Head-to-head AI inference benchmark comparison of MI300X (AMD CDNA 3) and MI325X (AMD CDNA 3) on DeepSeekv4 Pro 0813 1.6T. 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.
At 12 tok/s/user interactivity on DeepSeekv4 Pro 0813 1.6T, MI300X delivers 292 tok/s/chip at $0.90 per million tokens; MI325X delivers 387 tok/s/chip at $0.79. MI325X is 14% cheaper per token; MI325X delivers 32% more tok/s/chip at this point.
MI300X posts 130 tok/s/chip for $2.03 per million tokens at 19 tok/s/user on DeepSeekv4 Pro 0813 1.6T; MI325X posts 207 tok/s/chip for $1.47. MI325X is 38% cheaper per token; MI325X delivers 59% more tok/s/chip.
Throughput at 25 tok/s/user on DeepSeekv4 Pro 0813 1.6T: MI300X hits 88 tok/s/chip, MI325X hits 111. Per-million costs land at $2.99 and $2.74 respectively. MI325X is 9% cheaper per token; MI325X delivers 26% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | MI300X:292.2MI325X:386.8 | MI300X:130.0MI325X:207.3 | MI300X:88.2MI325X:111.4 |
| Cost ($/M tok) | MI300X:$0.903MI325X:$0.790 | MI300X:$2.030MI325X:$1.474 | MI300X:$2.992MI325X:$2.743 |
| tok/s/MW | MI300X:210222MI325X:228896 | MI300X:93512MI325X:122634 | MI300X:63442MI325X:65918 |
| Concurrency | MI300X:~101MI325X:~136 | MI300X:~29MI325X:~46 | MI300X:~15MI325X:~18 |
DeepSeek V4 Pro 0813 1.6T • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™ • Input $1/M tok · Output $1/M tok
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.
Gross token revenue a GPU could earn per hour at this operating point. The normalized source prices uncached input and output at $1 per million and cached input at $0.10 per million. The OpenRouter source uses the selected model’s current public prices, with a 10%-of-input fallback when no cache-read price is published. Revenue prices uncached input, cached input, and output separately. Input share comes from compatible measured input/output throughput. When disaggregated rates use different GPU denominators, fixed sequences use ISL:OSL and Agentic traces use measured prompt/generation tokens. Agentic cache hit combines GPU and external cache when external cache is reported, otherwise GPU and CPU cache. Historical Trends interpolates total throughput, input share, and cache hit separately before pricing. A partially measured cache frontier receives no cache discount. This turns the throughput/interactivity tradeoff into a business-facing SLA curve.
Formula: $/GPU/hr = (uncached input × input price + cached input × cache-read price + output × output price) per GPU-second, scaled to one hour
DeepSeek V4 Pro 0813 1.6T • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™ •