REGISTER | LOGIN
Breaking News
News & Analysis

India Preps RISC-V Processors

Two 64-bit projects run in parallel
1/27/2016 10:20 AM EST
34 comments
NO RATINGS
Page 1 / 2 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 4   >   >>
GSMD
User Rank
Author
Re: How do you get paid for open source work?
GSMD   1/31/2016 9:43:39 PM
NO RATINGS
1. We are not talking caps. Pure capitalism. As the other gentleman pointed out, we have lots of folks here , so salaries follow the law of supply and demand. Checked with the  heads of HR of 3 large companies (more than 10B USD turnover). My numbers are 40% inflated. Senior VPs make only 8k pm. A good condo (3 bedroom) in the near suburbs is only 100k, so we  are talking 2 years salary to pay of a house. These are high salaries by any yardstick. Ignore the high salaries paid by Google and Amazon and their ilk. They represent a miniscule majority who can justify salaries based on US sales. Real world salaries are much lower. I should know, our students are among the highest paid kids on the planet. Silicon valley CXO levels have lots of IIT folks.  But every engineer does not command IIT-M level salaries. The city of Chennai alone churns out 100k engineers (mostly unemployable) every year. Starting salaries in a country of high inflation have remained at 300-400 USD for the past 4 years. My building super. makes the same salary as a Java programmer !

2. That is precisley what I am saying. Open source will have a major deflationary effect. I am currently coordinating smart card, id card and IoT programs. These will go the reverse auction route of solar panels. Commodity IP manufactured by multiple companies with the auction discovering the lowest possible price. Welcome to the new dystopian world of low cost VLSI ! If you want pricing power, you better have a great story which the vast majority of companies will not have. By the way the ARM based industry looks exactly like this now. Differentiation is next to impossible in the sub USD 10 parts. So I am not forecasting something new. Ask Qualcomm.

3. Please stop quoting Govt. initiatives. At best the Govt. should just seed and we all agree on that. So does the govt. for that matter. That is why this program is staffed by industry veterans who know how to get real life products out.  All seeding programs are funded with the caveat that the private sector should take over .So the govt. also agrees with you. Akash was a really bad idea and I opposed it too. They should have just bought tabs in volumes and driven prices down. Happened with laptops (every govt. run  high school kid in Tamilnadu gets a free laptop)

4. The SOM tablet is my program., so you are talking to the culprit in chief ! It was not meant to revolutionize tablets. By the way, my company designed a tablet for Ericsson back in 1998 (Strong ARM 100 Mhz, mono. screen). So I know my tablets and the economics of manufacturing. I also used to be in the advisory board of a contract manufacturer.  This program is not a wooly headed program to revolutionize tablets. The BOM cost of the SODIMM itself woudl have killed costs. It is meant for high security tablets and the program is doing well. I used an i.MX6 processor in it since it is the best security processor around but is not cost effective. But for a 10$ USD BM increase I get MIL grade security. OS in it is secure L4 (a secure micro kernel). The SW is being transitioned to two phone makers. Has tamper detect with auto erase of sec. keys, secure boot, formally verifiable OS, NFC. HW is currently at USD 300 for low volume (sub 100 pieces). Trying buying a high sec. tablet south of USD 1000.

GSMD
User Rank
Author
Re: Yes Government corruption is the least of India's problems
GSMD   1/31/2016 9:16:31 PM
NO RATINGS
Since this is an august journal dedicated to engineering, let us stick to the facts and apply logic to those facts.

1. I am a so called "higher caste" guy who has no kids, so I seem to be following your advise! I rescue and rehabilitate  injured/sick cats  when I am not desiging CPUs. Way better than having kids !

2. If you look at India birth rate statistics, barring a few northern states, birth rate is pretty low. Tamilnadu where IIT-Madras is located has a birth rate lower than US average. Kerala has Finland level birth rates. So a lot of states are looking at population deflation in the future. But since you are a fan on forced birth control, by your logic we should start with the US. I suggest you bring up the following in the upcoming Republican Convention - federally supported birth control, large budgetary grants for planned parenthood clinics along with federally guarded clinics  and federally supported advertisements of birth control devices. All these are done in India by the way. We just extended our maternity leave (paid) to 6 months to reduce neo natal mortality (this reduces second births).

Loser99
User Rank
Author
Yes Government corruption is the least of India's problems
Loser99   1/31/2016 2:31:35 PM
NO RATINGS
They need to have more birth control or forced sterilization.  Maybe start with the higher up castes.

That would solve even more of Indias problems.

EverHappy
User Rank
Rookie
Re: How do you get paid for open source work?
EverHappy   1/31/2016 1:49:59 AM
NO RATINGS
@GSMD: "I suspect only 10% of the resources of a  propreitary model are needed... Cost per engineer - 4k USD per month. Add two leads at 6k per month Cost of tools zero if Chisel and verilator are used."


Interesting indeed. Two problems here.

First, why put a cap on the resources? So there is the likelihood of greater profit in chip sales and heftier margins on the sales of systems and products up the value chain? Obviously, somebody up the chain stands to benefit at the expense of the engineers designing cores who will only be eking out a living. You can't tell me this open-source initiative at the level of these cores also means chip makers and system and product developers up the chain will equivalently bring their prices down. I'm sure, the chip makers and product developers will let the market determine their price - free enterprise way, essentially. It would be downright unfair making people lower down in the chain eke out a living while those at the top swallow up the savings.

Secondly, those salaries are low even for India, if one only cares to do the math for the cost of living there - again a myth of the past that living and wages in India are cheap. That is why there has been a flight of software contracts and call centers out of India to places more favorable in Eastern Europe, North Africa and elsewhere. Wake up. You have to let the market determine salaries - that is democracy or economic freedom for the engineers. Furthermore, fixing salaries like that is like fixing prices. It will give you the engineers you want, but not what you really want - quality, good work, commitment and efficiency. And, when you don't get what you really want, then starts the downward spiral. Examples of failures with price-fixing are a dime a dozen in India alone - One of them is the under-$50 tablet "Akash" that the Ministry of IC&T had embarked on 5 years or so ago. The intent was for every student to be able to afford a low-cost tablet. Noble thought indeed. But if people ever understood how economics works (don't suppose the politicians in Delhi care or ever did), you don't ever fix prices. However, the Indian Government did just that at $50 a piece. And what did they get in return? An "Akash" tablet that fit the $50 bill exactly, but with substandard components and such poor assembly that there have been cases where this "Akash" literally came apart while merely unpacking. Look at Amazon on the other hand, they finally did bring out a relatively robust $50 tablet in late 2015 with exactly the capability wanted in the Akash for students - customer service and tech support with Mayday Screen Sharing that lets an Amazon expert guide you remotely.

More on why the open-source hardware model would not work. You indicate you have lined up industry members for this ecosystem. Wonder who they are and how far they will go with this open-source initiative. For an academic in India in the past had put together a System-on-Module (SOM) based tablet design. A novel revolutionary idea it was because a user need not change the entire tablet for an upgrade, only the SOM in it. This academic then spoke with a system developer or manufacturer about possible productization. The manufacturer liked the SOM idea very much, but not the open-source nature of the design. The manufacturer wanted the ability to differentiate in order to compete, which this open-source thingy kills.

EverHappy
User Rank
Rookie
Re: No Doubt A Splendid Intellectual Exercise
EverHappy   1/30/2016 11:35:44 PM
NO RATINGS
@krste: Once upon a time, there were many commentators making disparaging comments about how Linux would never compete with a serious commercial Unix like Irix, or Solaris, or AIX, or HPUX, or any number of dead or dying Unixen.

 

Those commentators were damn right! Linux did not compete, it killed competition because of the so-called open-source veil it presents behind which large companies sponsored their proprietary projects and gave it an overwhelming edge. Competition means a level playing field. Linux instead just took that away for their vested interests.

Just because Linux probably has all the features that its users would like in a standardized piece of infrastructure, it does not imply quality or make it the best. Competition alone ensures quality. Differentiation is one of the keys to quality - while product x may be efficient at task A, product y can be more efficient at task B even though product x is also capable of it.

Furthermore, there is this myth that open-source will address every use case out there with ease and you will not have to be in the clutches of makers of proprietary stuff. Unfortunately, for software/hardware to work efficiently in the various use cases possible, there absolutely has to be some differentiation for specific use cases in question. Afraid the one-size-fits-all model would not quite fit when it comes to efficiency and quality.

Who knows? There might be a whole lot of OS users out there feeling Linux is not their choice, but have to be content with it because of lack of alternatives owing to lack of competition. So presenting open-source as the panacea for the ills of being in the clutches of makers of proprietary stuff is probably not correct. Competition is the solution for which governments need to create an atmosphere of enterprise. Of course, that is like hoping for the impossible in India - with the 60 or 70 odd licenses entrepreneurs in India have to seek for starting an enterprise, even seed funding from any government source or VC will not suffice. Unfortunately, people in India also never appreciate the meaning of quality born out of freedom, given its history of governments trying to tell them how to live their lives and providing them with jobs rather than an atmosphere where people themselves can start an enterprise, employ and alleviate poverty amongst their brethren significantly. Since there is an economic value associated with every deed, democracy is actually economic freedom, not the voting process itself as most people tend to believe.

Matter of fact, this effort at IIT M is indeed commendable, but would really be relevant or sustainable socio-economically only if the person driving it were to kick off an enterprise with it, saying he believes in its success through a competitive enterprise. Calling this effort open source and done in the collective cause of society are mere red herrings. Adam Smith, the reknowned philospher, over 250 years ago had observed thus: "By pursuing his own interest he frequently promotes that of the society more effectually than when he really intends to promote it. I have never known much good done by those who affected to trade for the public good." And, by the way, this also speaks aptly to the world of Linux.

GSMD
User Rank
Author
Re: How do you get paid for open source work?
GSMD   1/30/2016 11:23:15 PM
NO RATINGS
This is another fallacy in the open source world. There does not need to be a whole lot of open source projects. There just needs to be a few. The rest are a distraction. I do a few of this and so I am dismissive of my own efforts ! Only 1 or 2 linux distributions are needed. The rest while good for diveristy do not add a great deal of eceonomic value. So as long as the core open source projects get funded through voluntary efforts, the model works. HCL where I started my career, was  contributing to X windows and BSD right from the mid 80s. Very profitable company and one of the leading companies in India today. So I am talking from a position of strength having worked only in open source for 3 decades now. I and the companies I worked for have done Ok financially !

In this case also the open source components is the ISA and the cores and key IP. This does not need more than 50-100 odd engineers around the world. I am talking about lower end cores. Server grade cores have different issues which I will address in another blog. The myth here is that tech industry needs large resources to develop and maintain core technologies. Once IP is shared, I suspect only 10% of the resources of a  propreitary model are needed. All the nay sayers here, refuse to go into the nitty gritty and work out some numbers. Get down to the basics - cost of engineers, cost of tools, verification lifecycle costs, cost of fixing bugs, cost of hosting, cost of support. Then some clarity will emerge. A Cortex A5 level core needs a team of 15, 10 in a pinch to develop and maintain (AHB level bus included). Cost per engineer - 4k USD per month. Add two leads at 6k per month Cost of tools zero if Chisel and verilator are used. Cost of FPGA boards - 4 boards - 750 USD per unit. We can get this free. You can have verified, FPGA proven RTL. Deliverable - Core + L1/L2 cache + bus + I2C + SPI + UART + JTAG + DMA + PIC. Cortex A72 class OO core, add 5 more members and increase FPGA costs to 10K. Thsi is what an open source foundation has to spend. This is FPGA level development. ASIC tapeout is more exepensive but Cadence and Synopsys aid in this. Service companies do provide free P&R services. Shuttle tapeout costs are low and you can get 2 free tapeout per year for open source foundations.

 

Now show me your numbers .....

DMcCunney
User Rank
Author
Re: How do you get paid for open source work?
DMcCunney   1/30/2016 10:06:37 PM
NO RATINGS
@GDSM: Please take a look at projects like Kudu or Docker or Spark and the myth of open source projects written on the side should vanish.

I did not say open source projects were written on the side.  I said most folks contributing code to open source projects were doing so.

How many people committing code to the Linux kernel or LLVM are being paid to write that code?  I think you'll find it's a small fraction of the total number with commit access.  And the ones who are paid fall into the category I mentioned of having a job with a company that uses open source code and pays engineers to work on what they use.

Once you get past the high profile stuff like the Linux kernel, LLVM, or the efforts you mentioned, you'll find an enormous number of open source projects that essentially are written on the side, by developers scratching a pesonal itch but doing something else for money.

>Dennis

GSMD
User Rank
Author
Re: How do you get paid for open source work?
GSMD   1/30/2016 9:44:13 PM
NO RATINGS
Please take a look at projects like Kudu or Docker or Spark and the myth of open source projects written on the side should vanish. I supposed GCC and LLVM and the Linux kernel are also hobby projects ! Open source is inherently a lower margin business, as long as you accept that, the economics work. 

I always advocate this, do not read books on open source, look at the balance sheets of companies who have open source as their core business. We have this bizarre notion that technology development is inherenty activity. In open source HW it is only fabrication that is high cost. The manpower costs are bloated due to tool costs and odd salary structures. For example in a company I used to run, the PCB P&R guy was paid 20% of the IC P&R guy. Granted the later requires more training but we did the smart thing and retrained PCB designers ! Cost of IC P&R manpower went down quite dramatically. In India a senior VLSI engineer can buy a small condo in 2 montths of salary. That needs to change and efforts like these will accomplish that by democratizing knowledge in this area. UCBs chisel is based on Sacla, would be interesting to at least get some QA enginners from the Java world to run Chisel tests. Salary differential between a Java test engineer and a VLSI RTL test engineer is 1:4. It is time we stop putting chip design on a pedestal, at least in the front part it is just another piece of SW engineering.

GSMD
User Rank
Author
Re: No Doubt A Splendid Intellectual Exercise
GSMD   1/30/2016 9:31:58 PM
NO RATINGS
1. Whether you like it or not, open source and a commons based services model as opposed to a product model does have political connotations. But that is a blog for another day

2. A lot of assumptions are being made here. It is the UCB effort in fact which has more of a Govt. involvement ! The Indian effort at IIT is a just a seeding effort from the govt. This was assumed right from the beginning. Everything from the RTL synthesis downwards (ASIC synthesis, P&R, board schematics, PCB design) has significant private sector participation from day 1. In fact we have no board level or P&R expert in house. All that is provided gratis by private sector participants. So I would consider the case closed since the private sector participation is now well established. The problem I think most folks from the west are facing is that the transition from a product based industry to a services industry is not readily apparent. Once core IP is commoditized, there is no such thing as a product. Is RH Linux a product or a service ? It is branded and packaged but ultimately only numbers matter, technology never matters. The numbers say it has the margins more akin to a services company. The obscene gross margins in the semicon industry are a thing of the past. Even physical parts are services in the sense that the financing model and the operational model is more akin to that of a services company. Please do not hypothesize, look at the balance sheets of a few top Indian services companies, look at their IP offerings and then make up your mind. Product IP has no value, only process IP.

A bit about my background. I am 3rd generation CFO and a second generation CEO. Been helping my Dad with his private equity stuff since I was 15, so I know the nos. better than most. I am also incubating two startups. I just advise and assist the govt., so please do not assume that this is a major govt. controlled initiative. 

I know of at least 3 FPGA companies who have enquired with us or our partners  for using RISC-V in their FPGAs and one major mobile phone vendor who wants to trial a RISC-V APU in their tablet. 

To add a bit of perspective, chip development costs tend to be bloated for a variety of reasons (another blog on that). But the combination of HGLs, open source and formal techniques can bring that down dramatically. So please do not extrapolate using existing economic models. I am not talking wooly headed Govt. projections. These are nos I give to VCs who know how to sniff out bad financial projections.

To sump up, the private sector in India agrees there is money in and is putting its money where its mouth is. Krste can shed light  for the US market.

DMcCunney
User Rank
Author
How do you get paid for open source work?
DMcCunney   1/30/2016 3:45:56 PM
NO RATINGS
@krste: Open-source hardware should have a much easier path to monetization than open-source software for the simple reason Andreas stated - you can sell chips! 

But can you sell them for enough?

I had a back and forth in email a while back with open sourrce advocate Eric S. Raymond, where the question was "How do you get paid for writing open source code?", and the answer was "You get a job with someone like Google or Facebook that uses open source code and pays engineers to hack on what they use."  Most open source code is written on the side by developers who make their actual living writing proprietary code.  They contribute to an open source project because it's something they use, and in some cases for bragging rights.  (Having commit access to something like the Linux kernel repository is a major geek status marker.)

You can indeed sell chips.  Making any money doing it is problematic.  The nature of the industry is that all hardware eventually becomes a commodity, with commodity pricing, paper thin margins, and the need to sell ginormous amounts of whatever it is to make any money at all.  As a rule, you largely can't make money on hardware.  The value is in what the buyer can do with the product containing the hardware, and that comes in at a higher level in the chain.

Open source hardware is an example of commoditization.  The win for those adopting it is that it is potentially cheaper, but that raises the whole question of how you make actual money designing and building it when there are market based restrictions on what you can successfully charge for it once you have.

>Dennis

 

Page 1 / 4   >   >>

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed