ASUS showcases new NVIDIA-based AI/HPC infrastructure solutions at SC24

ASUS announced their latest infrastructure solutions at SC24, showcasing new server systems and cooling solutions. In collaboration with NVIDIA and Ubitus/Ubilink, they also established Taiwan's largest supercomputing facility, which notably offers green energy options for customers.

Their AI server lineup includes the ASUS AI POD, which incorporates the NVIDIA GB200 NVL72 platform. This system integrates GPUs, CPUs, and switches with high-speed direct communication capabilities, featuring the NVIDIA GB200 Grace Blackwell Superchip and fifth-generation NVIDIA NVLink technology.

It supports both liquid-to-air and liquid-to-liquid cooling options. The ESC8000A-E13P, an HPC server compatible with NVIDIA MGX modular architecture, supports up to eight NVIDIA H200 NVL dual-slot cards at 600 watts each. It includes five PCIe 5.0 slots for NVIDIA ConnectX and BlueField SuperNICs, along with BlueField-3 DPUs for networking, storage, and security processing.

Additional systems include the ESC N8-E11V with NVIDIA HGX H200 platform and ESC N8A-E13 with NVIDIA Blackwell platform.

The brand also developed comprehensive cooling solutions in partnership with global providers, achieving up to 95% heat dissipation through either liquid-to-air with sidecar or liquid-to-liquid with in-row coolant-distribution unit options. Their verification process includes thorough testing of power, network, GPU cards, voltage, and temperature components.

A small note on he newly constructed supercomputing center, it is built in collaboration with Ubilink and it has achieved 45.82 PFLOPS performance capacity and was completed in three months. This facility provides public cloud services, AI computing rentals, and subscription-based cloud services, with the unique option for customers to choose renewable energy packages.