ASUS Unveils AI Server XA P2N-E2 - Inven Global

The XA P2N-E2 is based on the NVIDIA MGX architecture and combines CPU computing with GPU acceleration, supporting up to two NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. The server targets agentic AI and reinforcement learning workloads, designed to maintain responsiveness during complex execution tasks through high single-thread performance and specialized memory technology.

Rack-Scale Solutions

ASUS also presented the rack-scale XA VR721-E3, which integrates 36 NVIDIA Vera processors and 72 NVIDIA Rubin GPUs on the NVIDIA Vera Rubin NVL72 platform. This fully liquid-cooled architecture is designed for large-scale AI training and inference, addressing the growing demand for high-performance computing infrastructure.

Hybrid Cooling Technologies

The company demonstrated servers with hybrid cooling methods, including the XR R1I12B-M1BHFS that applies Direct-to-Chip Liquid Cooling to GPU baseboard chips while air-cooling Intel Xeon 6 processors. This flexible design enables diverse server configurations to meet various workload requirements.

Complete AI Infrastructure

For the NVIDIA Vera Rubin platform, ASUS showcased four core technologies: the NVIDIA HGX Rubin NVL8, NVIDIA Spectrum-6 Ethernet switch, NVIDIA CMX for context memory storage, and NVIDIA Groq 3 LPX for accelerating token-generation inference. The company also introduced Intel and AMD-based platforms including the RS700-E12-RS4U and RS720-E12-RS12U with Intel Xeon 6 processors, and the AMD-based RS700A-E14B-R16U with 6th Gen AMD EPYC 9006 Series processors.