diff --git a/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/sglang/mi355x-fp8/8k1k.yaml b/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/sglang/mi355x-fp8/8k1k.yaml
index f578a6a9f0..dd4422d75f 100644
--- a/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/sglang/mi355x-fp8/8k1k.yaml
+++ b/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/sglang/mi355x-fp8/8k1k.yaml
@@ -43,8 +43,8 @@ zip_override_concurrency:
   roles:
     agg:
       args:
-        cuda-graph-max-bs-decode: [4, 8, 16, 32, 64, 128, 256]
-        max-running-requests: [4, 8, 16, 32, 64, 128, 256]
+        cuda-graph-max-bs-decode: [1, 2, 4, 8, 16, 32, 64, 128, 256]
+        max-running-requests: [1, 2, 4, 8, 16, 32, 64, 128, 256]
   benchmark:
     env:
-      CONC: ['4', '8', '16', '32', '64', '128', '256']
+      CONC: ['1', '2', '4', '8', '16', '32', '64', '128', '256']
diff --git a/inferencex-e2e/configs/amd-master.yaml b/inferencex-e2e/configs/amd-master.yaml
index 17b2fa7039..8f855829e5 100644
--- a/inferencex-e2e/configs/amd-master.yaml
+++ b/inferencex-e2e/configs/amd-master.yaml
@@ -206,7 +206,7 @@ qwen3.5-fp8-mi355x-sglang:
       search-space:
       - tp: 4
         ep: 1
-        conc-start: 4
+        conc-start: 1
         conc-end: 256
         srt-recipe: 8k1k.yaml
 
diff --git a/inferencex-e2e/perf-changelog.yaml b/inferencex-e2e/perf-changelog.yaml
index 39b32d17e6..4c5f367644 100644
--- a/inferencex-e2e/perf-changelog.yaml
+++ b/inferencex-e2e/perf-changelog.yaml
@@ -10841,3 +10841,11 @@
     - "Request thinking explicitly on the AgentX client and enable checkpoint prefetch and multithreaded weight loading."
     - "Read SGLang dcp-size in the single-node recipe validator so SGLang points can use dcp-size > 1."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4389
+
+- config-keys:
+    - qwen3.5-fp8-mi355x-sglang
+  scenario-type:
+    - fixed-seq-len
+  description:
+    - "Extend the Qwen3.5 FP8 MI355X SGLang 8k1k (ISL 8192 / OSL 1024, TP4 EP1) concurrency ladder from 4-256 to 1-256 in amd-master.yaml and the matching srt-slurm zip_override. The reused sweep measured and validated GPU power at every point. Leave the MTP arm unchanged."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3380