diff --git a/.github/workflows/pr-test-amd-rocm720.yml b/.github/workflows/pr-test-amd-rocm720.yml
index c536369dc..a3dd3999d 100644
--- a/.github/workflows/pr-test-amd-rocm720.yml
+++ b/.github/workflows/pr-test-amd-rocm720.yml
@@ -272,8 +272,7 @@ jobs:
       - name: Run test
         timeout-minutes: 20
         run: |
-          docker exec -w /sglang-checkout/sgl-kernel/tests ci_sglang python3 -m pytest test_amd_deterministic_custom_allreduce.py
-          docker exec -w /sglang-checkout/sgl-kernel/tests ci_sglang python3 -m pytest test_amd_nccl_allreduce_determinism.py
+          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite sgl-kernel-unit-test-2-gpu-amd
 
   # =============================================== primary ====================================================
 
diff --git a/.github/workflows/pr-test-amd.yml b/.github/workflows/pr-test-amd.yml
index 0a569e352..e5991f614 100644
--- a/.github/workflows/pr-test-amd.yml
+++ b/.github/workflows/pr-test-amd.yml
@@ -283,17 +283,7 @@ jobs:
         env:
           CONTINUE_ON_ERROR: ${{ needs.check-changes.outputs.continue_on_error }}
         run: |
-          failures=0
-          run_pytest() {
-            if [[ "$CONTINUE_ON_ERROR" == "true" ]]; then
-              "$@" || failures=$((failures + 1))
-            else
-              "$@"
-            fi
-          }
-          run_pytest docker exec -w /sglang-checkout/sgl-kernel/tests ci_sglang python3 -m pytest test_amd_deterministic_custom_allreduce.py
-          run_pytest docker exec -w /sglang-checkout/sgl-kernel/tests ci_sglang python3 -m pytest test_amd_nccl_allreduce_determinism.py
-          exit $failures
+          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite sgl-kernel-unit-test-2-gpu-amd ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
 
   # =============================================== primary ====================================================
 
diff --git a/test/registered/jit/test_activation.py b/test/registered/jit/test_activation.py
index 832d73944..b01660eec 100644
--- a/test/registered/jit/test_activation.py
+++ b/test/registered/jit/test_activation.py
@@ -10,10 +10,11 @@ from sglang.jit_kernel.activation import (
     run_activation,
 )
 from sglang.jit_kernel.utils import get_ci_test_range
-from sglang.test.ci.ci_register import register_cuda_ci
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=20, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=30, suite="nightly-kernel-1-gpu", nightly=True)
+register_amd_ci(est_time=20, suite="jit-kernel-unit-test-amd")
 
 
 OPS = SUPPORTED_ACTIVATIONS
diff --git a/sgl-kernel/tests/test_amd_deterministic_custom_allreduce.py b/test/registered/jit/test_amd_deterministic_custom_allreduce.py
similarity index 98%
rename from sgl-kernel/tests/test_amd_deterministic_custom_allreduce.py
rename to test/registered/jit/test_amd_deterministic_custom_allreduce.py
index aa7125925..c5a0525a6 100644
--- a/sgl-kernel/tests/test_amd_deterministic_custom_allreduce.py
+++ b/test/registered/jit/test_amd_deterministic_custom_allreduce.py
@@ -23,6 +23,9 @@ import torch
 import torch.distributed as dist
 
 from sglang.srt.environ import envs
+from sglang.test.ci.ci_register import register_amd_ci
+
+register_amd_ci(est_time=120, suite="sgl-kernel-unit-test-2-gpu-amd")
 
 
 def get_open_port():
diff --git a/sgl-kernel/tests/test_amd_nccl_allreduce_determinism.py b/test/registered/jit/test_amd_nccl_allreduce_determinism.py
similarity index 98%
rename from sgl-kernel/tests/test_amd_nccl_allreduce_determinism.py
rename to test/registered/jit/test_amd_nccl_allreduce_determinism.py
index 492d54142..480786a8b 100644
--- a/sgl-kernel/tests/test_amd_nccl_allreduce_determinism.py
+++ b/test/registered/jit/test_amd_nccl_allreduce_determinism.py
@@ -20,6 +20,10 @@ import pytest
 import torch
 import torch.distributed as dist
 
+from sglang.test.ci.ci_register import register_amd_ci
+
+register_amd_ci(est_time=120, suite="sgl-kernel-unit-test-2-gpu-amd")
+
 
 def get_open_port():
     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
diff --git a/test/registered/jit/test_per_token_group_quant_8bit.py b/test/registered/jit/test_per_token_group_quant_8bit.py
index f370cb19a..4b4de2dec 100644
--- a/test/registered/jit/test_per_token_group_quant_8bit.py
+++ b/test/registered/jit/test_per_token_group_quant_8bit.py
@@ -23,10 +23,11 @@ from sglang.srt.layers.quantization.fp8_kernel import (
 from sglang.srt.layers.quantization.fp8_kernel import (
     per_token_group_quant_8bit as triton_per_token_group_quant_8bit,
 )
-from sglang.test.ci.ci_register import register_cuda_ci
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=16, suite="base-b-kernel-unit-1-gpu-large")
 register_cuda_ci(est_time=120, suite="nightly-kernel-1-gpu", nightly=True)
+register_amd_ci(est_time=16, suite="jit-kernel-unit-test-amd")
 
 configs = list(
     itertools.product(
diff --git a/test/run_suite.py b/test/run_suite.py
index accc5191a..d9c139e64 100644
--- a/test/run_suite.py
+++ b/test/run_suite.py
@@ -39,6 +39,7 @@ PER_COMMIT_SUITES = {
         "stage-b-test-1-gpu-large-amd",
         "stage-b-test-2-gpu-large-amd",
         "jit-kernel-unit-test-amd",
+        "sgl-kernel-unit-test-2-gpu-amd",
         "stage-c-test-4-gpu-amd",
         "stage-c-test-large-8-gpu-amd",
         "stage-c-test-large-8-gpu-amd-mi35x",