SiMa.ai MLSoC Modalix
Modalix is SiMa.ai's second-generation MLSoC (Machine Learning System-on-Chip), designed for multimodal and generative edge AI (physical AI). It combines an ML accelerator, eight Arm Cortex-A65 cores and a Synopsys EV74 DSP on one piece of silicon, running a full application pipeline (from sensor capture to text-to-speech output) on a single chip.
The 50 TOPS (INT8) variant draws roughly 8-10W, supports BF16, INT8 and INT16 precision and CNN, ViT, LLM and LMM models; the vendor cites over 10 tokens/s on Llama2-7B. The family spans 25, 50, 100 and 200 TOPS tiers (the 25 TOPS part is the same silicon as the 50 TOPS with limited performance). The 50 TOPS device is built on TSMC N6.
Modalix ships as a System-on-Module (pin-compatible with NVIDIA Jetson Orin NX and Nano), an HHHL PCIe card, a chip-down BGA package and a DevKit. It is programmed through the Palette software suite and the Palette Neat agentic environment. Announced in September 2024 with sampling from January 2025; used in robotics, drones, automotive, industrial systems and vision.
Classification
AI System-on-Chip (SoC) ยท serves as: AI Inference, AI acceleration, Computer vision processing, Embedded main processor. ยท compatible with: Humanoid, Mobile Robot, Quadruped, Industrial Robot.
Which group SiMa.ai MLSoC Modalix belongs to and how it is built
Covers integrated System-on-Chip circuits used as central processing units in embedded systems, robotics, AIoT, and edge AI devices. In this schema, parentCategory is set to 'other' because the parentCategory enum does not include a compute category.
Component type encompassing System-on-Chip designs used as primary compute units in embedded devices, edge AI, and robotics applications.
A single semiconductor component integrating CPU, GPU, AI accelerators, memory controllers, multimedia engines, and communication interfaces.
Technical specification
How much power SiMa.ai MLSoC Modalix needs and how it can be powered