Case Study: Texas Advanced Computing Center achieves efficient on-premises AI inference with SambaNova SambaStack

TACC runs 10x faster Horizon-scale AI inference with SambaNova Systems

The Texas Advanced Computing Center (TACC), a major research facility, required a high-performance AI solution to run within the strict power, cooling, and floor space constraints of its existing data center. Their challenge was to deploy a system capable of running the latest open-source coding models for researchers while maintaining energy efficiency and avoiding costly infrastructure retrofits. They turned to vendor SambaNova Systems for a solution.

SambaNova Systems implemented its SambaStack AI inference platform, an air-cooled system that eliminated the risk of coolant leaks and slotted into TACC's facility without retrofit. The solution's OpenAI-compatible API allowed for easy integration with existing tools, and its support for models like MiniMax M2.7 delivered the low-latency inference required for responsive coding agents. As a result, TACC successfully runs its diverse AI workloads, reduces operating costs, and stays within its physical and power constraints.

Texas Advanced Computing Center (TACC)