Run Massive-Scale UMAP in Minutes Using Multiple GPUs—Without Losing Accuracy

Summary

UMAP workflows on tens to hundreds of millions of vectors are constrained by the cost of constructing an all-neighbors k-nearest-neighbor graph, especially when repeated runs support exploration and parameter tuning. The NVIDIA cuML and cuVS 25.06 feature distributes this graph-construction stage across multiple GPUs while retaining the out-of-core design that partitions data into balanced clusters, overlaps vectors across nearby clusters, computes local graphs, and merges them into a global graph. knn_n_clusters and knn_overlap_factor expose the memory, runtime, and graph-quality trade-off: more clusters reduce per-GPU data, while more overlap can improve nearest-neighbor recall at higher computational and memory cost. On eight NVIDIA H100 GPUs, the reported evaluation found up to 74x end-to-end speedup over a projected CPU runtime for the 106M-vector MIRACL dataset, completing it in 8 minutes. The post states that performance continues to improve from one to eight GPUs while preserving comparable embedding quality across configurations.

Takeaways

1. 1

   The 106M-vector MIRACL benchmark completed end-to-end processing in 8 minutes on eight NVIDIA H100 GPUs, with up to 74x speedup over a projected CPU runtime.

2. 2

   Increasing knn_n_clusters reduces the amount of data assigned to each cluster and GPU, while increasing knn_overlap_factor generally improves kNN recall and embedding quality at higher memory and computation cost.

3. 3

   Independent per-cluster graph construction avoids expensive all-to-all communication, and the reported scaling improves from one to eight GPUs while preserving comparable trustworthiness across configurations.