One Platform for Complete AI Acceleration
Get the best of general-purpose compute, flexible vector acceleration, and powerful AI-specific
performance—on one integrated platform.
CPU + Vector + Tensor
Why It Matters
Integrated, Efficient, Low Power, and
Low Latency Computing
No need for separate accelerators
Everything is tightly integrated and easy to use based on a single ISA
More efficient
Lower power, lower latency, and no data-copying between blocks
Highest efficiency
Lowest power, lowest latency, and no data-copying between compute elements
Scales with your needs
From edge AI to heavy-duty inferencing and training
What You Get
Powerful CPU, Flexible Vector Unit, and
Scalable Tensor Unit for AI
Powerful CPU core (Atrevido)
Flexible vector unit (V4, V8, V16, V32)
Scalable tensor unit (T1, T2, T4, T8) for AI workloads
Runs Yolo, Llama and LLMs with ease
Simple software—no DMA, no kernel changes
Great For
Scalable AI Inference and
Advanced Applications.
AI inference
(vision, language, recommendation)
