Chinese semiconductor industry

[krautmeister]

An addendum on this: the equipment in question are fairly standard industrial products with no particularly high capital intensity and no particularly long product cycles and EDA is software and China has a ton of MechEs and EEs

EDA software isn't as easy you as you're making it out to be. There are only somewhere around 1500 EDA programmers in China and around 300 of them work for China based EDA companies with the rest working for global EDA providers. Synopsis alone has over 5000 EDA programmers worldwide. It doesn't matter how many Mechanical/Electronic engineers/programmers there are in China if the EDA niche is tiny.

Much more importantly, China needs its own EDA software ASAP because companies like Huawei/HiSilicon are banned from buying any semiconductors produced with American EDA software, of which they have a global monopoly with their triopoly of Cadence, Synopsis and Mentor Graphics, primarily because they are the largest end-to-end full design flow solutions. American politicians are now pushing for a global ban on sales of <=14nm semiconductors to China designed by American EDA software.

EDA software is further perfected by working closely with the fabs and their different kinds of equipment. Equipment from an SMEE or CETC isn't going to necessarily work that well with any of the big 3 American EDA providers when they aren't working with China and aren't even allowed to sell semiconductors to China designed with their software. Pirating copies of American EDA and continuing to use them will result in no support and if we can even speculate, somehow compromised designs that could potentially have "problems" intentionally planted into them surreptitiously by the EDA software.

Without a domestic China spec to delivery full design flow EDA solution, China is going to be trapped. Time is also of the essence. The Empyrean Technologies EDA allows China fabs to produce at 28nm for now and at 14-16nm in the future. Companies like Huawei/HiSilicon have no time to waste because their market window will close otherwise.

 

[latenlazy]

EDA software isn't as easy you as you're making it out to be. There are only somewhere around 1500 EDA programmers in China and around 300 of them work for China based EDA companies with the rest working for global EDA providers. Synopsis alone has over 5000 EDA programmers worldwide. It doesn't matter how many Mechanical/Electronic engineers/programmers there are in China if the EDA niche is tiny.

Much more importantly, China needs its own EDA software ASAP because companies like Huawei/HiSilicon are banned from buying any semiconductors produced with American EDA software, of which they have a global monopoly with their triopoly of Cadence, Synopsis and Mentor Graphics, primarily because they are the largest end-to-end full design flow solutions. American politicians are now pushing for a global ban on sales of <=14nm semiconductors to China designed by American EDA software.

EDA software is further perfected by working closely with the fabs and their different kinds of equipment. Equipment from an SMEE or CETC isn't going to necessarily work that well with any of the big 3 American EDA providers when they aren't working with China and aren't even allowed to sell semiconductors to China designed with their software. Pirating copies of American EDA and continuing to use them will result in no support and if we can even speculate, somehow compromised designs that could potentially have "problems" intentionally planted into them surreptitiously by the EDA software.

Without a domestic China spec to delivery full design flow EDA solution, China is going to be trapped. Time is also of the essence. The Empyrean Technologies EDA allows China fabs to produce at 28nm for now and at 14-16nm in the future. Companies like Huawei/HiSilicon have no time to waste because their market window will close otherwise.

Domestic EDA is constrained by adoption, not technical difficulty. It’s very hard to expand the suite of EDA features if you don’t have direct access to the instruments and processes being used by designers and fabs.

 

[krautmeister]

No, EBL, LDW and NIL are all just for mask-making and *very* small-volume IC

Touche, but if the process was more efficient and faster, it would open more possibilities. Every new technology is the precursor for new processes or at least good knowledge accumulation.

 

[WTAN]

According to Havok, 28nm immersion duv machine can do double patterning, will participate in 16nm domestic equipment trial production line, and will be used in big production line this year end.

This is excellent news. This shows that the much anticipated Indigenous 14nm/16nm Production Line may be developing much faster than expected.
I believe that SMEE was initially being very conservative in its roadmap for further improvements to the 28nm DUVL.
Afterall it was their first DUVL offering such a capability at the 28nm node.
SMEE Engineers after recent trials and evaluations now realised that their DUVL performed better than expected.

 

[Deleted member 15949]

EDA software isn't as easy you as you're making it out to be. There are only somewhere around 1500 EDA programmers in China and around 300 of them work for China based EDA companies with the rest working for global EDA providers. Synopsis alone has over 5000 EDA programmers worldwide. It doesn't matter how many Mechanical/Electronic engineers/programmers there are in China if the EDA niche is tiny.

Much more importantly, China needs its own EDA software ASAP because companies like Huawei/HiSilicon are banned from buying any semiconductors produced with American EDA software, of which they have a global monopoly with their triopoly of Cadence, Synopsis and Mentor Graphics, primarily because they are the largest end-to-end full design flow solutions. American politicians are now pushing for a global ban on sales of <=14nm semiconductors to China designed by American EDA software.

EDA software is further perfected by working closely with the fabs and their different kinds of equipment. Equipment from an SMEE or CETC isn't going to necessarily work that well with any of the big 3 American EDA providers when they aren't working with China and aren't even allowed to sell semiconductors to China designed with their software. Pirating copies of American EDA and continuing to use them will result in no support and if we can even speculate, somehow compromised designs that could potentially have "problems" intentionally planted into them surreptitiously by the EDA software.

Without a domestic China spec to delivery full design flow EDA solution, China is going to be trapped. Time is also of the essence. The Empyrean Technologies EDA allows China fabs to produce at 28nm for now and at 14-16nm in the future. Companies like Huawei/HiSilicon have no time to waste because their market window will close otherwise.

Your reading too much into it. EDA (or any sort of modelling software) is knowledge of computer algebra/symbolic computing and the underlying circuit models. There's are procedural gains to be made with optimizing along side a litho vendor but again, then that just becomes a SWE problem.

 

[Deleted member 15949]

Touche, but if the process was more efficient and faster, it would open more possibilities. Every new technology is the precursor for new processes or at least good knowledge accumulation.

EBL, LDW and NIL have been with the world for decades. They are not "new technology" and their capability for VLSI has been attempted for decades during the painful ArF immersion multipatterning years

 

[krautmeister]

Domestic EDA is constrained by adoption, not technical difficulty. It’s very hard to expand the suite of EDA features if you don’t have direct access to the instruments and processes being used by designers and fabs.

Concerning adoption, China really has no choice but to adopt China based EDA if they are working with any process <=28nm. This is especially true for Huawei/HiSilicon. If they don't and they go the pirated software route for those <=28nm designs, it's at their peril.

The crazy thing is, literally almost every China university is still using American EDA in their programs, generally either Synopsys or Cadence. These American EDA providers give the universities free licenses that get the students hooked on their softwares which they continue to demand once they graduate and work in the field. It's an obvious problem that needs to change ASAP.

 

[ansy1968]

This is excellent news. This shows that the much anticipated Indigenous 14nm/16nm Production Line may be developing much faster than expected.
I believe that SMEE was initially being very conservative in its roadmap for further improvements to the 28nm DUVL.
Afterall it was their first DUVL offering such a capability at the 28nm node.
SMEE Engineers after recent trials and evaluations now realised that their DUVL performed better than expected.

@WTAN Sir, again trying to temper my expectation, a big IF, SMIC with SMEE collaboration is able to trial produced 14nm and 16nm line within the year as @foofy posted then an indigenous 7nm line can be ready by next year? As you stated before if a 14nm line is successfully conquered going to 7nm is much quicker since its a matter of adjusting and tinkering since both processes used the same equipment?

 

[krautmeister]

EBL, LDW and NIL have been with the world for decades. They are not "new technology" and their capability for VLSI has been attempted for decades during the painful ArF immersion multipatterning years

Nobody said these were new technologies. There were recent breakthroughs in these fields by China that were mentioned.

 

[Deleted member 15949]

The crazy thing is, literally almost every China university is still using American EDA in their programs, generally either Synopsys or Cadence. These American EDA providers give the universities free licenses that get the students hooked on their softwares which they continue to demand once they graduate and work in the field. It's an obvious problem that needs to change ASAP.

No, it isn't. The UI for all EDA programs is going to be the same. You know one, you know all, type thing.