Micro-Benchmarks & Validation

Tools to validate your cluster, measure performance, and profile training workloads.

📡

Communication Benchmarks

NCCL Tests
NVIDIA Collective Communications Library benchmarks for GPU cluster networking
NCCOM Tests
AWS custom communication library benchmarks
NVSHMEM
GPU shared-memory communication benchmarks
Expert Parallelism
Mixture-of-Experts routing and communication benchmarks
✅

Validation & Health

GPU Cluster Health Checks
Validate GPU cluster readiness and detect hardware issues
📊

Monitoring

Prometheus & Grafana
Monitoring stack for GPU clusters with custom dashboards
🔬

Profiling

NVIDIA Nsight Systems
GPU profiling for training workload optimization