Muse Spark Open 816B — GB200 NVL72 vs TPU7x
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and TPU7x (Google Ironwood) on Muse Spark Open 816B. 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 →
GB200 NVL72 / TPU7x on Muse Spark Open 816B at 165 tok/s/user: 21295 / 18720 tok/s/chip, $0.02 / $0.02 per million tokens. TPU7x is 59% cheaper per token; GB200 NVL72 delivers 14% more tok/s/chip.
Around the middle of the 95–377 tok/s/user interactivity band, at 236 tok/s/user on Muse Spark Open 816B: GB200 NVL72 runs 14733 tok/s/chip at $0.04/M tokens, TPU7x runs 15517 at $0.02/M. TPU7x is 90% cheaper per token; TPU7x delivers 5% more tok/s/chip.
Setting 307 tok/s/user as the target on Muse Spark Open 816B, GB200 NVL72 produces 8773 tok/s/chip ($0.06 per million tokens) and TPU7x produces 11176 ($0.03). TPU7x is 130% cheaper per token; TPU7x delivers 27% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:21294.7TPU7x:18719.7 | GB200 NVL72:14733.2TPU7x:15517.1 | GB200 NVL72:8772.5TPU7x:11175.7 |
| Cost ($/M tok) | GB200 NVL72:$0.024TPU7x:$0.015 | GB200 NVL72:$0.035TPU7x:$0.018 | GB200 NVL72:$0.059TPU7x:$0.026 |
| tok/s/MW | GB200 NVL72:11387556TPU7x:15509294 | GB200 NVL72:7878704TPU7x:12855931 | GB200 NVL72:4691197TPU7x:9259054 |
| Concurrency | GB200 NVL72:~12TPU7x:~13 | GB200 NVL72:~9TPU7x:~9 | GB200 NVL72:~5TPU7x:~7 |