Recommendation inference — user-item scoring queries per second per GPU. · DLRM-v2 · collaborative filtering · MLPerf v4.1/v5.0 · on-demand pricing
| GPU | Scenario | samp/s per GPU | |
|---|---|---|---|
| NVIDIA H200-SXM-141GB | Offline | 80,470 samp/s | |
| NVIDIA H100-SXM-80GB | Offline | 75,264 samp/s | |
| NVIDIA H200-SXM-141GB | Server | 73,151 samp/s | |
| NVIDIA H100-SXM-80GB | Server | 69,513 samp/s |
| Provider | GPU | Scenario | Price | samp/$ | |
|---|---|---|---|---|---|
| RunPod | NVIDIA H100-SXM-80GB | Offline | $2.99/GPU·hr 1d ago | 90,618,261 | |
| RunPod | NVIDIA H100-SXM-80GB | Server | $2.99/GPU·hr 1d ago | 83,694,121 | |
| Crusoe | NVIDIA H100-SXM-80GB | Offline | $3.90/GPU·hr 1d ago | 69,474,000 | |
| Lambda Labs | NVIDIA H100-SXM-80GB | Offline | $3.99/GPU·hr 1d ago | 67,906,917 | |
| Crusoe | NVIDIA H200-SXM-141GB | Offline | $4.29/GPU·hr 1d ago | 67,526,853 | |
| RunPod | NVIDIA H200-SXM-141GB | Offline | $4.31/GPU·hr 1d ago | 67,213,503 | |
| Crusoe | NVIDIA H100-SXM-80GB | Server | $3.90/GPU·hr 1d ago | 64,165,493 | |
| Lambda Labs | NVIDIA H100-SXM-80GB | Server | $3.99/GPU·hr 1d ago | 62,718,151 | |
| Crusoe | NVIDIA H200-SXM-141GB | Server | $4.29/GPU·hr 1d ago | 61,385,431 | |
| RunPod | NVIDIA H200-SXM-141GB | Server | $4.31/GPU·hr 1d ago | 61,100,580 | |
| CoreWeave | NVIDIA H200-SXM-141GB | Offline | $6.31/GPU·hr 1d ago | 45,909,699 | |
| CoreWeave | NVIDIA H100-SXM-80GB | Offline | $6.16/GPU·hr 1d ago | 43,985,162 | |
| CoreWeave | NVIDIA H200-SXM-141GB | Server | $6.31/GPU·hr 1d ago | 41,734,311 | |
| CoreWeave | NVIDIA H100-SXM-80GB | Server | $6.16/GPU·hr 1d ago | 40,624,257 | |
| OCI | NVIDIA H200-SXM-141GB | Offline | $10.00/GPU·hr today | 28,969,020 | |
| Azure (eastus2) | NVIDIA H200-SXM-141GB | Offline | $10.60/GPU·hr today | 27,329,264 | |
| GCP (us-central1) | NVIDIA H200-SXM-141GB | Offline | $10.60/GPU·hr today | 27,327,038 | |
| OCI | NVIDIA H200-SXM-141GB | Server | $10.00/GPU·hr today | 26,334,350 | |
| OCI | NVIDIA H100-SXM-80GB | Offline | $10.75/GPU·hr today | 25,204,521 | |
| Azure (eastus2) | NVIDIA H200-SXM-141GB | Server | $10.60/GPU·hr today | 24,843,727 | |
| GCP (us-central1) | NVIDIA H200-SXM-141GB | Server | $10.60/GPU·hr today | 24,841,703 | |
| GCP (us-central1) | NVIDIA H100-SXM-80GB | Offline | $11.06/GPU·hr today | 24,495,297 | |
| OCI | NVIDIA H100-SXM-80GB | Server | $10.75/GPU·hr today | 23,278,644 | |
| GCP (us-central1) | NVIDIA H100-SXM-80GB | Server | $11.06/GPU·hr today | 22,623,611 | |
| AWS (us-east-1) | NVIDIA H100-SXM-80GB | Offline | $12.29/GPU·hr today | 22,046,265 | |
| Azure (eastus2) | NVIDIA H100-SXM-80GB | Offline | $12.29/GPU·hr today | 22,046,265 | |
| Azure (eastus) | NVIDIA H100-SXM-80GB | Offline | $12.29/GPU·hr today | 22,046,265 | |
| AWS (us-east-1) | NVIDIA H100-SXM-80GB | Server | $12.29/GPU·hr today | 20,361,711 | |
| Azure (eastus2) | NVIDIA H100-SXM-80GB | Server | $12.29/GPU·hr today | 20,361,711 | |
| Azure (eastus) | NVIDIA H100-SXM-80GB | Server | $12.29/GPU·hr today | 20,361,711 |