ISMC Announces Setting Up Of India's First Semiconductor Fab In Karnataka At An Investment Of $3 Billion

[HydraChess]

In context of my project I don't know what is a validation tool but from what I know the Xilinx software called Vivado used to program Xilinx FPGAs costs around five lakhs and the FPGA board itself ranges from more than a lakh to 10 lakhs depending on the sophistication and facilities on the board.

Vivado is Xilinx synthesis, place and route and target bit stream generator combined with a HDL editor. In the past it was called Xilinx ISE. This is STILL a FPGA and CPLD specific tool. Still a PLD / PLA tool. Actual silicon targetting tools and standard cell based ASIC tools are much more expensive and much highly protected.

startup company that designs a processor will not need more than one installation of that software and not more than one board. I am sure there are enough installations of Xilinx FPGAs locked into enough Xilinx Vivado softwares in China and this would be for prototyping and I think these installations can be disconnected from the internet to prevent Xilinx from disabling the software.

As I said, these are only PLA/PLD tools. They are useless without your prototyping board. So they are not locked down that hard.

The actual chip could be, as I said, produced in non-silicon forms or non-fully-silicon forms like graphene and photonics which China is researching on and for these the software tools for production needs will of course have to be adopted from open source means or developed anew.

Well, here is the basic problem.

<High Level Description> -> Gate Level Description -> Transistor Net-List -> Geometry on Chip.

This transformation will still apply. You can not prevent it. All of this requires tools. "High level" can be HDLs or more common these days, behavioral synthesis technologies. All of these are dominated by American companies. If they ban you, you are dead in water.

But what are the IP libraries you speak of and will they be applicable in a production environment where the materials and paradigms are different ?

Yes, they will be. What will change will be their geometries on new material and inter-connect. But there are only these may ways in which you can realize a PCI bus or AMBA bus or a SRAM block. Those are fixed. You will need those in your foundry's library of basic blocks.

Please read this article about a NASA research as of 2014 about bringing back of the vacuum tube transistor concept to nano scale miniaturization which the research found the transistor resistant to radiation and to work ( switch ) at 460 GHz ! Seems this transistor has to be fabricated using MEMS fabrication technique so it is possible that the software toolchain will need to be customized to this in addition to using open source software.

Does not matter how you make a transistor, you will still have to create a bus or a SRAM block out of it. A foundry like TSMC has a library of thousands of such blocks so user can just instantiate it in their design and they do not have to place and route every single component again and again for each customer.

 

[jamahir]

Does not matter how you make a transistor, you will still have to create a bus or a SRAM block out of it. A foundry like TSMC has a library of thousands of such blocks so user can just instantiate it in their design and they do not have to place and route every single component again and again for each customer.

Can you please elaborate with simple examples ?

Well, here is the basic problem.

<High Level Description> -> Gate Level Description -> Transistor Net-List -> Geometry on Chip.

This transformation will still apply. You can not prevent it. All of this requires tools. "High level" can be HDLs or more common these days, behavioral synthesis technologies. All of these are dominated by American companies. If they ban you, you are dead in water.

What is your opinion on an open source EDA called eSim developed by IIT Bombay ?

 

[Skull and Bones]

What do Tata Elxsi and other fabless companies in India design ?

I think they are into FPGA, but not in the scale of Xilinx.

 

[jamahir]

I think they are into FPGA, but not in the scale of Xilinx.

You mean they manufacture their own FPGA chips and boards at small scale ?

 

[Skull and Bones]

You mean they manufacture their own FPGA chips and boards at small scale ?

Not sure. I think mostly design.

 

[jamahir]

Not sure. I think mostly design.

I see. I will see what I can find.

 

[HydraChess]

What is your opinion on an open source EDA called eSim developed by IIT Bombay ?

Long way home. Not enough proven silicon. Debugging EDA is HARD. Validation (not testing) is harder. It may be great for synthesizing designs for FPGA or CPLD but I will wait till I see more and more designs being pushed out of clean-rooms if not fabs.

What do Tata Elxsi and other fabless companies in India design ?

Tata Elxsi (though I never worked there) is into outsourcing of product development. You want a MP3 player developed? They are your people. They do creative part, engineering part. You then manufacture your product in China or somewhere. They are not fundamentally an IP/design company. They are a product development company. They typically do integration of IPs sourced from a lot of places into SoCs. They (AFAIK) do not sell IPs, they sell integration, product development services.

You mean they manufacture their own FPGA chips and boards at small scale ?

No. They use FPGAs to do pre-silicon validation of SoCs. AFAIK. Think of it as TCS of hardware and product development.

Can you please elaborate with simple examples ?

Okay, say you are making a Processor. You need to put a register file in it. In FPGA you may choose to use BRAM (Block RAM). Thats a "Hard IP", right? Its build into FPGA itself, unlike soft IP (VHDL code). Now when you want to build a design into a chip in TSMC, they will have such IPs. But instead of "Hard IPs" of FPGA, these will be pre-validated and converted into silicon geometry and ready to use IP. You say you need a 64x10 Register file, made out of high speed SRAM on chip, they have the most optimised SRAM possible on their fabrication process. So when you make a new process, you will need such library of IPs.

 

[Black Tornado]

 

[INDIAPOSITIVE]

https://twitter.com/x/status/1634088091886534656

https://twitter.com/x/status/1634157578254000129

https://twitter.com/x/status/1634339819231891456

 

[nahtanbob]

Yes this looks promising.

They are still searching for a technology partner. Vedanta and Foxconn both don't the technology.

Who is the technology partner here ? ISMC isn't a technology company either

Its a cow and a milk thing. Should you buy a cow if you can buy just milk and thats what you need? I think India right now can do fine with importing chips and building products out of them --like phones, laptops, TVs. "Cow" here is REALLY expensive, needs lot of water and needs recurrent expense.

The reason you hear these "Lets get Semiconductor Industry" has more to do with becoming sanction resistant. But thats a 50 year project or more. Even China after pouring 100s of billon dollars in it is not able to achive it.

Unless China breaks the monopoly you will be dependent on the West for semiconductors, semiconductor equipment, EDA tools, raw materials etc for decades to come
You first learn to crawl, then walk, run and then maybe fly

 

[Zhukov]

Its always good to hear advance tech industry starting outside western countries.
But I have a very serious question. Why are the "announcements" such big news in case of India. I mean its a very good developement that Fab industry is getting foothold in India. But be it defense procurement, foreign policy developement, economic forecast, industrial developement etc there is so much hype for announcement even before the start of a project. Is it a cultural thing?
So many newly industrialised countries are achieveing so much breaking the western dominance on industry. Brazil, Phillipines, Vietnam are making a lot of progress lately. But we only hear from them when something is materialized
But you will never find such sensationalism in "announcements of plans". it is just annoying at best.

 

[nahtanbob]

Its always good to hear advance tech industry starting outside western countries.
But I have a very serious question. Why are the "announcements" such big news in case of India. I mean its a very good developement that Fab industry is getting foothold in India. But be it defense procurement, foreign policy developement, economic forecast, industrial developement etc there is so much hype for announcement even before the start of a project. Is it a cultural thing?
So many newly industrialised countries are achieveing so much breaking the western dominance on industry. Brazil, Phillipines, Vietnam are making a lot of progress lately. But we only hear from them when something is materialized
But you will never find such sensationalism in "announcements of plans". it is just annoying at best.

In all fairness they are trying to get a private entity to do something that requires massive financial and technical inputs

 

[INDIAPOSITIVE]

https://twitter.com/x/status/1648583789231370240

https://twitter.com/x/status/1647662353042010112

https://twitter.com/x/status/1648597325298937856

https://twitter.com/x/status/1645321548419710977

 

[Sam6536]

https://twitter.com/x/status/1648583789231370240

https://twitter.com/x/status/1647662353042010112

https://twitter.com/x/status/1648597325298937856

https://twitter.com/x/status/1645321548419710977

Oo mediatek too interesting

 

[nahtanbob]

Oo mediatek too interesting

You are a bunch of financial/business types trying to use government money to enter the semiconductor business. this one is ending in failure.