Chimera™ GPNPU IP now available to power efficient AI/ML inference for the next generation of silicon startups

Silicon Valley, CA, June 28, 2023 - Silicon Catalyst, an incubator exclusively focused on accelerating semiconductor solutions, announced that Quadric® has joined as a new member of its In-Kind Partner (IKP) ecosystem. The Silicon Catalyst IKPs provide technical and business support to the Portfolio Companies in the incubator, enabling them to tap into the products and services available to enhance the growth of their companies.

Quadric’s Chimera family of GPNPUs is a semiconductor intellectual property (IP) offering that blends the ML performance characteristics of a neural processing accelerator with the full C++ programmability of a modern DSP. Chimera GPNPUs provide one unified architecture for ML inference plus pre-and-post processing, greatly simplifying both system-on-chip (SoC) hardware design by the semiconductor developer today and subsequent software programming months and years later by application developers.

Chimera GPNPUs are now added to the roster of silicon IP building blocks available to Silicon Catalyst member companies, enabling those innovative startups to focus on value-add algorithm solutions that deliver differentiated value.

“Silicon Catalyst is instrumental in bootstrapping new and innovative silicon companies” said Veerbhan Kheterpal, CEO and co-founder of Quadric. “Quadric itself was a part of the Silicon Catalyst program back in 2018, and now we have a chance to pay that back to the next cohort of incubating companies.”

High Efficiency + Flexibility: First to Support Vision Transformers

The Chimera GPNPU architecture excels at convolution layers, the heart of convolutional neural networks (CNNs). Quadric’s Chimera cores deliver ML inference performance similar to the efficiency of dedicated CNN offload engines but with full programmability. Unlike conventional accelerators that can only run a handful of predetermined ML operators, Chimera GPNPUs can run any ML operator. Custom operators can be added by the software developer simply by writing a C++ kernel utilizing the Chimera CCL application programming interface then compiling that kernel using the Chimera SDK.

Quadric’s support of the latest innovation in machine learning – the Transformer class of models – is evidence of the architecture’s flexibility. The invention of the Vision Transformer (ViT) class of models has rendered many traditional edge and device NPU accelerators obsolete because they cannot support ViT class models. But Quadric’s Chimera GPNPU does support performant implementations of ViT graphs. Interested users can see and evaluate ViT on the Chimera platform in the Quadric DevStudio, found at studio.quadric.io.

"Silicon Catalyst has built an ecosystem to support semiconductor hardware companies that includes over 60 In-Kind Partners, and we are thrilled to have Quadric become our first Portfolio-Company-turned-IKP," said Nick Kepler, COO of Silicon Catalyst, "As they build a successful market presence, it is gratifying to see Quadric evolve from building their first chip with the help of our In-Kind Partners to providing that same support to our current and future Portfolio Companies.”

Quadric at DAC60 in San Francisco

Quadric’s Chief Marketing Officer, Steve Roddy, will be participating in a panel session, “AI on the Edge! Challenges and solutions for Edge AI products ”, Monday July 10, 2023, 3:30 to 5pm, Room 2008, level 2. The session details can be found here in the areas of communications, energy, IoT and embedded systems.”