GLM 5/5.1 · Chip comparison
GLM 5/5.1 — B300 vs MI325X
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) 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.
Near the low end of the 10–28 tok/s/user interactivity band, at 14 tok/s/user on GLM 5/5.1: B300 runs 1,658 tok/s/chip at $0.38 per million tokens, MI325X runs 491 at $0.62. B300 is 39% cheaper per token; B300 delivers 3.4× the tok/s/chip.
Setting 19 tok/s/user as the target on GLM 5/5.1, B300 produces 1,551 tok/s/chip ($0.40 per million tokens) and MI325X produces 309 ($0.99). B300 is 59% cheaper per token; B300 delivers 5.0× the tok/s/chip.
At 24 tok/s/user interactivity on GLM 5/5.1, B300 delivers 1,458 tok/s/chip at $0.43 per million tokens; MI325X delivers 200 tok/s/chip at $1.53. B300 is 72% cheaper per token; B300 delivers 7.3× the tok/s/chip at this point. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B3001,658MI325X491 | B3001,551MI325X309 | B3001,458MI325X200 |
| Cost ($/M tok) | B300$0.38MI325X$0.62 | B300$0.40MI325X$0.99 | B300$0.43MI325X$1.53 |
| tok/s/MW | B300872,815MI325X290,711 | B300816,521MI325X182,855 | B300767,274MI325X118,096 |
| Concurrency | B300~107MI325X~34 | B300~75MI325X~15 | B300~56MI325X~8 |