SiFive Intelligence XM Series

High-performance AI computing engine that integrates multiple computing units and offers high memory bandwidth.

CommonProductOthersAI ComputingMemory Bandwidth

The SiFive Intelligence XM Series is a high-performance AI computing engine launched by SiFive, which delivers exceptional performance-to-power efficiency for compute-intensive applications through the integration of scalar, vector, and matrix engines. This series continues SiFive's tradition of providing efficient memory bandwidth while leveraging the open-source SiFive Kernel Library to accelerate development time.

SiFive Matrix Engine: Designed with Fat Outer Product
tightly integrated with four X-Cores for deep fusion of vector units.
Four X-Cores per cluster: Each equipped with dual vector units to execute all other layers
such as activation functions
with added exponential acceleration instructions.
New matrix instructions: Retrieved by scalar units with source data coming from vector registers
targeting each matrix accumulator.
1 cluster = 16 TOPS (INT8)
8 TFLOPS (BF16) per GHz.
Each XM series cluster provides a sustained bandwidth of 1TB/s.

\Target audience includes enterprises and research institutions that require AI workload processing
data flow management
object detection
voice processing
and recommendation handling. The high performance and scalability of the SiFive Intelligence XM Series make it an ideal choice for these applications.\

For processing large-scale AI model training and inference tasks.
For efficient data flow management and analysis in data centers.
For real-time object detection and voice processing on edge computing devices.

1. Visit the SiFive official website and download the SiFive Kernel Library.
2. Select the appropriate XM Series configuration based on project requirements.
3. Integrate the XM Series into the existing hardware or software architecture.
4. Utilize open-source libraries to accelerate the development process.
5. Conduct system testing to ensure performance meets expectations.

Visit