In Data Centers, the workload requirements of generative artificial intelligence are forcing an architectural revolution. Compute resource requirements are growing and becoming increasingly heterogeneous, and memory and storage hierarchies are flattening to reduce latency. Socionext's custom Solution SoCs allow architects to integrate differentiating IP optimized for the target applications and meet the performance requirements cost-effectively and within stringent power constraints.
Socionext develops high-performance custom SoCs by utilizing its multi-core design capability, low-power AI engine, and accelerator to meet customers' exact needs for features and performance.
- Target Markets
- 5G Network
- Data Center
As the demand for cloud-based data centers grows, power-efficient devices are needed. Socionext develops power-efficient AI Accelerator SoCs to meet the growing demand for hyperscalers.
Watch Socionext AI Accelerator video:
Data Center Architecture Example
Example of Data Center SoC Development
- HPC Server CPU SoC
- Leading edge process node
- Multi-core CPU complex
- High-speed interfaces
- HBM3
PCIe Gen6
DDR5, multiple channels
CoWoS package
