sglang/ GB200 perf

dsr1-fp4-1k1k-mid-curve

concurrency 4,096 · 5 days ago

cron
passed
re-run
Commit

[AMD] Document Mori XGMI for Single-Node PD Disaggregation (#25094)

clintg6·5 days ago
PR #25094 · [AMD] Document Mori XGMI for Single-Node PD Disaggregation
Run
GitHub Actions27803336065-1
Slurm job5322
GPUs48·prefill 16 / decode 32
ISL / OSL1024 / 1024

Metrics

28 captured

best_of

1.00

burstiness

1.00

completed

40,960

duration

315

max_concurrency

4,096

mean_e2el_ms

30,725ms

mean_itl_ms

1,180ms

mean_tpot_ms

24.09ms

mean_ttft_ms

8,550ms

median_e2el_ms

30,353ms

median_itl_ms

1,159ms

median_tpot_ms

24.19ms

median_ttft_ms

9,027ms

num_prompts

40,960

output_throughput

119,884tok/s

p99_e2el_ms

50,154ms

p99_itl_ms

2,390ms

p99_tpot_ms

29.06ms

p99_ttft_ms

23,859ms

peak_output_tokens_per_s

160,907s

request_throughput

130tok/s

std_e2el_ms

4,604ms

std_itl_ms

353ms

std_tpot_ms

3.15ms

std_ttft_ms

4,172ms

total_input_tokens

37,769,666

total_output_tokens

37,742,239

total_token_throughput

239,856tok/s