Server, storage and network solutions vendor Supermicro has released details on an “innovative” new multi-node GPU server solution. Supermicro claims its multi-node system to deliver up to 10% TCO savings utilizing shared power and cooling. The new 2U 2-Node NVIDIA A100 based GPU server system is designed with up to 64 cores and 128 PCIe 4.0 lanes with three double-width PCIe 4.0 GPUs or six single-width PCIe GPUs at full speed per node.
“Our new 2U 2-node multi-GPU server is the perfect platform for video streaming, high-end cloud gaming, and countless social networking applications,” said Charles Liang, President and Chief Executive Officer (CEO) of Supermicro. “With our advanced server Building Block Solutions® design, and resource-saving architecture, customer deployments will be the most energy-efficient systems available. Supermicro’s market-leading system flexibility and cost savings will deliver uninterrupted performance. We have already experienced overwhelming market interest for this unique platform from global customers.”
Powered by AMD EPYC 7002 series and “next-generation” processors with 64 cores and 128 PCIe 4.0 lanes per node, the thermally efficient, streamlined airflow would allow sustained top performance of this dense multi-GPU system.
The solution can be ideal for multi-instance high-end cloud gaming and many other compute-intensive data center applications. Equipped with Supermicro’s advanced I/O Module (AIOM) for fast and flexible networking capabilities, the system can also process massive dataflow for demanding AI/ML applications, deep learning training, and inferencing.
The unique multi-GPU node design would allow great serviceability, according to Supermicro. The two-node drawers in the 2U system can be pulled out for easy serviceability thanks to the shared power and cooling resources. This unique design and the accessibility to the GPUs may lower the cost of maintenance and upgrades for demanding GPU accelerated applications. One example of such applications includes cloud gaming which typically requires sustaining high-power usage and frequent maintenance.