Ever heard of GPUs that may be programmed on the fly? Vividsparks, an organization from Hubli, Karnataka, is engaged on making this potential. Dr. Vijay Holimath, President and CEO of VividSparks, talks about how their firm introduced a revolution within the computing world.
Q. VividSparks focuses on utilizing the POSIT quantity system in computing. What market or technical drawback do you search to resolve along with your product?
A. Allow us to put it in a simplified type. All of us have a smartphone, and everytime you watch a film, YouTube, or play a sport, your telephone’s battery goes down drastically. These days, you get quick cell phones with high-end and excessive efficiency, however that velocity comes with a price. Therefore, whenever you require velocity, you eat extra battery energy.
Therefore, these issues have been addressed in many alternative methods. Consequently, manufacturing semiconductor chips with a small measurement within the vary of 5 nanometers, seven nanometers, or 10 nanometers requires decrease fabrication and therefore lowers the expertise nodes, which gives greater velocity. Because it consumes a smaller chip, your telephone or pc turns into smaller. Persons are addressing the issue within the above-mentioned manner. However we tried to deal with it in a really basic manner. The elemental lie in how we compute math in pc persons are utilizing the floating level (FP) quantity system which was invented within the Eighties. This outdated FP structure is modified as per the necessities and expertise nodes, which ends up in the design of a brand new processor. On the similar time, we got here up with the concept of addressing this battery drain concern by altering the quantity system.
Computation has a broader perspective, together with writing a program on a pc, taking part in video games, and many others. There’s arithmetic behind it. Therefore, whenever you watch a film or when taking part in a sport, the graphics or lighting results you see on the telephone are resulting from arithmetic similar to trigonometric features, transformational features, divisional features, sq. roots, partial differential equations, likelihood idea, and many others. Persons are fixing this drawback with a floating level quantity system, invented within the Eighties. We got here throughout a novel quantity system referred to as the POSIT quantity system, which was invented in 2017. It took virtually 10 years of analysis and improvement efforts from 2012 to return throughout this novel structure.
We’re utilizing a POSIT quantity system to make the chip smaller and decrease the facility. Therefore, no matter you implement on the chip shall be cheaper. In different phrases, the carbon footprint is decrease; therefore, energy consumption is decrease. So, for those who run a sport, play a sport, or watch a film in your cell phone or pc, your battery received’t drain drastically. You would possibly now must cost your battery each 5 – 6 hours, regardless that it’s possible you’ll use high-end merchandise whereas utilizing our system, it’s possible you’ll not must cost your battery for 2 or three days utilizing our system. So, within the meantime, you’re additionally providing excessive efficiency. Therefore, the answer to all the issues is packed into one quantity system referred to as the POSIT quantity system.
Q. What sort of finish units do your chips produce?
A. Now we have quite a few finish units. It isn’t restricted to smartphones or cell phones but in addition consists of high-performance computer systems and workstations as nicely. We intently work with varied vital prospects throughout the globe. These computation models are usually not restricted to laptops or cell phones it’s also utilized in (IoT) the Web of Issues, Edge computing, Industrial Automation, and in addition in Self-drive help automobiles. All this stuff are pushed by our processors.
Q. How did you give you this concept? What’s the story behind it?
A. Initially, it was not about funding an organization or earning profits. It began once I (Vijay) was sitting at house, very uncommon. I used to be pursuing engineering, and after ending the sensible, I got here within the afternoon. I began taking a look at pc structure books and started turning the pages. I observed a floating level unit, which bought me . For instance, at any time when we do all this arithmetic utilizing pen and paper, it’s additional fed to the pc in a quite simple strategy to examine and provides the outcomes. Within the Maclaurin collection, Taylor collection, or partial differential or Gaussian theories, which you get to study in engineering diploma books or Bcom or BSC diploma books, you see all these mathematical equations.
“We’re all the time motivated in the direction of arithmetic, arithmetic runs in our blood.”
I used to be all the time fascinated by how this stuff we remedy utilizing paper simply by including integration zero to infinity and fixing them, however how the identical factor is solved by a pc. This bought me into insanity. I began studying quite a lot of books, after which I got here to know the technical time period “pc arithmetic.” In different phrases, you’re doing all this computation on a pc. Additionally it is referred to as pc arithmetic. Therefore, after ending my engineering, I did my MPhil and Ph.D. in the identical area, i.e., Pc Arithmetic. These are R&D levels, and I revealed some papers and located novel concepts and novel implementations. After that, I began engaged on it, and I labored on the identical area in Tokyo. I got here again to India in 2014. Then I suppose the POSIT quantity system modified the entire story.
Q. How did you make your prototype, the primary MVP, the proof of idea, first one you made? How did you do it?
A. We intently labored with Prof. John Gustafson. We’re the primary to show the POSIT number-based system, and that’s our working prototype.
“We launched the world’s first POSIT-based Common Goal Graphic Processing Unit (GPGPU).”
Many corporations have GPGPUs primarily based on totally different applied sciences, however they aren’t POSIT-based. Therefore, now we have bought the world’s first posit-based processor, GPGPU, that’s made in India, and this crown additionally goes to me. We even have the compiler on this POSIT-based quantity system as nicely. It’s additionally the world’s first compiler. So now we have all of the prototypes.
Q. What’s your organization’s primary USP for all its merchandise or applied sciences? One Distinctive Factor?
A. These days within the semiconductor market, the whole lot is licensed-based. In different phrases, if you wish to develop a processor and wish to promote it as a processor, normally individuals generate income after they license their IP cores and promote them to the businesses, after which they generate income whereas we make income on a subscription-based. For instance, sooner or later, you get a cell phone free of charge whereas paying solely $1 per 12 months or solely $2 per 12 months, after which the cell phone is yours. In that case, we generate income on a subscription-based mannequin, which is our USP. The merchandise now we have consumes half the reminiscence measurement in comparison with all of the merchandise available in the market. One other USP is that our merchandise are programmable, one can program each {hardware} and software program. In different phrases, if there’s a bottleneck in any of your functions, you possibly can change this system software program or you possibly can change the {hardware}, each the contouring, which is not like within the business. So these are all our market penetration or go-to-market methods, or you can too say USP.
Q. As you talked about, one can change the {hardware} and software program like FPGAs and rewrite the entire software program and alter the inner circuitry. Are you able to elaborate on that?
A. Essentially the most fascinating USP is, e.g., whenever you purchase a clean telephone and it doesn’t encompass something, it solely has a reminiscence and scratch card. Then the salesperson will ask you about your necessities, similar to high-speed, low-speed, or medium-speed. Once you inform the gross sales man your necessities, he’ll program your telephone accordingly and provides it to you. Then it’s a must to pay, let’s say, $1 or $2 per 12 months. Then, after one 12 months, you are feeling your telephone could be very sluggish. It’s type of quite a lot of reminiscence. Then you definately go to the salesperson and inform him to make the modifications as per your new necessities. He’ll then change the entire thing. He’ll provide you with one massive file to put in, after which it is going to be like a brand new telephone once more.
“These days processors in cell phones are usually not programmable and are static, however utilizing our expertise you possibly can change each {hardware} and software program.”
This will provide you with excessive efficiency, after which you possibly can run it for just a few extra years. So that is how the mannequin works. You’ll be able to have a brand new telephone yearly with out shopping for one. On common, a brand new telephone prices round $135 to $200 within the vary. Should you use our expertise, you don’t want to purchase a telephone each time. Solely updating it should give you all the brand new options. Therefore, there are quite a lot of benefits for the businesses and the shoppers, and the carbon footprint is decrease, therefore its helpful nature as nicely. Should you use one telephone for 10 years or 20 years, you don’t want to purchase a brand new telephone. So the concept behind that is to extend the lifecycle of the product.
Q. Are you able to inform us about your AI Multicore processor?
A. As AI/ML comes into quite a lot of issues, I’ll give one instance. In protection, there may be picture recognition that is available in. Let’s say the Navy has to detect objects beneath the ocean or detect missiles. They should detect it and should take precise motion. Therefore, environment friendly picture video processing is required, for which synthetic intelligence is available in. There are algorithms to make this detection, and quite a lot of mathematical computation is concerned. Therefore, the 2 most vital elements are inference and coaching. So it’s a must to prepare it after which infer it. On this case, the POSIT has a major position as a result of it has the next dynamic vary and a decrease knowledge width. In different phrases, with decrease knowledge width, you may get good outcomes. So, regardless of the expertise you see, these days, state-of-the-art, the dynamic vary is low. In different phrases, regardless of the enter or output you do to the circuit, it should both go underflow or overflow. The consequence has two extremes, that’s, for those who have a look at arithmetic, there are greater and decrease knowledge ranges, i.e., zero to 1, one to 2, and many others. If it goes on the best aspect, it’s larger; if it goes on the left aspect, it’s smaller. Therefore, the smaller the numbers go, it underflows; the larger the numbers go, it overflows. Within the present processor, it overflows, and they’re going to chop it off right here and chop it up there. Utilizing our expertise, there aren’t any overflows or underflows; it stays the identical. Therefore, greater precision means extremely correct detection and correction algorithms and good mathematical outcomes.
Q.What sort of analysis and improvement operations occur in India?
A. We did all of the analysis and improvement for chips in India, similar to RacEr, GPGPU, and FalKon. Thus, all of the R & D work was carried out in India, however our engineers are everywhere in the world, like in Madrid, Tokyo, Paris, and Washington. So, our HQ is in India, however our engineers and our primary crew are throughout the globe.
Q. What has been your commercialization story?
A. These days individuals don’t wish to pay, which is the laborious fact. Individuals don’t wish to pay for IP cores, particularly in semiconductors. Therefore, we determined to have subscription-based gross sales, to allow them to use it at a really minimal price. Thus, utilizing our product, they don’t want any infrastructure, lab, or {hardware} to run the board, and many others. They will apply it to the cloud. They will use it, let’s say, for $100 for a 12 months or $100 monthly. So, that is how we’re doing the commercialization. For our companions, we offer the IP cores freed from cost. Thus, they combine it, after which we cost the top prospects the identical subscription-based as I discussed, both month-to-month or yearly. Our chips can be found commercially everywhere in the world.
Q. What sorts of corporations are your prospects? What sort of individuals typically method you for subscriptions in India?
A. Corporations, institutes, and R&D facilities in AI/ML, HPC, GPU computing, and automotive computing. In India, we intently work with authorities institutes. Normally, overseas, we’re approached to speed up their functions. They wish to see how way more correct, how significantly better efficiency, or low reminiscence consumption they’ll get by utilizing our product. Thus, persons are very enthusiastic about utilizing this expertise.
Q. Are you presently searching for funding or funding in your firm?
A. For now, we don’t require funding, but when now we have a great alternative that’s mutually helpful, we are able to contemplate it. However I’m not so eager on the funding, as now we have good prospects offering us with a great revenue.
Q. What are your normal hiring developments?
A. Since our discipline is very technically specialised, among the college students are usually not so geared up with these sorts of applied sciences, i.e., semiconductors. College students are well-versed with VLSIs, VERILOG, and FPGAs, however in our case, it requires excessive mathematical expertise. Analysis professionals, perhaps we plan to offer internships.
Q. What are the plans for the enlargement of ViviSparks in India and overseas?
A. In India, I don’t require any branches as a result of solely HQ is sufficient, however we’re planning to broaden to different international locations. We’re primarily an export-oriented firm, so we don’t give attention to India alone. Now we have robust distribution throughout the globe.
Q. Are there any alternatives for different companies to associate with you to broaden your buyer base?
A. Ya, after all, we’re enthusiastic about partnering with any institutes or corporations. Whether it is for offering funding and is useful for them or mutually helpful, then we might like to be partnered.
Q. What do you wish to add or convey to our viewers?
A. I might add that our expertise is subscription-based which can present telephones or any digital gear at a less expensive price. Therefore, we consider expertise must be novel—not costly. And novel expertise needs to be accessible and reasonably priced for everybody.
