8U 16GPU Industry-Leading Supercomputing Server for AI


Inspur AGX-5 is equipped with 16 of the most powerful NVIDIA Tesla® V100 Tensor Core GPUs with 32GB and will support next-generation GPU accelerators, providing incredibly powerful stand-alone AI computing performance with 2 Petaflops. AGX-5 brings revolutionary advances in AI computing power, and will surely provide an unprecedented AI innovation acceleration experience for commercial companies and research and innovation institutions dedicated to leading AI technology.

Key Features

Extreme Performance

Provides incredibly powerful stand-alone AI computing performance with 2 Petaflops Compared to previous GPU servers of the same type, the AGX-5 supports a 4x ultra-large-scale deep neural network model and a 10x faster training speed.

Extreme Connectivity

Based on NVIDIA’s latest HGX-2 platform, AGX-5 uses the industry’s most advanced NVIDIA NVSwitch™ interconnect fabric, enabling 48-channel, 2.4TB/s full-chip cluster high-speed interconnects

Extreme Throughout

512GB HBM2 globally shared ultra-fast graphics cache, and provides nearly linear AI computing performance scaling

Flexible GPU topology

Free to choose different topologies for various applications

Model AGX-5
Form Factor 8U
Dimensions 448mm x 351.6mm x 850mm

Server: 2*Intel® Xeon® Scalable Processors, 3*UPI
JBOG: 2×8*GPU or 16*GPU

GPU NVIDIA® Tesla® Volta, Volta Next (SXM3)
Memory 24× DDR4 DIMM, 6 channel
Storage 8 x 2.5” SATA storage (Includes 4*2.5” NVMe), 2*M.2
Performance 2 PFlops
Network Controller Integrated LAN controller; up to 4*10GbE
Power Supply (2+2)*2 Redundant, up to 12 KW total
Fan & Cooling Redundant Hot swap System Fan, Air cooling
Air Cooling or Air / Liquid Hybrid Cooling

Interested in Learning More?