AI Solutions

The only constant in the world of AI is change. Higher processing requirements and new algorithms are introduced regularly. In this environment a solution must have high flexibility and programmability, since everything evolves rapidly, making any hard-coded approach non-viable from the start. The processing power requirements keep increasing as new AI inference and real-time processing requisites are announced. Edge AI needs low latency as well as lower energy consumption and has power constraints. Ultimately it comes down to cost = silicon area. VSORA offers the most flexible solution on the market, by giving the user the ability to select the best combination of performance, power and cost.

Click here to obtain all standard benchmark results.


  • Fully programmable solution
  • Algorithm agnostic (CNN, RNN,…)
  • High-level framework support (TensorFlow, Caffe2, PyTorch, ONNX)
  • Graph support
  • Scalable number of MACs – 256 to 65,536 in a single core
  • Unlimited number of cores in parallel
  • Very high memory bandwidth to load high number of MACs
  • Data memory bandwidth follows processing power = no bottleneck
  • No hardware accelerators
  • No performance impact from compute precision
  • Solution works from IoT to ADAS and beyond


VSORA ADAS Sensor Solutions

VSORA has developed a unique, algorithm-agnostic architecture that allows you to optimize power, performance and silicon size.

For AI applications a single core is scalable from 256 to 65,536 MACs and the user has the ability to design a system with multiple cores, providing the user a wide range of choices. Even used as a regular signal processor VSORA provides a very powerful, flexible and scalable solution. A single core is capable of handling in excess of 1TMAC/second. In both cases there is no limitation on the number of parallel cores that can be used.

An innovative combination of software and hardware can reconfigure a system in a single clock cycle, to accelerate all current and future algorithms without hardware changes.

For the specific example provided by OSRAM above, two cores using 4,096 MACs each would provide 25 TOPS computing power. There is plenty of room to scale this up, as a single core of 65,536 MACs can provide up to 290 TOPS, and the ability to run multiple cores in parallel makes the selection process very easy and extremely flexible.


Artificial Intelligence Relative Performance


Artificial Intelligence Conditions