DeepSeek R1 · MI355X · Precision comparison

MI355X: FP4 vs FP8 precision comparison

How FP4 and FP8 precision affect DeepSeek R1 inference on MI355X (AMD CDNA 4). Throughput, latency, and cost across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.

At 66 tok/s/user on DeepSeek R1 (MI355X), FP4 delivers 1,848 tok/s/chip at $0.23 per million tokens; FP8 delivers 807 tok/s/chip at $0.52. FP4 delivers 129% more tok/s/chip on the same chip, which makes it 56% cheaper per token. Lower-precision quantization trades model accuracy for throughput — check the evaluation page for quality impact.

FP4 posts 642 tok/s/chip for $0.65 per million tokens at 116 tok/s/user on DeepSeek R1 (MI355X); FP8 posts 295 tok/s/chip for $1.41. FP4 delivers 118% more tok/s/chip on the same chip, which makes it 54% cheaper per token. Quantization-level accuracy differences are tracked on the evaluation tab.

Throughput at 166 tok/s/user on DeepSeek R1 (MI355X): FP4 hits 206 tok/s/chip, FP8 hits 130. Costs per million tokens land at $2.02 and $3.20 respectively. FP4 delivers 59% more tok/s/chip on the same chip, which makes it 37% cheaper per token. The cost-throughput tradeoff from lower precision is only part of the picture — see the evaluation page for accuracy data. (Numbers reflect the default 1k/1k selection for this URL — table and chart below update if you change sequence or model in the controls. Each side uses the best available serving configuration for that precision, which may include speculative decoding such as MTP where recipes exist — the same convention as the other comparison pages.)

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)
FP41,848FP8807
FP4642FP8295
FP4206FP8130
Cost ($/M tok)
FP4$0.23FP8$0.52
FP4$0.65FP8$1.41
FP4$2.02FP8$3.20
tok/s/MW
FP4884,065FP8386,158
FP4307,198FP8141,218
FP498,754FP862,242
Concurrency
FP4~232FP8~115
FP4~64FP8~10
FP4~16FP8~4
DeepSeek R1: FP4 versus FP8 throughput and cost at matched interactivity levels on MI355X
FP4 versus FP8 throughput and cost per million tokens on MI355X for this comparison's canonical default workload
Open full-size image ↗