Qwen 3.5 397B-A17B · Chip comparison
Qwen 3.5 397B-A17B — MI300X vs TPU7x
Head-to-head AI inference benchmark comparison of MI300X (AMD CDNA 3) and TPU7x (Google Ironwood) 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 →
Throughput at 110 tok/s/user on Qwen 3.5 397B-A17B: MI300X hits 6,427 tok/s/chip, TPU7x hits 84,364. Costs per million tokens land at $0.041 and $0.003 respectively. TPU7x is 92% cheaper per token; TPU7x delivers 13× the tok/s/chip.
MI300X / TPU7x on Qwen 3.5 397B-A17B at 140 tok/s/user: 4,689 / 74,144 tok/s/chip, $0.056 / $0.004 per million tokens. TPU7x is 93% cheaper per token; TPU7x delivers 16× the tok/s/chip.
Toward the upper edge of the 82–198 tok/s/user interactivity band, at 169 tok/s/user on Qwen 3.5 397B-A17B: MI300X runs 3,510 tok/s/chip at $0.075 per million tokens, TPU7x runs 65,182 at $0.004. TPU7x is 94% cheaper per token; TPU7x delivers 19× the tok/s/chip. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | MI300X6,427TPU7x84,364 | MI300X4,689TPU7x74,144 | MI300X3,510TPU7x65,182 |
| Cost ($/M tok) | MI300X$0.041TPU7x$0.003 | MI300X$0.056TPU7x$0.004 | MI300X$0.075TPU7x$0.004 |
| tok/s/MW | MI300X4,623,967TPU7x69,895,634 | MI300X3,373,415TPU7x61,428,647 | MI300X2,525,002TPU7x54,003,207 |
| Concurrency | MI300X~17TPU7x~583 | MI300X~14TPU7x~519 | MI300X~9TPU7x~460 |