AIME GX8-H200 HGX Server

The AIME GX8-H200 is the HGX deep learning server based on the Gigabyte G593-ZD1-AAX3 barebone, it hosts 8 of the most advanced NVIDIA deep learning accelerators: the NVIDIA H200 SXM.

The AIME GX8-H200 is the ultimate deep learning accelerator and ai inference server. Packed into a dense form factor of only five height units the AIME GX8-H200 delivers:

• Dual AMD EPYC Genoa/Bergamon CPU performance with up to 2x 128 CPU cores with a total of 512 CPU threads

• 8x NVIDIA H200 SXM with 141 GB HBM3e memory each, with a total of 1.128 GB (1,1 TB) GPU memory

• 900 GB/s NVLINK switch connectivity between all accelerators

• up to 3072 GB (3TB) DDR5 memory, 12 channel memory with 24 equipped RAM slots

• up to 8x native NVMe SSD drives with up to 120 TB SSD storage

• 128 GB/s PCI 5.0 bus speeds between CPU memory and GPUs

• up to 8x 400 GBit/s OSFP network connectivity for cluster computing with a total network through put of 400 GB/s

Built to perform 24/7 for most reliable high performance cluster computing. Either at your inhouse data center, co-location or as a hosted solution.