Indian Electronics Manufacturing Developments : News, Updates and Discussions

Atleast now Tata is entering in the components manufacturing for apple and other OEMs and investing a billion dollar in tamilnadu.....
All that they will be doing is assembling the components that will be sourced from south east asia, at the best we might manufacture those pcb's and outer casing. soc chips , display screens and other critical component will still have to be imported.

Remember Tatas also build fuselage for chinook helicopters, that does not make them a aircraft manufacturer. We still have a long way to, we need to capability to manufacture the critical parts on our own. We should simply create another glorious HAL.
 
BTW, did you check the price of the same processor in 2020? Quoting 2017’s price in 2020? Very relevant!!
i9 is a higher core count processor targetted towards workstation crowd. Its not what most of the folks in India will buy as a consumer product. Its more suited for someone buying for a CAD/CAM work or for video editing workflow or likewise. Even for software development its not exactly needed. Even gaming crowd won't buy it, gaming crowd needs great single core performance with a really powerful GPU. This proc is solid performance on per core but massive number of cores and perhaps PCI express lanes to match.

I doubt seeing a 10 million sales per year of this processor.
 
Its not possible to "slap sanction" on a commodity items like commodity SoC. You can deny technology but trying to ban sale of commodity items is very very very hard to implement. There are many ways it will get circumvented. Supply chains in this are pretty long. You can ban sale of SoC chips, but can you ban the sale of PCB with SoC mounted? Or PCBs with SoC mounted together in a knocked down kit? At one point or the other you realize, its too difficult to enforce such a ban.

What is a chip ban btw?
Any chip having a core that is designed in America?
Any electronic PCB having such a chip?
Any SoC having such a core with other cores?

How and at what point will you ban it? And how will you enforce it?

Best you can do is what US is doing with Huawei. Ban sale of IP cores and software tools and put policies to deny access to market.

Rest all bans are impractical. One cann't legislate the reality.
We are talking about mass scale, even china is vulnerable. Once US asked tsmc , qualcomm to stop selling to huawei it was game over for them. Huawei reportedly set to sell Honor budget phone division for $15B – TechCrunch

When I say chip ban, china will no longer able to buy or get chips manufactured from private players like TSMC or use IP from ARM or Qualcomm.
Bcos all these players use US intellectual property. You are living in lala land if you think they cant enforce it on big companies. US forced UK not to sell ARM to china , it got bought over by nvidia. US will force every other country in the world to ban products that violate their IP.

This is not iran or north korea smuggling few chips by creating fake companies to run their nuclear program, this is war at large scale.
 
We are talking about mass scale, even china is vulnerable. Once US asked tsmc , qualcomm to stop selling to huawei it was game over for them. Huawei reportedly set to sell Honor budget phone division for $15B – TechCrunch
Qualcomm does not manufacture chips. Not anymore. Its an IP licensing company. US can block sale of their IPs to India at best that all. Thats what they are doing to Huawei. The solution to this is simple. Get that IP from elsewhere, there are others who also provide that IP. OR get your chips manufactured with the help of a 3rd party consultant who have access to qualcomm's IP as a part of their own portfolio. US can block sale of IP to Indian companies, how will it block someone else to sell you products built around that IP? Especially if the same IP is available as a part of consortium licensing obligations? You block the sale of your IP, we buy products from someone else whom you have licensed your IP.

US cann't order TSMC and even if TSMC tips toes, Indian companies can get their chips manufactured by another SoC vendor with the customization they need. Actually thats how most of the product development happen. You approach a SoC vendor and get a semi-customized SoC manufcatured with your own custom cores as well. You ban TSMC from manufacturing chips for us (supposedly) and we buy manufactured chips from another vendor with our customizations done. The industry is too complex and layered to be legislated.

US can hold absolute grip on say US companies but it cann't hold same grip on non-US companies. No one will police for US.
 
i9 is a higher core count processor targetted towards workstation crowd. Its not what most of the folks in India will buy as a consumer product. Its more suited for someone buying for a CAD/CAM work or for video editing workflow or likewise. Even for software development its not exactly needed. Even gaming crowd won't buy it, gaming crowd needs great single core performance with a really powerful GPU.

I doubt seeing a 10 million sales per year of this processor.
i9 is outdated and outpriced by ryzen9 processor, proof that better technology & competition lowers the price.
These processors are now more affordable in developed countries but not in India. Thats our problem.

Your argument of saying it is suited for CAD/CAM is not correct, if it is affordable why would not rest of the ppl buy? Would you buy a dual core processor if 8 core is affordable and makes your tasks run faster.

My first computer had a single core proc that was bought at astronomical price and now I use a 8 core proc waiting to move on to 16 core. Now this 16 core procs (same as i9) are becoming consumer grade processors they are no longer the privilege of enterprise companies. It is only that Indians are falling behind in buying capacity.

You are simply confusing things we are talking about mass adaption and availability for economy as whole for the country to progress and compete not just the gaming crowd. ARM and AMD is coming out with 32 , 64, 128 core chips which will become norm in the future while your banks will still be stuck with processing transactions on 4 core chips.
 
Qualcomm does not manufacture chips. Not anymore. Its an IP licensing company. US can block sale of their IPs to India at best that all. Thats what they are doing to Huawei. The solution to this is simple. Get that IP from elsewhere, there are others who also provide that IP. OR get your chips manufactured with the help of a 3rd party consultant who have access to qualcomm's IP as a part of their own portfolio. US can block sale of IP to Indian companies, how will it block someone else to sell you products built around that IP? Especially if the same IP is available as a part of consortium which usually telecom IPs are.

US cann't order TSMC and even if TSMC tips toes, Indian companies can get their chips manufactured by another SoC vendor with the customization they need. Actually thats how most of the product development happen. You approach a SoC vendor and get a semi-customized SoC manufcatured with your own custom cores as well.

US can hold absolute grip on say US companies but it cann't hold same grip on non-US companies. No one will police for US.
I dint say qualcomm manufactures chip it is TSMC and samsung which manufactures chip. you cannot buy IP from any where other than western companies. If that was the case china would have already done that ,there was no reason for them to attempt to buy ARM.

Tell me which non - western company has the IP which is on par to ARM or Qualcomm?

Which company will manufacture your chips? name them.

All these companies including TSMC/Samsung use IP from US , buy equipment from us manufacturers like applied materials to build them.
 
i9 is outdated and outpriced by ryzen9 processor, proof that better technology & competition lowers the price.
These processors are now more affordable in developed countries but not in India. Thats our problem.

Your argument of saying it is suited for CAD/CAM is not correct, if it is affordable why would not rest of the ppl buy? Would you buy a dual core processor if 8 core is affordable and makes your tasks run faster.

My first computer had a single core proc that was bought at astronomical price and now I use a 8 core proc waiting to move on to 16 core. Now this 16 core procs (same as i9) are becoming consumer grade processors they are no longer the privilege of enterprise companies. It is only that Indians are falling behind in buying capacity.

You are simply confusing things we are talking about mass adaption and availability for economy as whole for the country to progress and compete not just the gaming crowd. ARM and AMD is coming out with 32 , 64, 128 core chips which will become norm in the future while your banks will still be stuck with processing transactions on 4 core chips.
Processors like i9 and Threadripper are best suited for workloads where there is lots of Thread level parallelism. Simulation, CAD/CAM, video editing come to mind. Their single thread performance may not be best but they deliver more throughput by parallelizing tasks.
 
Processors like i9 and Threadripper are best suited for workloads where there is lots of Thread level parallelism. Simulation, CAD/CAM, video editing come to mind. Their single thread performance may not be best but they deliver more throughput by parallelizing tasks.
Dude stop teaching me technology I know where they are used. I said multicore processors are becoming pervasive and are consumer grade now.
Your mobile has more cores for the simple reason that it needs more cores to do tasks efficiently. companies are creating multicore chips bcos there is a huge market for that called data center.

your single core performance is for gaming only bcos you are still using DX11 , once DX12 comes on board even gaming will move on to multicore processors. Its only that gaming studios dont optimize or dont have the ability to create their games to run on multicore.

Your are just skimming the surface and forming your opinion.
 
you cannot buy IP from any where other than western companies.
Not completely.

There are different level of IPs.

ISA and micro-architectures.

For instance, ARM sells ARM ISA architecture IP and its implementation (as a microarchitecture). Both are "IPs". A 3rd part SoC can license ARM ISA and implement it itself. Such an implementation is free from US bans.

Take Phytium for instance.


They sell ARMv8 compatible ISA core within a SoC. They hold license for ARM v8 ISA, they have their own or acquired implementation of that ISA and they will sell a SoC customized for your need.
Dude stop teaching me technology I know where they are used. I said multicore processors are becoming pervasive and are consumer grade now.
All I was saying was that i9 and TRs were workstation class processors and they don't scale at same pace as cheaper processors do. They are expensive for more than one reason and not all people need them because they don't have workloads with kind of thread level parallelism they bring.

Consumer grade hardware gone multicore is a very old story from 2005 or 6.
 
Last edited:
your single core performance is for gaming only bcos you are still using DX11 , once DX12 comes on board even gaming will move on to multicore processors. Its only that gaming studios dont optimize or dont have the ability to create their games to run on multicore.
Err... DX12 was launched with windows 10 or did I miss something? That was 2015 and my observation still holds.
 
Not completely.

There are different level of IPs.

ISA and micro-architectures.

For instance, ARM sells ARM ISA architecture IP and its implementation (as a microarchitecture). Both are "IPs". A 3rd part SoC can license ARM ISA and implement it itself. Such an implementation is free from US bans.
ARM is now under NVIDIA its no longer an UK company. chinese have the ability to manufacture processors that is nearly equivalent pentium i5 2500k or some thing like that but it is all under licensing from AMD. This was all before US-china war started. Once china started to get to the top US is pulling all the curtains. US holds all the IP, there is no escape from that.


Only other processor company is russian , the only non x86 based cpu. (Russian CPU is still behind AMD and Intel)

All I was saying was that i9 and TRs were workstation class processors and they don't scale at same pace as cheaper processors do. They are expensive for more than one reason and not all people need them because they don't have workloads with kind of thread level parallelism they bring.

Consumer grade hardware gone multicore is a very old story from 2005 or 6.
thats your opinion not any fact. what do you mean by cheaper processor?
Once it is mass produced it will become cheaper, only reason it was expensive until now was bcos intel had the monopoly at top end and charged more, post 2017 ryzen launch its all changed. Now more ppl have these systems and more programs will be written which take advantage of it. It is affordability that matters.

My first pc had 256mb ram and now my cell phone has 6gb ram. Why dint engineers stick to only 256mb ram and write code for it?
As I said its affordability & economics not parallelism or some technical nonsense, those kind of argument are good for when designing systems for engineers. You are confusing economics with technology.
Err... DX12 was launched with windows 10 or did I miss something? That was 2015 and my observation still holds.
your are only googling and posting answers here, thats the problem. You should have seen the industry grow and change to know more.

DX12 was launched long time back as an interface which would take advantage of multiple cores and use asynchronous computing. It was spearheaded by AMD. But it was just a specification not implementation. GPU manufacturers have to implement it in hardware and software has to take advantage of it.

But as usual industry adoption is not that fast bcos intel never released its hold on cpu monopoly. They only stuck to single thread performance and every year increased incrementally produced processors which were just 200- 400mhz faster than prev generation. Obviously game developers also implemented games assuming every one will have few cores and only few threads will be available for game. So our whole game industry never took advantage of Dx12.

But post ryzen launch once intel monopoly was shattered, market got affordable ryzen pc's every one started to take notice of multicore cpu's. Games performance could now be improved using more cores as more ppl were expected to have such systems. Some game developers like RTS games used multicores easily but most of them were used to single thread approach as it was easier to implement games.
So problem is not with hardware but ppl not able to write software , not all of them have skills to write such code especially FPS games.

Core Scaling and Gaming Performance — How Many Cores Do You Need?

For example latest games like red dead redemption 2 takes full advantage of multicores. More games in future will take advantage of cores, it is just a matter of time. More than that single thread perf has nearly reached the limit it is no longer scalable in physical terms, it would only be through IPC or more cores.


so if you write a code that does not take advantage of all the cores and make it run faster it is your problem, they would say it is bad software implementation.
 
Last edited:
your are only googling and posting answers here, thats the problem. You should have seen the industry grow and change to know more.
Before DX12, there were many attempts to adapt current software development (not just games) into something more amenable to multi core / thread-level-parallelism. I was a part of the team which designed and developed Task Parallel Library and parallel extensions in .Net framework to promote harnessing multi-core systems way way back. The results have been, less than optimal till this day. True some places have shown a lot of progress but its not exactly people have started writing code more amenable for multithreading environments. Why? Its still very hard to write quality parallel code and old languages / SDKs/ habbits die slower. I am still waiting for the functional revolution that was promised in 2004 or 05.

so if you write a code that does not take advantage of all the cores and make it run faster it is your problem, they would say it is bad software implementation.
Errr... That in itself explains why parallelism is not everyone's cup of tea and why a lot of games etc will not be able make best use of multicores. Single core performance will still limit a lot of consumer grade workloads.

Anyways, I will love a discussion on concurrency and parallelism but not in this thread.

My main point is simple : India is not going to become a market place where 2000 or 3000 dollar processors are going to be sold by 10 millions per year anytime soon. That price point is for workstations and servers and they don't scale in sales (in India atleast) as well as consumer grade processor sale.
 
Only other processor company is russian , the only non x86 based cpu. (Russian CPU is still behind AMD and Intel)
Errr... there are quite a few. From the top of my head :

MIPS is owned by Chinese now. MIPS moves to China via Samoa
There is a new RISC-V ISA which China has the fastest implementation : Alibaba On The Bleeding Edge Of RISC-V With XT910

As I said before, the current ban by US government is not on manufacturing muscle but on specific IPs, EDA Softwares, Patents and market access.

It will be circumvented. Lesson for India is to grow its home grown gallery of IPs, EDA software, specification and standards for security and not waste money on manufacturing silicon. Manufacturing silicon is not the sword. Its IPs etc are.

Thankfully, there are new open ISAs, possibly their microarchitecture. They are as free as linux is. More and more companies implement them in microarchitecture, the better.

Top of the line Silicon manufacturing at scale is a losers game. No need to replay that level. We did it in 2004 or 05 and it came undone. Better climb the value chain and invest in IPs, EDA tools, standards.
 
Before DX12, there were many attempts to adapt current software development (not just games) into something more amenable to multi core / thread-level-parallelism. I was a part of the team which designed and developed Task Parallel Library and parallel extensions in .Net framework to promote harnessing multi-core systems way way back. The results have been, less than optimal till this day. True some places have shown a lot of progress but its not exactly people have started writing code more amenable for multithreading environments. Why? Its still very hard to write quality parallel code and old languages / SDKs/ habbits die slower. I am still waiting for the functional revolution that was promised in 2004 or 05.


Errr... That in itself explains why parallelism is not everyone's cup of tea and why a lot of games etc will not be able make best use of multicores. Single core performance will still limit a lot of consumer grade workloads.

Anyways, I will love a discussion on concurrency and parallelism but not in this thread.

My main point is simple : India is not going to become a market place where 2000 or 3000 dollar processors are going to be sold by 10 millions per year anytime soon. That price point is for workstations and servers and they don't scale in sales (in India atleast) as well as consumer grade processor sale.
.net ? really enterprise software gets written in .net? functional programming is already happening it is just not every body's cup of tea. languages like java have already embraced functional programming and have introduced features from other languages like scala,haskell..etc
What do you think kafka or spark runs on? Most of the telecom, high frequency trading,banking software is written in functional programming languages bcos of their support for asynchronous programming ground up. Those $2000 cpus you are cribbing about were $11000 before amd came into picture, they will dominate the data centre where the big market is. In another two years those will drop to $500-800. There is a good reason why every thing runs on multicore from ur pc to mobile, single thread is not scalable. Soon we will see mobile computing converge with pc computing while your still cribbing about cost, cost is India's problem, apple still sells $1200 iphone to masses.
 
Errr... there are quite a few. From the top of my head :

MIPS is owned by Chinese now. MIPS moves to China via Samoa
There is a new RISC-V ISA which China has the fastest implementation : Alibaba On The Bleeding Edge Of RISC-V With XT910

As I said before, the current ban by US government is not on manufacturing muscle but on specific IPs, EDA Softwares, Patents and market access.

It will be circumvented. Lesson for India is to grow its home grown gallery of IPs, EDA software, specification and standards for security and not waste money on manufacturing silicon. Manufacturing silicon is not the sword. Its IPs etc are.

Thankfully, there are new open ISAs, possibly their microarchitecture. They are as free as linux is. More and more companies implement them in microarchitecture, the better.

Top of the line Silicon manufacturing at scale is a losers game. No need to replay that level. We did it in 2004 or 05 and it came undone. Better climb the value chain and invest in IPs, EDA tools, standards.
Thats the mistake Indian did , to jump into services sector without mastering manufacturing sector for which we are paying a heavy price. Even if you have IP's or software you need to have balls to ensure it is not violated anywhere which India hardly can do. For western countries they mastered and moved up the value chain hence spun off manufacturing to china. They still have a tight grip on hardware, you cannot give example of few RISC-V which are custom manufactured for super computing needs as an example. I worked on those systems (HP-ux ) long time back havent heard about them afterwards. Long time back when there was some thing called solaris, Sun microsystems offered machines for free if you bought license then. Guess where there are now, in historys dust bin. EDA software, you mean morons who release software once in 5 years? Only reason they exist is bcos is to service other morons who forgot to board the train post dot com crash.
 
Thats the mistake Indian did , to jump into services sector without mastering manufacturing sector for which we are paying a heavy price. Even if you have IP's or software you need to have balls to ensure it is not violated anywhere which India hardly can do.
Errr... ummm... In semiconductor manufacturing, IPs are not useful commercially unless you do actual manufacturing. Also, EDA software while prone to piracy are seldom used in pirated version in any major design and manufacturing houses. Its not worth to cheap out on IP cost while manufacturing millions or billions of chips.
So, any manufacturer worth its salt just licenses it. It keeps their life simple.

They still have a tight grip on hardware, you cannot give example of few RISC-V which are custom manufactured for super computing needs as an example.
Errr nope. There are quite a few.

MIPS comes to mind and China has been manufacturing it for quite sometime. Loongson - Wikipedia . RMS famously used or uses a Lemote notebook based on these.

There is x86 microarchitecture available from VIA and its Taiwanese partners. Its used in a number of single board computers used in industrial applications.

UltraSPARC RTL code is available under GNU GPL if you want to make your own chips.

RISC-V are available in all types of targets like Supercomputer, embedded etc. Heck our own Shakti Processor from IIT-M is actually an implementation of RISC-V microarchitecture. RISC-V based dev boards are available in plenty. RISC-V is roughly looking same what Linux kernel was looking in 1996-7. Its a royalty free, Open ISA with possibly many many implementation (like opensource IIT-M Shakti processor).

PowerPC is ISA available as an open ISA under OpenPower Foundation initiative. Its members are not just US enterprises. And no, IBM cannot stop anyone from getting PowerPC chip manufactured if they take a vendor's microarchitecture. Its VIA all over again.
 
Last edited:
You want to use technology that is nearly 5-10 years old? why spend so much for it just buy the finished product then. Companies have moved on to 14nm bcos they are power efficient, better yields and to make profit. Simply put 65nm/32nm will not be profitable and sooner or later we will be shutting down or asking for 14nm to break even.

Yes thats the right thing to do , it is a very critical industry. Given that we dont have private players pitching in , govt should step in.

In terms of numbers, most of the processors today still use the 45-180nm processes. I suppose the main agenda for our own fab plants is to tap the 45 to 180nm bracket first. The military also use processors within this bracket. 32nm and below requires brand value, IP, market etc which doesn't exist yet.

This is what china is facing now, they are importing most of the chips used in their products. They are virtually at the mercy of US and their allies.

We are in a position that is 10 times worse than china , in future it will become even more horrible as our economy grows. It will be just like 1960-70's food security, we will be at mercy of external forces.

Exactly. Hence the very strategic need to create capacities at home so that we are not blackmailed later on. All the main topics here concern strategic policies, but very few have been able to think in those terms.

There you go, @Bali78.
Technology takes up 21% of China’s imports, with US$449 billion in 2018, up 19% year-on-year. Technology has become the largest part of China’s imports.

Semiconductor imports have reached US$311 billion. It has become the largest part of imported technology, accounting for 70% of total technology imports.

Semiconductor memory imports amounted to US$122 billion, with 27% of total technology imports in 2018. Imports of other semiconductor products were worth US$189 billion, accounting for around 42% of total technology imports.


And as jetray has already pointed out, it's gonna be even worse for India.
 
IF 200 million people become rich enough to afford a new iPad in India each year, then we are looking at a bigger economy. That or cheaper knock offs of iPad or both.

No, we are going to see vast income inequality because half our population is still farming and don't give a damn about education. And we are not in a position to offer manufacturing like China does because we have to compete with a lot of other countries that compete with us, especially those with simpler regulations.

By 2030-35, we are going to get a very large population that can afford expensive gadgets, say about 200 million, while the remaining 1.2 billion or so won't be able to. But there's the trickle effect. Once technology reaches the hands of one group, companies will start offering cheaper versions of the same technologies to those who can't afford the high end versions. And these are going to be at least 2 or 3 years behind, or you can say at least 1 or 2 generations behind, the high end. Which is why you see iPhones at one end of the spectrum and a Redmi phone at the other end.

Now the govt can't do anything about the high end segment, but they can do something about the low end. But the real back breaker will be the low end segment because of the volumes, since at the high end, taxes can take care of some of the pressure.

India's overall GDP is still going to be small and we will still continue to have a weak exchange rate for a long time. It's expected that at constant prices, India's per capita income will be around $5000 in 2040, which is just $1000 over the threshold for calling ourselves an upper middle income country, something China achieved in 2011. So at constant prices, at $7.5T, our GDP will be half the size of today's China in 2040. But if the market climbs to even 50 million high end $1000 iPhone sales a year, we are talking about a forex outflow of $50B, which is half the size as our oil bill today. And this is merely iPhone sales. And 50 million is a significant underestimation. What do you think will happen if Apple ends up with a 50 million market before 2035?

But what about $300 phones? And what if the market in the 2030-35 period is 250 million a year for $300 phones? We are talking about a forex outgo of $75B. This is the low end segment.

Your PPP numbers don't matter here.

China will escape because, as an advanced economy, their yuan will find international buyers. But no one's gonna give a crap about our rupee unless we become at least close to an advanced economy, say $10,000 per capita.

What won't work is India subsidizes a massive fabrication ecosystem effort only to realize that it cann't offer the same prices as taiwan or china even while selling at cost price due to sheer scale of economies.

Actually this entire business of have a fab is a deja vu. It has happened before in 2000s.

It doesn't have to be 100% of the market. It needs to be enough that we are in a position to scale it up in an emergency so that we are not blackmailed. We are going to get into a position in the near future where public blackmail will be the main way to get something out of India. You already see that happening with China.

The Chinese made a huge mistake of not entering the electronics market, especially fab plants, even earlier that they did, which they admit. This has nothing to do with making a profit. Countries like India and China pay many times more than the profit margin of the semiconductor industry for security. Even if a fab plant makes a $1B loss every year, I doubt that's possible, it's peanuts compared to the overall security budget and will be subsidised for the sake of economic security. It's a strategic industry and making a profit is a secondary consideration.
 
  • Agree
Reactions: jetray