Versatile Integration
Provides exceptional flexibility for seamless integration with various NPUs, including Cadence’s Neo NPUs, in-house developed NPUs or third-party NPU IP, making it an ideal choice for developers and enhancing compatibility with existing AI infrastructure
Optimized AI Workload Performance
Built on Xtensa architecture, uses VLIW and SIMD to efficiently handle complex AI tasks. It has optimized instructions for operations such as sigmoid, tanh, and relu, excelling in executing layers not suited for NPUs
Highly Customizable
Offers extensive configuration options to optimize area and performance, surpassing traditional CPU and GPU solutions. This customization enables developers to create efficient, high-performing AI subsystems, driving innovation for evolving AI workloads
Unified Software Development
Cadence NeuroWeave Software Development Kit (SDK) is a single SDK used across all of Cadence’s AI IP. Leveraging the TVM stack, the NeuroWeave SDK is easy to use and allows architects to tune, optimize, and deploy their AI models on Cadence’s AI IP
Offerings
Tensilica NeuroEdge 130 AI Co-Processor
|
NeuroEdge 130 |
|
---|---|---|
MAX SIMD Width |
512 |
|
AI Enhancements |
Neural Network Quantization Acceleration Non-Linear Operator Acceleration |
|
Connecting to NPU |
AXI New! High Bandwidth Direct Interface (HBDI) |
|
MAC for AI |
8x8 |
512 |
8x16 |
128 |
|
16x16 |
128 |
|
FP16 |
64 |
|
Software |
NeuroWeave Software Development Kit (SDK) AI-LIB |