diff --git a/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi325x.sh b/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi325x.sh
index 0f6a801136..78a60cec69 100644
--- a/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi325x.sh
+++ b/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi325x.sh
@@ -37,7 +37,7 @@ python3 -m sglang.launch_server \
 --model-path=$MODEL --host=0.0.0.0 --port=$PORT --trust-remote-code \
 --tensor-parallel-size=$TP \
 --mem-fraction-static=0.8 \
---cuda-graph-max-bs=128 \
+--cuda-graph-max-bs-decode=128 \
 --chunked-prefill-size=131072 \
 --num-continuous-decode-steps=4 \
 --max-prefill-tokens=131072 \
diff --git a/configs/amd-master.yaml b/configs/amd-master.yaml
index 9b4fcadeb7..11a38ee897 100644
--- a/configs/amd-master.yaml
+++ b/configs/amd-master.yaml
@@ -78,7 +78,7 @@ dsr1-fp8-mi300x-sglang:
       - { tp: 8, conc-start: 4, conc-end: 64 }
 
 dsr1-fp8-mi325x-sglang:
-  image: lmsysorg/sglang:v0.5.19-rocm700-mi30x
+  image: lmsysorg/sglang:v0.5.20-rocm720-mi30x
   model: deepseek-ai/DeepSeek-R1-0528
   model-prefix: dsr1
   runner: mi325x
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index c5b4309f5f..35e5afc5c8 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -8455,3 +8455,9 @@
   description:
     - "Update B200 vLLM AgentX to DSpark6 and a new image with TP8 and DEP8 configurations."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3274
+
+- config-keys:
+    - dsr1-fp8-mi325x-sglang
+  description:
+    - "Update SGLang image from v0.5.19-rocm700-mi30x to v0.5.20-rocm720-mi30x and rename the CUDA-graph max-bs flag."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3318