Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) on Qwen 3.5 397B-A17B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics β same interactions as the main inference chart.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX β
Throughput at 18 tok/s/user on Qwen 3.5 397B-A17B: B300 hits 57140 tok/s/chip, MI325X hits 11757. Per-million costs land at $0.01 and $0.03 respectively. B300 is 137% cheaper per token; B300 delivers 386% more tok/s/chip.
B300 / MI325X on Qwen 3.5 397B-A17B at 28 tok/s/user: 57140 / 9825 tok/s/chip, $0.01 / $0.03 per million tokens. B300 is 183% cheaper per token; B300 delivers 482% more tok/s/chip.
Toward the upper edge of the 8β48 tok/s/user interactivity band, at 39 tok/s/user on Qwen 3.5 397B-A17B: B300 runs 57140 tok/s/chip at $0.01/M tokens, MI325X runs 5555 at $0.06/M. B300 is 401% cheaper per token; B300 delivers 929% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp8 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:57140.4MI325X:11757.0 | B300:57140.4MI325X:9824.8 | B300:57140.4MI325X:5554.8 |
| Cost ($/M tok) | B300:$0.011MI325X:$0.026 | B300:$0.011MI325X:$0.031 | B300:$0.011MI325X:$0.055 |
| tok/s/MW | B300:30073909MI325X:6956810 | B300:30073909MI325X:5813477 | B300:30073909MI325X:3286839 |
| Concurrency | B300:~56MI325X:~48 | B300:~56MI325X:~21 | B300:~56MI325X:~10 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom β’ Drag to pan β’ Double-click to reset β’ Click a point to pin tooltip
Qwen3.5 397B 8K / 1K Total Tokens per $1 TCO vs. Interactivity