With almost 30 years of experience in the server industry, ASUS has revolutionized Generative AI with its solutions based on NVIDIA Blackwell, designed for advanced applications, marking a new era of performance for data centers. One of the featured products is the ASUS ESC AI POD, an innovative rack solution equipped with the NVIDIA GB200 NVL72 system, which integrates 36 NVIDIA Grace CPUs and 72 NVIDIA Blackwell GPUs. All this in a single rack-scale NVIDIA NVLink domain, functioning as a massive GPU.
The ESC AI POD offers combined liquid-air and liquid-liquid cooling options, optimized for maximum energy efficiency, from single racks to large data centers. This solution is designed to accelerate inference of large language models (LLM), providing real-time performance for applications that require enormous computational resources, such as language models with trillions of parameters.
Another important release is the ASUS ESC NM2-E1, a 2U server based on the NVIDIA GB200 NVL2 platform. It is designed for generative AI and high-performance computing (HPC). This device offers high-bandwidth communication, optimized for the entire NVIDIA software stack, providing a powerful platform for developers and researchers looking for new ways to drive innovation in AI.
AI innovations powered by NVIDIA Blackwell: ASUS has presented the ESC NM2-E1 with the platform NVIDIA GB200 NVL2 and the ESC AI POD with NVIDIA GB200 NVL72designed for advanced generative AI applications
Improving AI performance with NVIDIA MGX architecture and NVIDIA H200 Tensor Core GPUs
The company also introduced its ASUS 8000A-E13P server, a 4U model that supports up to eight NVIDIA H200 Tensor Core GPUs. Fully supported by the NVIDIA MGX architecture, this server is designed for rapid deployments into large-scale AI infrastructures.
With a configuration that includes NVIDIA 2-8-5 (CPU-GPU-DPU/NIC) topology and four high-bandwidth ConnectX-7 or BlueField-3 SuperNICs, it significantly improves data traffic between nodes (steam traffic). west) as well as overall system performance.
Featured Session at OCP Global Summit 2024: Modular Architecture for Data Centers
The ASUS special session at the OCP Global Summit 2024, held on October 16, had a notable impact. During this talk, ASUS presented the Data Center Modular Hardware System (DC-MHS) developed by OCP, which uses the NVIDIA MGX modular architecture.
In addition, it was explored how this architecture is revolutionizing server design, providing flexibility, simplified maintenance and efficient scalability, key characteristics for modern data centers.
