Key Features and Benefits

■ Featuring eight Intel® Gaudi® 2 AI accelerators that deliver leading training and inferencing performance

■ Various cluster architectures can be built using the same servers to scale the AI training and inference cluster

utilizing thousands of Intel Gaudi®2s

■ Dual 4th Gen Intel® Xeon® scalable processors

■ Expanded networking capacity with 24 x 100GE RoCE ports integrated into every Gaudi® 2

■ 21 RoCEv2 100GbE connectivity allows training across all 8 Intel Gaudi® 2 processors no external Ethernet

switch required

■ Maximum 24 x 100GbE, 3 ports per Intel Gaudi® 2 accelerator, 6 400GbE QSFP-DD per server scale out

bandwidth

■ Ease of system build or migration with Habana® SynapseAI® soſtware stack

■ Standardized architecture and Ethernet instead of proprietary Infiniband and NVLINK