Llama 3.3 70B — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) 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.
Throughput at 54 tok/s/user on Llama 3.3 70B: H200 hits 2974 tok/s/GPU, MI355X hits 2216. Per-million costs land at $0.13 and $0.18 respectively. H200 is 39% cheaper per token; H200 delivers 34% more tok/s/GPU.
H200 / MI355X on Llama 3.3 70B at 75 tok/s/user: 2062 / 1536 tok/s/GPU, $0.19 / $0.26 per million tokens. H200 is 37% cheaper per token; H200 delivers 34% more tok/s/GPU.
Toward the upper edge of the 34–116 tok/s/user interactivity band, at 96 tok/s/user on Llama 3.3 70B: H200 runs 1312 tok/s/GPU at $0.30/M tokens, MI355X runs 750 at $0.55/M. H200 is 82% cheaper per token; H200 delivers 75% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:2974.5MI355X:2215.8 | H200:2062.3MI355X:1536.1 | H200:1312.1MI355X:749.9 |
| Cost ($/M tok) | H200:$0.131MI355X:$0.182 | H200:$0.190MI355X:$0.261 | H200:$0.300MI355X:$0.548 |
| tok/s/MW | H200:1719355MI355X:836163 | H200:1192078MI355X:579660 | H200:758430MI355X:282963 |
| Concurrency | H200:~64MI355X:~50 | H200:~63MI355X:~46 | H200:~29MI355X:~16 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.