Kubernetes v1.34 introduces the beta feature prefer-align-cpus-by-uncorecache for the CPU Manager static policy. This option optimizes workloads on processors with split uncore cache architectures. Uncore cache, also known as last-level-cache, is shared across CPU cores. Modern processors use split uncore caches to reduce latency by dividing the cache among CPU groupings. The feature allows Kubernetes to place container CPUs within the same uncore cache, minimizing latency and contention. This cache-aware placement improves throughput for sensitive applications. By default, Kubernetes uses a packed methodology, which can lead to noisy neighbor issues and cross-cache latency. Enabling this feature isolates containers onto individual caches, resolving contention. Use cases include telco applications like vRAN, but benefits depend on the workload, especially for memory-bandwidth-bound applications. To enable, set the CPU Manager policy to static and enable CPUManagerPolicyBetaOptions. Configuration involves modifying the kubelet configuration file to set the policy and options. The feature works on monolithic uncore cache processors by mimicking socket alignment.