# NUMA

Learn how NVIDIA's NUMA-aware GPU VM reference architecture improves GPU performance by aligning CPUs, memory, and GPUs to reduce latency and maximize throughput.

[Read Now](/content/ai-and-cloud-native-blog/numa-aware-gpu-vms-how-nvidias-reference-architecture-actually-fixes-it/index.html)

Two identically configured GPU VMs can deliver very different performance because the GPU, CPU cores, and memory may not be placed on the same NUMA node.

[Read Now](/content/ai-and-cloud-native-blog/why-is-your-gpu-vm-slower-than-its-twin-p/index.html)