A leading European search engine provider needs strong AI computing power, suitable for a variety of large-scale training, including search engine, E-commence and map. Low latency and high bandwidth communication between CPU and GPU as well as between GPU and GPU, and high computational efficiency are highly demanded.
Inspur provides the company with AI server NF5488A5. NF5488A5 delivers up to 5petaFLOPS per node; its CPUs communicate with GPUs via 64Gb /s PCIe GEN4, and GPUs communicate with GPUs via 600Gb /s NVLink3.0, which can ensure efficient data communication for various AI workloads. It has 8 NVIDIA SXM4 A100 GPUs equipped in 4U form factor, simplified upgrade and maintenance, optimized heat dissipation design, and can stand 35℃ environment temperature.
Inspur NF5488A5 helps the company dramatically speed up model training and development.