GLM 5/5.1 — B300 vs MI355X
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on GLM 5/5.1. 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.
B300 posts 882 tok/s/GPU for $0.74 per million tokens at 38 tok/s/user on GLM 5/5.1; MI355X posts 835 tok/s/GPU for $0.49. MI355X is 51% cheaper per token; B300 delivers 6% more tok/s/GPU.
Throughput at 60 tok/s/user on GLM 5/5.1: B300 hits 584 tok/s/GPU, MI355X hits 373. Per-million costs land at $1.10 and $1.09 respectively. Cost per token is essentially tied; B300 delivers 56% more tok/s/GPU.
B300 / MI355X on GLM 5/5.1 at 81 tok/s/user: 417 / 240 tok/s/GPU, $1.54 / $1.73 per million tokens. B300 is 12% cheaper per token; B300 delivers 74% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:882.0MI355X:834.6 | B300:584.0MI355X:373.3 | B300:416.9MI355X:239.8 |
| Cost ($/M tok) | B300:$0.739MI355X:$0.491 | B300:$1.102MI355X:$1.092 | B300:$1.543MI355X:$1.729 |
| tok/s/MW | B300:406438MI355X:314949 | B300:269127MI355X:140880 | B300:192098MI355X:90503 |
| Concurrency | B300:~93MI355X:~121 | B300:~40MI355X:~13 | B300:~21MI355X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.