Manjeera Digital Systems Builds First Indian Processor to Enable AI Processing2 months ago Saritha Keshamoni
Made-in-Hyderabad High-Performance Computing Engine Successfully Gets Validated on Silicon
With this announcement, the CIE-incubated Hyderabad-based startup Manjeera Digital Systems is all set to take on the global giants in the microprocessor industry.
As we hurtle towards a future dominated by AI and Computer Vision, it is becoming increasingly clear that we are going to have to rely ever more on sophisticated hardware that can support and scale up to deliver usable performance needed for small devices to big servers. Traditional processors do not have the abilities to process such big data typically needed for neural network processing. Therefore it is not just traditional chip makers but also tech giants that are after high-performance computing.
High Performance Computing Engine
“Unlike other players in the field who have been tweaking and making incremental changes or modifications to the processor while keeping the fundamental architecture at the core the same, we, at Manjeera took a step back and created an Indian processor with a unique approach to computing. We call this Universal Multifunction Accelerator or UMA,“ says Dr. Venu Kandadai, Co-Founder and CEO of Manjeera Digital Systems.
The team tested their high-performance computing engine against existing products in the market for vital parameters such as Performance, Power Consumption, Programmability and Silicon Area. UMA technology was the only one in the market that achieved the desirable results of High (Performance), Low (Power Consumption), High (Programmability) and Low (Silicon Area). They also conducted a competitive performance analysis on standard benchmarks for relevant application areas such as deep learning, image processing, computer vision, and DSP where UMA again outperformed other leading competitive products.
Applications of UMA
The Universal Multifunction Accelerator (UMA) series of products has an exhaustive range of applications from multimedia (audio, video, image) processing, to data security (encryption, decryption, hashing), to graphics, algo-trading, general computing and so on that are used across industries such as health care, robotics, defence and space, automative, surveillance, smart phones and drones. UMA architecture facilitates run time configurability and hence can be used to accelerate any future algorithms. According to Dr. Venu, “The unique feature of UMA is that it can be used even in applications where it is not possible to have any hardware accelerators. In fact, no new tools or infrastructure are needed to use UMA”.
Validation on silicon
The team has been working on tape-out of a chip with UMA over the last 6 months and achieved success in this effort. With the successful validation on silicon, the IP is ready to be integrated into an SoC. “In addition to silicon validation of UMA IP, this achievement also shows our team’s ability to design and tape-out an involved SoC successfully in the very first attempt,” says Dr. Venu.
A study by Assocham and consultancy firm EY in 2016 had reported that “Growing reliance on imports for electronic components and rapidly increasing demand for electronic products is making it indispensable to grow and strengthen India’s electronics manufacturing capabilities.” With electronic hardware and components such as integrated circuits, printed circuit boards (PCBs) and other active components, constituting the third largest drain on the exchequer (the first two being fuel, oils and gems and other precious stones) in terms of imports, Manjeera Digital Systems’ announcement showcasing our country’s native capabilities in the electronic components market is to be lauded.
About Manjeera Digital Systems
Christened after the “lifeline of the city” (Hyderabad), the Manjeera river, the 5-year-old startup, which received it’s initial seed fund investment from CIE in 2013, and is largely angel-funded, describes itself as a DSP R&D house designing next generation computing architectures for high performance computing. Their aim is to address the existing bottlenecks present in processor/co-processor/accelerator architectures. Their patent pending Middle Stratum Operations (MSO)-based architecture leads to a disruptive technology bringing a fundamentally different approach to computing with very high performance at low power consumption.
The core team consists of four vastly experienced professionals in semiconductor industry with strong research and entrepreneurial background: Dr Venu Kandadai, professor-turned-entrepreneur and Inventor of UMA technology, Dr. Govindacharyulu PA, their Silicon Implementation Advisor, Suresh Cheppalli, the CTO, and Srini Chandupatla, in charge of Operations, Strategy, Business Development. With a current employee strength of 24, who have expertise in VLSI design, SoC architecture, FPGAs, Computer vision and machine learning algorithms and embedded programming, the company has already won awards in the Innovation and IP space: Winner of India Innovation Growth Programme Award – 2013 and the Winner of Best Paper Award in Systems and IP at SNUG India 2015.
The company currently caters to ISRO, Intel, Astra Microwave Products and OmniPhy, creating IP for ASIC, or bit files. The base-band SoC chip is also going into the development of different types of NavIC receivers for ISRO. “The market for high-performance computing devices is set to be worth $44.98 bn by 2022, and as far as our break-through technology is concerned, the sky is the limit for expanding our customer base”, says Srini Chandupatla, Co-founder and COO.