DeepSeek V4 Pro 0813 1.6T · Chip comparison

DeepSeek V4 Pro 0813 1.6T — GB300 NVL72 vs MI355X

Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on DeepSeek V4 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.

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 →

GB300 NVL72 posts 122,275 tok/s/chip for $0.005 per million tokens at 90 tok/s/user on DeepSeek V4 Pro 0813 1.6T; MI355X posts 24,434 tok/s/chip for $0.017. GB300 NVL72 is 69% cheaper per token; GB300 NVL72 delivers 5.0× the tok/s/chip.

Throughput at 143 tok/s/user on DeepSeek V4 Pro 0813 1.6T: GB300 NVL72 hits 54,943 tok/s/chip, MI355X hits 20,260. Costs per million tokens land at $0.012 and $0.021 respectively. GB300 NVL72 is 43% cheaper per token; GB300 NVL72 delivers 171% more tok/s/chip.

GB300 NVL72 / MI355X on DeepSeek V4 Pro 0813 1.6T at 196 tok/s/user: 10,677 / 6,694 tok/s/chip, $0.060 / $0.062 per million tokens. GB300 NVL72 is 3% cheaper per token; GB300 NVL72 delivers 60% 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.)

View performance per dollar →

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)
GB300 NVL72122,275MI355X24,434
GB300 NVL7254,943MI355X20,260
GB300 NVL7210,677MI355X6,694
Cost ($/M tok)
GB300 NVL72$0.005MI355X$0.017
GB300 NVL72$0.012MI355X$0.021
GB300 NVL72$0.060MI355X$0.062
tok/s/MW
GB300 NVL7257,676,918MI355X11,690,810
GB300 NVL7225,916,658MI355X9,693,814
GB300 NVL725,036,314MI355X3,202,874
Concurrency
GB300 NVL72~760MI355X~48
GB300 NVL72~234MI355X~39
GB300 NVL72~18MI355X~11