Llama 3.3 70B — B200 vs MI325X
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and MI325X (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.
Setting 56 tok/s/user as the target on Llama 3.3 70B, B200 produces 6225 tok/s/GPU ($0.09 per million tokens) and MI325X produces 1359 ($0.26). B200 is 201% cheaper per token; B200 delivers 358% more tok/s/GPU.
At 79 tok/s/user interactivity on Llama 3.3 70B, B200 delivers 4229 tok/s/GPU at $0.13 per million tokens; MI325X delivers 771 tok/s/GPU at $0.45. B200 is 257% cheaper per token; B200 delivers 449% more tok/s/GPU at this point.
B200 posts 3031 tok/s/GPU for $0.18 per million tokens at 102 tok/s/user on Llama 3.3 70B; MI325X posts 409 tok/s/GPU for $0.87. B200 is 384% cheaper per token; B200 delivers 642% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:6225.3MI325X:1359.3 | B200:4229.3MI325X:770.6 | B200:3031.4MI325X:408.7 |
| Cost ($/M tok) | B200:$0.087MI325X:$0.263 | B200:$0.127MI325X:$0.454 | B200:$0.179MI325X:$0.866 |
| tok/s/MW | B200:2868811MI325X:623549 | B200:1949008MI325X:353491 | B200:1396967MI325X:187481 |
| Concurrency | B200:~128MI325X:~64 | B200:~124MI325X:~44 | B200:~63MI325X:~16 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.