Llama 3.3 70B — MI300X vs MI325X
Head-to-head AI inference benchmark comparison of MI300X (AMD CDNA 3) 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.
At 48 tok/s/user interactivity on Llama 3.3 70B, MI300X delivers 1295 tok/s/GPU at $0.24 per million tokens; MI325X delivers 1642 tok/s/GPU at $0.22. MI325X is 11% cheaper per token; MI325X delivers 27% more tok/s/GPU at this point.
MI300X posts 781 tok/s/GPU for $0.41 per million tokens at 69 tok/s/user on Llama 3.3 70B; MI325X posts 966 tok/s/GPU for $0.37. MI325X is 12% cheaper per token; MI325X delivers 24% more tok/s/GPU.
Throughput at 91 tok/s/user on Llama 3.3 70B: MI300X hits 398 tok/s/GPU, MI325X hits 572. Per-million costs land at $0.77 and $0.62 respectively. MI325X is 25% cheaper per token; MI325X delivers 44% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MI300X:1294.7MI325X:1642.4 | MI300X:780.7MI325X:965.7 | MI300X:397.9MI325X:571.6 |
| Cost ($/M tok) | MI300X:$0.239MI325X:$0.215 | MI300X:$0.410MI325X:$0.366 | MI300X:$0.771MI325X:$0.618 |
| tok/s/MW | MI300X:723271MI325X:753416 | MI300X:436149MI325X:442973 | MI300X:222267MI325X:262183 |
| Concurrency | MI300X:~61MI325X:~62 | MI300X:~32MI325X:~63 | MI300X:~18MI325X:~26 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.