Puget Systems has introduced a new custom generative AI and machine learning server with powerful Nvidia GPUs and AMD Epyc processors.
Puget Systems, which specializes in high-performance custom-built computers, earlier this month released its newest custom generative AI and machine learning server.
The LLM AI Training and Inference server, for enterprises or data centers, are rack-mount workstations designed for GPU-intensive generative AI workflows. They support up to eight Nvidia RTX Ada-generation GPUs for a total of 752GB of VRAM. Optionally, they can be configured with Nvidia L40S GPUs or Nvidia H100 NVL tensor-core GPUs, the latter targeting LLM inference, notes Puget, due to its high compute density, memory bandwidth, and energy efficiency, as well as its NVLink architecture. It also delivers extraordinary acceleration to power high-performing elastic data centers for AI, data analytics, and HPC applications, Puget adds.
The AI Training and Inference Server also come with AMD’s Epyc line of processors offering up to 128 cores with support for 1.5TB DDR5 ECC RAM and 128 PCIe Gen 5 lanes. The server uses a pair of those CPUs to enable up to eight dual-width GPUs in a 4U rack-mount chassis.
The server can be configured for a wide range of generative AI applications. Pricing is based on configuration.