Fujitsu Marries GPUs and Xeon in High-Density HPC Server

Print Friendly, PDF & Email

While Fujitsu has made headlines with their Sparc-powered K computer, they continue to develop more conventional x86 servers for technical computing. They’re doing hybrid computing as well, as evidenced by today’s announcement that Nvidia Tesla GPUs are now available in Fujitsu’s Primergy CX400 S1 servers.

The Fujitsu PRIMERGY Multi-Node CX400 demonstrates Fujitsu’s philosophy of field innovation and a constant drive to increase the power and flexibility of our scale-out server systems for both HPC and cloud computing environments,” said Jens-Peter Seick, senior vice president of Product Development Group at Fujitsu Technology Solutions. “With support for NVIDIA GPUs, the system dramatically increases computational performance, while reducing costs and power consumption. This combination is exactly what today’s HPC customers are looking for.”

As a high-density platform, the new Primergy servers use half-wide, two-socket server nodes—doubling the number of cores available per unit height. The system houses either four 1U hot-plug PRIMERGY CX250 S1 server nodes with two Intel Xeon CPUs per node, or two 2U hot-plug CX270 S1 server nodes with two CPUs and one NVIDIA Tesla GPU per node.

Read the Full Story.