Qwen 3.5 397B-A17B — GB200 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on Qwen 3.5 397B-A17B. 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.
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 →
At 68 tok/s/user interactivity on Qwen 3.5 397B-A17B, GB200 NVL72 delivers 82001 tok/s/chip at $0.01 per million tokens; MI355X delivers 6874 tok/s/chip at $0.06. GB200 NVL72 is 862% cheaper per token; GB200 NVL72 delivers 1093% more tok/s/chip at this point.
GB200 NVL72 posts 80080 tok/s/chip for $0.01 per million tokens at 72 tok/s/user on Qwen 3.5 397B-A17B; MI355X posts 5922 tok/s/chip for $0.07. GB200 NVL72 is 990% cheaper per token; GB200 NVL72 delivers 1252% more tok/s/chip.
Throughput at 76 tok/s/user on Qwen 3.5 397B-A17B: GB200 NVL72 hits 78291 tok/s/chip, MI355X hits 4971. Per-million costs land at $0.01 and $0.08 respectively. GB200 NVL72 is 1170% cheaper per token; GB200 NVL72 delivers 1475% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:82001.3MI355X:6873.5 | GB200 NVL72:80080.4MI355X:5922.3 | GB200 NVL72:78291.3MI355X:4970.9 |
| Cost ($/M tok) | GB200 NVL72:$0.006MI355X:$0.061 | GB200 NVL72:$0.006MI355X:$0.070 | GB200 NVL72:$0.007MI355X:$0.084 |
| tok/s/MW | GB200 NVL72:43850976MI355X:3288777 | GB200 NVL72:42823736MI355X:2833617 | GB200 NVL72:41867006MI355X:2378408 |
| Concurrency | GB200 NVL72:~38MI355X:~6 | GB200 NVL72:~36MI355X:~5 | GB200 NVL72:~35MI355X:~5 |
Total Tokens per $1 USD (Owning - Neocloud Giant) vs. Interactivity
Qwen3.5 397B • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™
Qwen3.5 397B • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™TCO $/chip/hr: VR200: 3.61H100: 1.55H200: 1.59B200: 2.07B300: 2.52GB200: 2.26GB300: 2.79MI300X: 1.16MI325X: 1.32MI355X: 2.09RTX6000PRO: 0.75Jalapeño (Teacup): 1.56
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.
No data available
Please change the model, sequence, precision, date range or chip selection.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
Interactivity is the rate at which a single user receives generated tokens while the model streams its answer — how quickly new words appear on screen. Higher values feel snappier; operators trade it against batch throughput.
How many total tokens (input + output) one US dollar of infrastructure spend buys, priced with the all-in hourly ownership cost of a Neocloud Giant operator. It is the reciprocal of cost per token, so higher means cheaper.
Formula: tok/$ = (total tok/s/chip × 3,600) ÷ all-in cost per chip-hour ($)
1 The ATOM engine is promising, however it has yet to serve production tokens. It is still in its infant stage.