AMD has introduced its new Helion rack solution, designed to provide flexible infrastructure for AI agent workloads. Unlike previous rack designs that relied primarily on GPU power, this system allows for a CPU-centric configuration, offering significant computational density for data centers.
The maximum configuration of the Helion rack supports up to 128 Epyc 9006 Venice processors based on the Zen 6 and Zen 6C architecture. When equipped with 256-core models, the system delivers a total of 32 768 processor cores, providing high-performance computing capabilities for demanding tasks.
For hybrid needs, the rack can also be configured with GPU accelerators. It supports up to 16 Instinct MI350P units, which provide 74 PFLOPS of performance in FP4 mode. AMD also noted the existence of Epyc 9006 LP Verano processors, which feature up to 192 cores in their top-tier configuration.
The Helion platform is based on the open AMD Helios design. It allows clients to balance CPU performance, core density, and power consumption alongside GPU acceleration to optimize the infrastructure specifically for the requirements of their artificial intelligence agents.
