AMD EPYC 9006 LP targets rack-scale AI servers

AMD EPYC 9006 LP, code-named Venice-LP, is a purpose-built enterprise processor for rack-scale AI systems. The chip follows the same system philosophy as NVIDIA’s Grace and Vera CPUs, handling serial processing tasks and the operating system stack for AI GPU servers with bandwidth-heavy networking requirements.

AMD is keeping the design on x86 rather than using specialized Arm CPU cores. EPYC 9006 LP is built around the company’s Zen 6 microarchitecture and includes a 72-core/144-thread Zen 6 complex with boost frequency of up to 5.00 GHz.

The platform includes a 24-channel LPDDR5X memory interface with support for SOCAMM2 modules. Connectivity between the processor and AI GPUs is handled through 112 Gbps-per-pin xGMI, with optimization for AMD Instinct MI455X GPUs based on the CDNA 5 architecture.