Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on DeepSeek V4.1 Flash 552B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics β 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 β
Throughput at 175 tok/s/user on DeepSeek V4.1 Flash 552B: B300 hits 112531 tok/s/chip, MI355X hits 76921. Per-million costs land at $0.01 and $0.01 respectively. MI355X is 3% cheaper per token; B300 delivers 46% more tok/s/chip.
B300 / MI355X on DeepSeek V4.1 Flash 552B at 308 tok/s/user: 28906 / 34676 tok/s/chip, $0.02 / $0.01 per million tokens. MI355X is 81% cheaper per token; MI355X delivers 20% more tok/s/chip.
Toward the upper edge of the 43β574 tok/s/user interactivity band, at 442 tok/s/user on DeepSeek V4.1 Flash 552B: B300 runs 11416 tok/s/chip at $0.05/M tokens, MI355X runs 15557 at $0.03/M. MI355X is 105% cheaper per token; MI355X delivers 36% more tok/s/chip. (Numbers reflect the default agentic-traces Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:112531.3MI355X:76920.5 | B300:28906.4MI355X:34675.9 | B300:11415.5MI355X:15557.4 |
| Cost ($/M tok) | B300:$0.006MI355X:$0.005 | B300:$0.022MI355X:$0.012 | B300:$0.055MI355X:$0.027 |
| tok/s/MW | B300:59226983MI355X:36804075 | B300:15213917MI355X:16591319 | B300:6008169MI355X:7443715 |
| Concurrency | B300:~29MI355X:~32 | B300:~16MI355X:~9 | B300:~2MI355X:~6 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
