An automotive-grade NPU chip has revolutionised AI applications with unparalleled energy efficiency and security, further reshaping the edge AI landscape by supporting offline GPT models.
Kneron, the San Diego-based full-stack AI company known for pioneering neural processing units (NPUs), have launched the KL730. This automotive-grade NPU chip integrates an Image Signal Processor (ISP). This chip offers secure, energy-efficient AI functionality for diverse applications, from enterprise-edge servers to smart appliances and advanced driving assistance systems. The NPU chip, crafted with Artificial Intelligence (AI) at its core, showcases adaptable, power-savvy, and secure technological advancements. With a versatile interface, it smoothly links various digital signals, such as images, videos, audio, and millimetre waves, fostering the creation of varied AI applications across sectors. The chip is low-cost and achieves a three- to fourfold increase in energy efficiency compared to earlier Kneron iterations. It is 150- to 200-percent more energy-conservative than its major industry counterparts.
The KL730 offers an energy efficiency improvement of 3 to 4 times compared to earlier Kneron models, and it outperforms major industry counterparts by being 150 to 200 percent more energy-efficient. The edge AI chip backs all Generative Pre-trained Transformer (GPT) models by supporting transformer neural networks. The NPU chip augments this lineup, providing chips with base compute capacities ranging from 0.35 to 4 effective tera operations per second. This enhancement broadens its capability to handle advanced lightweight GPT large language models, such as the nanoGPT. The KL730 stands distinctively poised to revolutionise security within the AIoT domain, allowing users the flexibility to operate GPT models either entirely or partially offline.
“Running AI requires AI-dedicated chips with an architecture completely different from anything we’ve seen before. A simple re-appropriation of adjacent technologies, such as graphics-dedicated GPU chips, doesn’t do the job,” says Albert Liu, Founder and CEO of Kneron. “The KL730 is a game-changer for edge AI. With its unprecedented efficiency and support for transformer neural networks, we empower users across industries to unlock the full potential of AI without compromising on data privacy and security.”
For more information, click here.