ASUS RS720A-E11/H100 Serveur GPU 90SKS000-M7BAN0

Modern AI and HPC workloads demand extreme parallel processing power that conventional servers simply cannot provide. The ASUS RS720A-E11/H100 tackles this challenge head-on, delivering a 2U rack-optimized platform engineered to house up to eight dual-slot NVIDIA H100 Tensor Core GPUs. With dual AMD EPYC 7003/7002 processors, 32 DIMM slots for up to 4TB of DDR4 memory, and 24 hot-swap drive bays, this server is built to accelerate deep learning, data analytics, and virtualization at scale.

Who is it for

This server is ideal for data scientists, AI researchers, and IT infrastructure teams deploying GPU-intensive applications such as large language model training, inference, scientific simulations, and high-performance computing clusters. It is also well-suited for enterprises seeking a dense, scalable platform for virtual desktop infrastructure (VDI) or hybrid cloud environments.

Main features

• Supports up to 8 dual-slot PCIe Gen4 x16 GPUs, including NVIDIA H100 and A100, for massive parallel compute.
• Dual-socket AMD SP3 design compatible with EPYC 7003 and 7002 series processors, offering up to 128 cores and 256 threads per node.
• 32 DIMM slots supporting up to 4TB of DDR4 3200MHz ECC RDIMM/LRDIMM for memory-intensive workloads.
• 24 front-access 2.5" hot-swap bays: 12 bays support NVMe/SATA/SAS, and 12 bays support SATA/SAS for flexible storage.
• Redundant 2200W 80 PLUS Platinum power supplies ensure high efficiency and continuous uptime.
• Comprehensive thermal design with 8 hot-swap 60x60mm fans and optimized airflow for dense GPU configurations.
• Integrated ASMB10-iKVM with IPMI 2.0 for out-of-band management, plus dual 10GbE LAN ports for high-speed networking.
• Expansion-ready with 8 PCIe Gen4 x16 slots for GPUs, 2 additional PCIe Gen4 x16 slots, and 1 OCP 3.0 Gen4 x16 slot.

Specifications

• Form Factor: 2U Rackmount
• CPU Sockets: 2 x Socket SP3
• Supported CPUs: AMD EPYC 7003/7002 Series Processors
• Chipset: AMD SP3
• Memory: 32 x DIMM slots, up to 4TB DDR4 3200MHz ECC RDIMM/LRDIMM
• GPU Support: Up to 8 x dual-slot PCIe Gen4 x16 GPUs (NVIDIA H100, A100)
• Storage Bays: 24 x 2.5" hot-swap (12x NVMe/SATA/SAS, 12x SATA/SAS)
• PCIe Expansion: 8 x PCIe Gen4 x16 (GPUs), 2 x PCIe Gen4 x16 (other cards)
• OCP Slot: 1 x OCP 3.0 Gen4 x16
• Network: 2 x 10GbE LAN (Intel X710-AT2 or equivalent)
• Management: ASMB10-iKVM (IPMI 2.0)
• Power Supply: 2 x 2200W 80 PLUS Platinum Redundant
• Cooling: 8 x hot-swap 60x60mm fans
• Front I/O: 2 x USB 3.2 Gen1
• Rear I/O: 2 x USB 3.2 Gen1, 1 x VGA, 1 x COM, 2 x 10GbE LAN, 1 x Management LAN
• OS Support: Windows Server, Linux (RHEL, SLES, Ubuntu), VMware ESXi

Tips

When populating this server with H100 GPUs, ensure your facility can provide adequate power and cooling—each GPU can draw up to 700W. Use NVMe drives in the front bays for high-speed data access to keep GPUs fed with data. For remote management, configure the dedicated management LAN port to isolate IPMI traffic from data traffic for security.