Llama 3.3 70B · Chip comparison

Llama 3.3 70B — H100 vs MI355X

Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and MI355X (AMD CDNA 4) on Llama 3.3 70B. 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.

H100 / MI355X on Llama 3.3 70B at 38 tok/s/user: 1,848 / 4,213 tok/s/chip, $0.18 / $0.099 per million tokens. MI355X is 44% cheaper per token; MI355X delivers 128% more tok/s/chip.

Around the middle of the 19–97 tok/s/user interactivity band, at 58 tok/s/user on Llama 3.3 70B: H100 runs 1,171 tok/s/chip at $0.28 per million tokens, MI355X runs 2,735 at $0.15. MI355X is 45% cheaper per token; MI355X delivers 133% more tok/s/chip.

Setting 78 tok/s/user as the target on Llama 3.3 70B, H100 produces 877 tok/s/chip ($0.37 per million tokens) and MI355X produces 1,324 ($0.31). MI355X is 15% cheaper per token; MI355X delivers 51% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 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)
H1001,848MI355X4,213
H1001,171MI355X2,735
H100877MI355X1,324
Cost ($/M tok)
H100$0.18MI355X$0.099
H100$0.28MI355X$0.15
H100$0.37MI355X$0.31
tok/s/MW
H1001,348,638MI355X2,015,740
H100855,073MI355X1,308,691
H100640,410MI355X633,355
Concurrency
H100~20MI355X~19
H100~16MI355X~25
H100~8MI355X~8