Head-to-head AI inference benchmark comparison of GB300 NVL72 (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.
Setting 36 tok/s/user as the target on GLM 5/5.1, GB300 NVL72 produces 12331 tok/s/chip ($0.05 per million tokens) and MI355X produces 1218 ($0.34). GB300 NVL72 is 557% cheaper per token; GB300 NVL72 delivers 912% more tok/s/chip.
At 43 tok/s/user interactivity on GLM 5/5.1, GB300 NVL72 delivers 12059 tok/s/chip at $0.05 per million tokens; MI355X delivers 949 tok/s/chip at $0.44. GB300 NVL72 is 725% cheaper per token; GB300 NVL72 delivers 1171% more tok/s/chip at this point.
GB300 NVL72 posts 11271 tok/s/chip for $0.06 per million tokens at 50 tok/s/user on GLM 5/5.1; MI355X posts 670 tok/s/chip for $0.62. GB300 NVL72 is 992% cheaper per token; GB300 NVL72 delivers 1582% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:12331.4MI355X:1218.0 | GB300 NVL72:12059.4MI355X:948.6 | GB300 NVL72:11270.6MI355X:670.1 |
| Cost ($/M tok) | GB300 NVL72:$0.052MI355X:$0.342 | GB300 NVL72:$0.053MI355X:$0.439 | GB300 NVL72:$0.057MI355X:$0.622 |
| tok/s/MW | GB300 NVL72:5816715MI355X:582775 | GB300 NVL72:5688417MI355X:453886 | GB300 NVL72:5316301MI355X:320643 |
| Concurrency | GB300 NVL72:~2253MI355X:~8 | GB300 NVL72:~2052MI355X:~8 | GB300 NVL72:~1492MI355X:~7 |
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 • Drag to pan • Double-click to reset • Click a point to pin tooltip
1 The ATOM engine is promising, however it has yet to serve production tokens. It is still in its infant stage.
GLM5/5.1 744B 8K / 1K Total Tokens per $1 TCO vs. Interactivity