of markup from a smartphone SoC. For TSMC customers high-end SoCs get the latest node and Nvidia GPUs are usually a node behind, but they still get away with it due to Nvidia's monopolistic position in the AI GPU market.
However, it's very likely that it would be the other way around for domestic fabs. Given that any potential domestic 5nm would be in high demand for AI GPUs, and how much performance gains they are able to make by optimizing the SoC design it's very likely that AI GPUs would always get the best node and SoCs stay a node behind.
 
Top