Ultra-low-power on-device chips for generative AI
A leader in AI technology, DEEPX is set to revolutionize on-device AI with its cutting-edge ultra-low-power AI chips.
With a significant presence at the Mobile World Congress (MWC) 2024, DEEPX will unveil a comprehensive blueprint for on-device AI chips and plans to expand partnerships with global collaborators. Currently, DEEPX is forging alliances with domestic and European telecommunications companies and global data center corporations, pushing for compatibility and optimization across network and cloud systems.
DEEPX’s journey also includes the development of groundbreaking technology that enables the federated operation of server-scale AI and on-device large AI models. The company plans to launch on-device AI chips capable of operating at just a few watts while delivering server-scale AI intelligence in personal devices by the second half of next year. Although Nvidia’s GPU-based offerings are currently the most cost-effective for LLM services such as ChatGPT and Gemini, the total power energy consumed by GPUs operating worldwide has reached levels exceeding the electrical energy of an entire country. Further, the high power consumption of GPUs leads to overheating and reliability issues. This collaborative operation technology between server-scale AI and on-device large AI models is expected to reduce energy consumption, carbon emissions, and costs by at least ten to a thousand times compared to relying solely on data centers.
DEEPX CEO, Lokwon Kim, stated, “DEEPX has gained international recognition for possessing global leading-edge core technology in power and cost efficiency for AI operation. Furthermore, we aim to develop new technology that enables artificial intelligence services at the level of super-scale AI with less than 5 W, providing the core technology that moves giant artificial intelligence technology from the realm of science to widespread use by humanity. We are committed to becoming a leading comprehensive AI chip company in the global market.”
With a focus on revolutionizing application areas such as smart cities, surveillance, smart factories, and other industries, DEEPX unveiled a line of low-power AI chips in 2023.
The suite of four high-performance, energy-efficient AI semiconductor products, designated DX-L1, DX-L2, DX-M1, and DX-H1 leverage Samsung Electronics’ foundry processes, harnessing the power of 5nm, 14nm, and 28nm technologies. Each product has been specifically engineered to cater to the unique demands of various market segments, from ultra-compact sensors with minimal data processing requirements to AI-intensive applications such as robotics, computer vision, autonomous vehicles, and many others.
DEEPX chips deliver high-performance AI capabilities while ensuring power efficiency, computational efficiency, AI accuracy, and support for the latest algorithms. This technological advantage has propelled DEEPX to the forefront of the AI semiconductor industry, surpassing the NPU (Neural Processing Unit) technology of global competitors. Furthermore, the chips offer superior cost competitiveness compared to GPUs and other market solutions, creating a significant edge in the NPU field.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
