India Microprocessor

We all know about Simputer and $10 laptop. Not to mention, they were disasters. And now Government of India is talking about its own microprocessor.

The project to make the India Microprocessor, as it is being tentatively called, will see scientists from the Bhabha Atomic Research Centre (BARC), Centre for Development of Advanced Computing (C-DAC), Indian Space Research Organisation (ISRO), Centre for Scientific and Industrial Research (CSIR), and IIT Delhi coming together under the aegis of the department of IT.

Well, all one can say is how bigger disaster this one is going to be? If I am not wrong, DRDO already made computers as in 32 bit micro-processors. ABACUS was touted by folks at DRDO as one of the most competent micro-processors of the time. They even made a 32-bit RISC processor called ANUPAMA. It is also available as an IP core. Now the question to be asked is why are these processors not being used to make computers for use atleast in Indian Governmental services? The answer is terribly complicated. But let me explain why from the little knowledge I have about DRDO and computing technologies. DRDO ANURAG is the Hyderabad Center. They have had some interesting projects to their name. ABACUS, ANURAG are its 32 bit micro-processors while ANAMICA is its Medical Imaging Software. They even made some DSP cores. They also made Gigabit Switches and other networking technologies. They also made some equipment for Parallel Computing. But all this sounds state-of-art, doesnt it? Where does it break apart? A visit to ANURAG achievement page reveals this:

1- µ CMOS Fabrication Technology Facility is available to fabricate designs targeted to 1µ technology with following features:

  • Cell based designs upto 1,00,000 gates.
  • Design support option.
  • Full contract design stating from functional description to physical design
  • Full ‘customer’ design [upto GDS II tape stage]
  • Die sizes upto 14 mm x 14 mm

1 – µ CMOS Fabrication Technology? Seriously? Are you kidding me? 1 – µ CMOS Fabrication Technology is technology of my grandmother’s time! This inevitably brings us to how fast these systems are! ANUPAMA works on a 33 Mega Hertz clock. This means ANUPAMA can execute approximately 30,000 30 million instructions per second. Even an undergrad student with an FPGA development kit could achieve this without any problems whatsoever on a simple microprocessor. Most of students in IITs as well as other premier institutes in India develop such computers for their lab assignments! Today the best of RISC processors can execute several tens of billions of instructions. Comparatively, where is ANUPAMA? Where does ABACUS stand? Does ongoing research on this kind of technology excite a graduate to pursue post graduation in India? Does this kind of technology persuade a post graduate to pursue doctoral studies in India? I dont understand what stops DRDO in working with TSMC/UMCA or other chip manufacturing technology gaints to develop these processors and system components using state-of-art sub nano meter technologies. By not opening up on these technologies, MoD and DRDO are living like frogs in a well! Back in Dec 2005, I attended IMAPS IINC conference on packaging and manufacturing technologies held in IITB, Powai. This was the time ABACUS and ANUPAMA were being developed in ANURAG. Out of the three days, second half of the last day was dedicated to showcase the projects being done at ANURAG and other DRDO centers. ANURAG presented its micro-processors. I am sorry to say but both the projects sounded pathetically over-rated! MoD and DRDO need to get individuals from outside and try to develop business plans so that products could be made and sold to general public! DRDO and MoD should work as independent VCs and let the enterprenuers grow businesses and thus enable employment and research on next generation products! If they say we want to develop indigenous technologies in this field, they are not going to evoke even a speck of interest! Today, we are approaching sub 30nm technologies in building chips, especially micro-processors. As Zerone, they want to take a bite into a plausible $315 million electronics market by 2015.

Demand for microchips from India’s booming technology sector is expected to touch $315 billion by 2015, but a semiconductor policy of previous years to encourage firms to manufacture them locally evoked little interest from the private sector …. The chip could also help India develop a low-cost mobile phone, worth say just Rs 500, high-tech defence precision systems and a host of other applications in areas including healthcare and weather forecasting. “Apart from defence sector, it will be a shot in the arm for the $10-billion IT hardware industry, if the government is successful in its move,” said Vinnie Mehta, executive director of hardware industry body MAIT, who was present at a recent meeting on the issue.

There they go again! Rs 500 worth of mobile phone? Shot in the arm for IT hardware industry? Dude, SPARC as a workstation is long dead! You expect companies to switch to OpenSparc based systems? And even as you are busy doing this, do you expect IBM, Oracle and otehr gaints to be quiet? Get a grip. IT hardware industry is already talking about server sharing, cloud computing? Lets not even get into that now! Well for one thing Zerone wants to use OpenSparc in its efforts.

The India Microprocessor is likely to adopt Sun Microsystem’s Open Sparc open source chip design technology, along with Linux operating system and MySQL database software.

But if they seem to forget the cut-throat competition in this field, they are mistaken! In today’s electronics, systems sell even if chips dont! If MoD and DRDO wants to really make any difference or wants to boost the snail-paced growth in electronics and chip development sector in India, they should

  • allow people to leverage on these technologies
  • allow enterpenuers to use these technologies in designing products
  • stop being a producer of products but start being enabler of products
  • stop being proud of 1- µ CMOS Fabrication Technology because seriously, no-body gives a damn if you are not in talking sub micron technology today!
  • start guiding the enterprenuers to build systems based on these technologies so that they can sell them in India as well as other countries

If the protectionist route is followed, newer, more efficient industries will have less scope to expand, and overall output and economic welfare will suffer” – Alan Greenspan

Advertisements

7 comments

  1. […] CDAC and ISRO was supposed to develop a micro processor called “India Microprocessor”. An analysis of this endeavor was done on in 2009. The project was ambitious and a seperate entity called “Zerone Corporation” was […]

  2. […] CDAC and ISRO was supposed to develop a micro processor called “India Microprocessor”. An analysis of this endeavor was done on in 2009. The project was ambitious and a seperate entity called “Zerone Corporation” was […]

  3. […] One of the oft-repeated bullet points in mandatory state sponsored Rajiv Gandhi bhajans these days is that it was Rajiv Gandhi’s PMO that brought telephony to India. As a matter of fact, the statement is true but the key question is why PMO was spearheading it instead of allowing Indian entrepreneurial spirit to bring this technology to people. The C-DoT might have had great achievements to its name but the fact is C-DoT was a State sponsored enterprise. The end result – personal memoirs like this where bureaucrats express their gratitude to the Gandhi family while in reality, government owned enterprises in semiconductor industry have done no good whatsoever! For some reality check, here are a couple of critical commentaries on “achievements” of government owned enterprises in this and other technologies – 1, 2. […]

  4. First things first. Anupama would do a few million instructions per second any day of the week. It does 33 million cycles per second, not 33 thousand. A few MIPS is still nothing to brag about but it does the job of tracking and guiding missiles. Advanced Numeric Processor for Aircraft and Missile Applications (ANUPAMA). The goal of this project was to make an indigenous processor to make sure their are no backdoors compromising our defence systems. And no – engineering students don’t “make” processors for lab assignments. Running verilog on FPGA packages and actually designing and etching a CPU are wo entirely different things.

    Talking of making – they make it at 780 nanometres now. At the time of writing the smallest process technology is 14nm. But those are made by companies who sell hundreds of millions of chips a year and hence make a profit out of it. And the driving force behind shrinking nodes is power efficiency – performance increments have plateaued after 32nm. ANUPAMA was never intended to play video games in a smartphone – it was meant to go into flight and guidance systems where it will be exposed to the elements of nature and expected to withstand EMP attacks. And 780nm makes a lot less heat per unit die area and can withstand higher potential differences dur to the wider depletion regions.

    I agree they have done a poor job of making a processor and our semiconductor industry is lagging far behind. But pursuing a full blown IP and manufacturing ecosystem is a lost cause. it would take years if not decades of in-house research just to get to a respectable process node. And it will still remain unprofitable if it is not outsourced to other chip companies. Outsourcing could be done but designing and making sophisticated IP would have no returns if nobody else is using it. USA is the only country with a proper semiconductor ecosystem. At that semiconductor is the only space where USA has an edge now.

    So ANUPAMA as a processor for military applications is good.

    1. This is a typical knee-jerk reaction to a potent argument. First I would like to acknowledge my mistake. 30,000 instructions was a typo and I didn’t observe it until it was pointed out. Apologies for that. It is now corrected.

      Now to the criticism that you have played out in your comment. Taking statements out of context has become a kind of time pass for most of the literate class, especially those on the internet.

      1) “First things first” – The article is neither about DRDO’s defense technologies, nor about using sub nanometer technologies for defense applications. The article is about shortsightedness of the famous defense establishment. DARPA’s project to connect computers was also a defense enterprise intially but what it is today is a well known fact. The difference is that DARPA wasn’t afraid in opening up technologies to public (after of course sufficiently and thoroughly analyzing it revealing any defense secrets). Thats the crux of the article. But no, how can we understand that? We have to now cry that DRDO is not getting its due.

      2) The exact statement from my article about graduate students doing lab assignments to build microprocessors was “develop” and not “make”. It is of course impossible to see the difference when the reader is not dispassionate. Yes, they do develop quite brilliant microprocessor architectures right in their labs. Why they cannot etch their microprocessors is a different story (of which I have written extensively in another blog post, which I am sure you may find fault with as well!).

      3) Architecture has nothing to do with Manufacturing Process. If you disagree with the established methodology (ala Mead and Conway methodology), it has to be based out of strong reasoning, which I am not sure if you have time or intention to do. Systems in aircraft guidance etc don’t need state of the art but that doesn’t mean you have to limit your designs to measly performance. Process technologies could be limited by the ESD protection required etc but that doesn’t mean designs have to limited in the same way. Spawning out designs from the process technologies will definitely funnel newer and better performing architectures. DRDO doesn’t do that because well, it is in the end DRDO.

      4) ANUPAMA architecture, if opened up to public, it could have probably landed in cell phones and fanned a new set of options to local manufacturers. Our local customer needs are now being met by international players and the result is that most of our electronics consumption is imported. What little is manufactured in India is by MNCs. Thats the loss. ANUPAMA may fly but DRDO’s secretive attitude doesn’t.

      5) In IMAPS-IINC, the presenters displayed proudly that they could run Linux with GNOME (or KDE? … I don’t remember) UI on ANUPAMA and ABACUS processors. I suppose the missile guidance system needs a full blown UI based OS so that pilot can play Super Mario after firing the missile.

      6) Finally, you have completely ignored my observations on the ridiculousness of “India” microprocessors and stuck on to ANUPAMA technology as if thats the final and the ultimate architecture in guidance systesm. You also miss the point about adopting Open Sparc to build hardware for IT industry. My only gripe was that you could have taken any of your legacy microprocessors like ABACUS to achieve it. But no, how can that be!

      All this discussion is moot because the whole idea of Zerone Corporation and India Mircroprocessor was shut down after they realized how useless it is. ANUPAMA stays in the missile guidance system. You may cry at a different blog where other great technologies that DRDO developed are being ridiculed.

      1. “First things first..” point 5 is hilarious..
        I failed to see that this article was about the shortsightedness of the Indian weapons program. In which case I should concede the discussion. I totally ignored the part where you mentioned Zerone and the India microprocessor. The same way I ignored the original project when news came out. I was skeptical of the project back then. I knew it was headed nowhere. But your article needs a little restructuring.

        Moving on, the DARPA had the first mover’s advantage. They were the first to do it and hence found no competition and quick adoption. And that was in a capitalist America where anything and everything found investors back then. Not to mention America already had the base of core technologies. And to my knowledge ANUPAMA and ABACUS are already available as VHDL code. Did it find any adopters? Did it make it into any phone? Let me tell you why it didn’t work.

        Say I run a company that “makes” chips. Why would I pour billions of dollars into an unproven CPU architecture? I could as easily get an ARM core to do the job. Let’s say I turn out to be a risk-taker and start making ABACUS core chips for phones. Do note that the cores account for less than 20% of the die size. As you know, a chip without software is just a piece of rock. Quite literally. So to see my chips in any phone in production I would need to port the OS at least. So either I or DRDO must develop a compiler for the ISA (instruction set architecture). An activity allegedly so complex that only Intel embarks on it in the x86 space. After I make a compiler I would need to port an OS which would be easy for open source projects like android and impossible for proprietary OS like Windows Phone. Let’s say I managed to port one version of android after getting hold of the android and debugging it enough to make it run properly on my chip. What do I do when the next version of android comes out? Repeat the whole process? Any sane person would give up by now. Or bribe google into supporting the ISA. Had it had a sizable user base, Google would have automatically done it. Even with that problem out of the way, we would have to deal with potential compatibility problems with the 1 million apps in app store. The bytecode ones included. In practice there are compatibility issues between even Snapdragon and Exynos chips – both of which are based on the same ISA. Then what do you think would come out of cross ISA operation? They won’t even make angry birds for my Lumia and you expect them to optimize their apps for a new ISA. Core technologies are like currency. They have inertia attached to them. You simply cannot expect everybody to use an ISA you came up with last night. A famous example would be Intel struggling to gain market share in the mobile space. They actually pay vendors Rs. 1500 to use their chips.
        There are a lot of deserving architectures out there Sharma Ji. A lot of them proven. None of them make it into devices. If the Sparc didn’t work, why do you think OpenSparc will?
        I admire your spirit. The way you want indigenous industries to usher. To reduce imports. I used to be like that when I was 16. But I am sorry to tell you – semiconductor business is a dead end. The best we could do is set up fabs like Taiwan and Korea.

        “In theory, theory and practice are the same. In practice, they are not.” – Albert Einstein
        I do not know the ala Mead and Conway methodology that you talk of. Pure ignorance. But I have followed computers for long enough to know a few things. In practice, architectures have always depended on manufacturing process. Computer science is a discipline of engineering. Which means it has been dealing with the scale of economies since forever. It has always mattered how complex a design can be shipped economically. In the real world even the business models affect design choices. In the last 4 decades it became increasingly easy to make more complex designs cheaply and make them tick faster. The difference between a complete 8086 system and an i7 (forget speed, look at the system architecture) of today is because of the 4 decades of evolution primarily driven by node shrinks.
        The number of good chips made from a wafer falls quadratically with the size of a chip, which can be explained with simple probability. And size depended on process node. As designs became faster as more clock rates became feasible, a need for instruction level parallelism came up which spawned the development of all the out of order designs and tricks we see today. The discrepancy between different parts of the computer spun off the cache and the idea of dedicated controllers. The growth of all of these were still capped by process nodes. Add to this another multidimensional variable called software and you see what makes theory different from practice. And I haven’t even covered all the variables yet for the sake of simplicity. This comment is already too big.

        Finally coming to the defence part – ANUPAMA is still capped by process node. The first time I read about it, they were making it on a 10 micrometer process. So 0.78 is a huge improvement. And I think abacus was developed to exploit the 0.78 node. And I didn’t say ANUPAMA is enough. For a simple heat seeker, it will suffice. But for something with countermeasure suppression it won’t. From a military standpoint, it is reliable and secure. So the ANUPAMA stays in missiles as God intended. Actually it goes into aircrafts too – but. Meh. I am not defending DRDO. But I would like people to understand its place – a defence R&D establishment. I do hate the fact that they made too many surface to surface missiles doing the same thing and most of them aren’t even operational yet. There is only one expensive air to air missile and no shoulder launched missile at all. We made the world’s most lethal cruise missile and made nothing that could stop it. Also we miss a reliable and effective automatic rifle and other small arms are non-existent. It is only recently that they started working on armored personnel carrier. But we also need to consider the fact that the entire burden of R&D in our country rests on a handful of public sector organizations like DRDO. The only solution is to completely remove all the caps on defence R&D in our country and privatize these companies.

        So the assumptions that make the basis of this article have some glaring inconsistencies. Criticizing anything and everything government does has become a kind of time pass for most of the literate class, especially those on the internet…
        Otherwise I’m right there with you. The “India microprocessor” was a bust. It wouldn’t have succeeded. No “new microprocessor” project has ever succeeded. That’s just how things are. We should instead invest in fabs.
        I would also like to read your other articles on technology. Good desi bloggers are a rarity.
        Meanwhile scientists at DRDO have started working on a new missile because there was no match on TV today.

      2. Such a long comment itself shows incoherency in your thoughts. The post was from 2009. Andriod was announced in 2008 (September?) and really took off only in 2010. There was a clear opportunity for Indian scientists to drive an indigenous processors for Indian markets. Coming to ANUPAMA. I think the links in my post are dead because the sites have been moved to .gov.in. Heres what DRDO ANURAG site declares proudly:

        “ANURAG was established on May 2nd 1988 to execute specific, time-bound projects/programmes leading to the development of custom designed computing systems and software packages for numerical analysis and other applications. To design and develop advanced computing systems, especially those based on state-of-the art concepts like parallel architectures as well as associated systems and sub-systems. To develop systems and application software packages for mission oriented tasks to build-up technology in these areas. To undertake forward-looking research and development in futuristic concepts and to create an environment and infrastructure which fosters such advanced research leading to the application of advanced computing concepts and technologies.”

        These chips were not made for specific applications, they were General Purpose.

        Under the head “General Purpose Processors“, they proudly say:

        “ANURAG has designed and developed general-purpose microprocessors. ANUPAMA is a 32-bit RISC processor. ANUPAMA works at 33 MHz clock speed. The complete software development tool kit consisting of C compiler, assembler, linker, loader, debugger etc is available for application development. A single-board computer based on ANUPAMA is available for evaluation and software development. ANUPAMA is also available as an IP core.

        ABACUS is a 32-bit processor for multi-tasking applications with virtual memory support. It is designed around ANUPAMA core with additions like MMU, two levels of cache, double precision FPU, SDRAM controller. The IP core of ABACUS is available in Verilog RTL code. This processor is suited for desktop applications. Complete software platform is available for ABACUS processor and a single board computer with ABACUS is implemented. Linux Kernel is ported.”

        We are the biggest emerging democratic market. So if you quoting geo-politics and America’s advantages, I don’t buy it. We had a clear opportunity to make the mobile world do our bidding. While political will was totally lacking, DRDO failed on technology front. If only …. DRDO had the propensity to bring out these technologies (not just provide Design Kits etc.,) but collaborate with enterprises, drive them, it would have led to a different mobile manufacturing statistics in India.

        Fab is a failed game for India. We are too late. It takes decades of continuous research. With the tech nodes already hitting sub-nano meter levels, there is nothing much we can do. If anything, we have a bigger role to play in VLSI Design Testing (post silicon) and Verification (pre silicon). On this topic, I have written elsewhere. Instead of finding inconsistencies in my study, you should probably build a reasoning based argument on why India continues to lack in electronics industry. That would help public discourse.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: