diff --git a/benchmarks/single_node/dsv4_fp4_b300_sglang.sh b/benchmarks/single_node/dsv4_fp4_b300_sglang.sh
index 6fae10837..dededd071 100755
--- a/benchmarks/single_node/dsv4_fp4_b300_sglang.sh
+++ b/benchmarks/single_node/dsv4_fp4_b300_sglang.sh
@@ -110,7 +110,7 @@ PYTHONNOUSERSITE=1 sglang serve \
     --port $PORT \
     --trust-remote-code \
     --tp $TP \
-    --max-running-requests "$((CONC * 3 / 2))" \
+    --max-running-requests "$(( CONC * 3 / 2 > 8 ? CONC * 3 / 2 : 8 ))" \
     --mem-fraction-static "$MEM_FRACTION_STATIC" \
     --swa-full-tokens-ratio "$SWA_FULL_TOKENS_RATIO" \
     "${PARALLEL_ARGS[@]}" $EVAL_CONTEXT_ARGS >> $SERVER_LOG 2>&1 &
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 0bce77831..77c2dd31e 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -1862,3 +1862,9 @@
     - "Restore the recipe-per-CONC split (low-latency / balanced / max-throughput) on top of the low-latency-only fallback from #1143; the DeepEP FP8 weight-postprocess path is fixed, so the high-throughput scenario runs again"
     - "Recipes from https://docs.sglang.io/cookbook/autoregressive/DeepSeek/DeepSeek-V4"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1158
+
+- config-keys:
+    - dsv4-fp4-b300-sglang
+  description:
+    - "Floor --max-running-requests at 8 in dsv4_fp4_b300_sglang.sh so low-CONC sweeps don't drop below the queue depth needed for stable benchmarking (CONC * 3 / 2 still applies above CONC=5)"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1173