Quadric partners with Silicon Catalyst, bringing edge AI chip tech to wider audience
Quadric, a provider of edge AI chips, has joined Silicon Catalyst, an incubator focused on semiconductor solutions, as a new member of their partner ecosystem.
Members of the Silicon Catalyst In-Kind Partner (IKP) ecosystem provide technical and business support to portfolio companies in the incubator, enabling them to tap into the products and services available to enhance the growth of their companies. Quadric’s intellectual property building blocks will be available to the 97 companies currently in the incubator’s fold.
Quadric recently revealed that their Chimera general-purpose processor IP will be compatible with the vision transformer ML model. Quadric says that its edge computing System-on-Chip (SoC) will allow customers to effectively leverage the capabilities of advanced large language models for computer vision applications.
Quadric’s chip design, the Chimera core processor, integrates integer multiple-accumulate (MAC) hardware alongside a 32-bit arithmetic and logic unit (ALU) functionality. According to the company, this architectural design plays a crucial role in facilitating the efficient integration of the vision transformer model.
The vision transformer combines MAC-intensive operations such as convolutions and dense layers with DSP/CPU-centric code like normalization and softmax. Quadric believes that its architecture efficiently supports this combination.
“Back in 2018, when we conceived the Chimera GPNPU architecture, we knew the rapidly evolving nature of machine learning meant we had to build processor IP that was both matrix-performance optimized and general purpose,” says Veerbhan Kheterpal, CEO at Quadric.
The artificial intelligence and machine learning landscape for edge inference has evolved since the introduction of large language models. The subsequent innovation through the vision transformer has significantly improved performance in various image and vision processing tasks.
Vision transformers have a completely different approach to solving vision computation. But Quadric believes that the Chimera general purpose processor IP addresses the limitations of the current ecosystem of edge computing SoCs.
“We knew something would be invented that pushed the ResNet family of networks into the history bin, but we did not know that transformers would be the current winner. And we do not know what state of the art will look like in 2027, but we do know that Chimera GPNPU licensees will be ready to tackle that next challenge,” Kheterpal continues.
Quadric joined forces with Ams Osram earlier this year to create an intelligent sensing solution, using its advanced Chimera neural processors. This collaborative solution possesses a low-power smart sensing module capable of capturing images. It excels at conducting machine learning inference tasks.
Article Topics
AI/ML | chip | edge AI | inference | Quadric | Silicon Catalyst | vision processing | vision transformer
Comments