SC20 Inspur Virtual Showcase

Record-Setting AI Server
#1 training and inference performance in MLPerf v0.7 benchmark, set 18 performance records and boasting 549,000 inferences per second in the ResNet50 inference task

Extreme AI Performance
Support 8 NVLink3.0 fully interconnected A100 GPUs, and 2 AMD Rome processors enabling PCIe4.0 in 4U, it delivers outstanding 5PetaFlops of AI performance

Accelerate AI Training
Besides A100 40G GPU, it supports the latest A100 80G GPU with HBM2e, which doubles the memory capacity of its predecessor and a 30% increase in GPU memory bandwidth for in single node, delivering 15% performance increase in training task for AI model with tens of billions of parameters

Empowering AI Innovation
Suitable for the most challenging AI applications, including intelligent customer service, financial analysis, smart cities, NLP and more

Powerful Computing Performance
Besides A100 40G GPU, it supports 8x NVLink3.0 fully interconnected A100 80G GPU with HBM2e memory in 4U, providing 300X computing power of a general-purpose dual-socket server, optimized for larger and more complex training scenarios.

Extreme Hardware design
The 4U 8GPU design helps to set up AI infrastructure and development environment efficiently, enabling lower deployment and operating costs with higher computing performance.

Comprehensive Ecological Support
Based on the NVIDIA CUDA, it supports the Delta system is seamlessly transplanted; based on the Intel platform development, it minimizes and avoids the cost of business migration, and meets the needs of AI users for extreme computing power and fast deployment.

Turbocharge AI Application
Ideal choice for a wide range of AI applications such as virtual assistant, financial analysis, smart cities, natural language processing and more.