Llama 3.3 70B — H200 vs MI300X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI300X (AMD CDNA 3) on Llama 3.3 70B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
At 44 tok/s/user interactivity on Llama 3.3 70B, H200 delivers 3756 tok/s/GPU at $0.11 per million tokens; MI300X delivers 1401 tok/s/GPU at $0.22. H200 is 110% cheaper per token; H200 delivers 168% more tok/s/GPU at this point.
H200 posts 2381 tok/s/GPU for $0.16 per million tokens at 67 tok/s/user on Llama 3.3 70B; MI300X posts 844 tok/s/GPU for $0.38. H200 is 131% cheaper per token; H200 delivers 182% more tok/s/GPU.
Throughput at 90 tok/s/user on Llama 3.3 70B: H200 hits 1514 tok/s/GPU, MI300X hits 409. Per-million costs land at $0.26 and $0.75 respectively. H200 is 191% cheaper per token; H200 delivers 270% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:3755.5MI300X:1400.7 | H200:2381.2MI300X:843.9 | H200:1514.1MI300X:408.8 |
| Cost ($/M tok) | H200:$0.106MI300X:$0.222 | H200:$0.163MI300X:$0.377 | H200:$0.256MI300X:$0.745 |
| tok/s/MW | H200:2170824MI300X:782500 | H200:1376440MI300X:471439 | H200:875220MI300X:228379 |
| Concurrency | H200:~96MI300X:~64 | H200:~64MI300X:~32 | H200:~37MI300X:~20 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.