SUPERCHIP (Scalable Unified Processor Enhancing Revolutionary Computing) is our answer to the surging demand for real-time supercomputing across edge AI, autonomous driving, generative AI, and decentralized AIoT applications. Funded under the HORIZON-EIC-2023-ACCELERATOR-01 call, this project tackles the exponential growth of sensor data and AI workloads that today’s semiconductor providers struggle to serve.
Modern automation—especially Level 4/5 autonomous driving—relies on ingesting vast streams of sensor inputs and running complex algorithms to maintain a dynamic, 360° perception model in under 20 ms. Yet no commercially available solution can efficiently combine AI inference and digital signal processing (DSP) at that speed and scale. Current multi-chip assemblies introduce high latency, wasted silicon utilization (often under 20 percent), and excessive power draw, undermining both performance and sustainability. To achieve true autonomy and maximum safety, the industry must deliver enormous compute throughput, ultra-low latency, minimal energy consumption, seamless AI-DSP integration, deterministic execution, in-field programmability, and cost-effectiveness in a single platform.
Our companion chip, TYR, revolutionizes this space by adding “extra brains” to conventional processors. The TYR family (TYR1, TYR2, TYR4) plugs into a central compute unit and operates semi-autonomously, asynchronously processing data alongside the main CPU. Each chip delivers between 400 and 1,600 TFlops of raw performance with 70–80 percent implementation efficiency, all with very low power consumption and with a silicon footprint small enough to keep costs competitive. Fully programmable in high-level languages and agnostic to host algorithms, Tyr transforms how AI and DSP tasks coexist—unlocking real-time, context-aware processing that makes Level 4/5 autonomy, green transitions, and next-generation AI applications commercially viable. SUPERCHIP is our answer to the surging demand for real-time supercomputing across edge AI, autonomous driving, generative AI, and decentralized AIoT applications. Funded under the HORIZON-EIC-2023-ACCELERATOR-01 call, this project tackles the exponential growth of sensor data and AI workloads that today’s semiconductor providers struggle to serve.
Co-funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union. Neither the European Union nor the granting authority can be held responsible for them.
HQ
13 rue Jeanne Braconnier
Immeuble Le Pasteur
92360 Meudon-La-Forêt
France
Asia
Taipei
Taiwan
Japan
Tokyo
Japan
Korea
Seoul
Korea
USA
San Diego, CA
USA
Unmatched Performance at the Edge with Edge AI.
Fully programmable
Algorithm agnostic
Host processor agnostic
RISC-V core to offload & run AI completely on-chip
Tyr 4
fp8: 1600 Tflops
fp16: 400 Tflops
Tyr 2
fp8: 800 Tflops
fp16: 200 Tflops
Tyr 4
fp8/int8: 50 Tflops
fp16/int16: 25 Tflops
fp32/int32: 12 Tflops
Tyr 2
fp8/int8: 25 Tflops
fp16/int16: 12 Tflops
fp32/int32: 6 Tflops
Close to theory efficiency
Fully programmable
Algorithm agnostic
Host processor agnostic
RISC-V cores to offload host
& run AI completely on-chip.
fp8: 3200 Tflops
fp16: 800 Tflops
fp8/int8: 100 Tflops
fp16/int16: 50 Tflops
fp32/int32: 25 Tflops
Close to theory efficiency