diff --git a/.github/workflows/pr-test.yml b/.github/workflows/pr-test.yml
index 23e94b866..793a3b3d9 100644
--- a/.github/workflows/pr-test.yml
+++ b/.github/workflows/pr-test.yml
@@ -672,6 +672,58 @@ jobs:
           fi
           python3 run_suite.py --hw cuda --suite stage-b-test-large-1-gpu $CONTINUE_ON_ERROR_FLAG
 
+  # 5090 dry run stage - only runs on scheduled CI or when explicitly targeted via /rerun-stage
+  # Used to validate 5090 compatibility before full integration
+  stage-b-test-small-1-gpu-5090:
+    needs: [check-changes, call-gate, stage-a-test-1, sgl-kernel-build-wheels]
+    if: |
+      always() &&
+      (
+        (inputs.target_stage == 'stage-b-test-small-1-gpu-5090') ||
+        (
+          !inputs.target_stage &&
+          github.event_name == 'schedule' &&
+          (!failure() && !cancelled()) &&
+          ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
+        )
+      )
+    runs-on: 1-gpu-5090
+    continue-on-error: true
+    env:
+      RUNNER_LABELS: 1-gpu-5090
+      IS_BLACKWELL: "1"
+    strategy:
+      fail-fast: false
+      max-parallel: 8
+      matrix:
+        partition: [0, 1, 2, 3, 4, 5, 6, 7]
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.pr_head_sha || inputs.ref || github.sha }}
+
+      - name: Download artifacts
+        if: needs.check-changes.outputs.sgl_kernel == 'true'
+        uses: actions/download-artifact@v4
+        with:
+          path: sgl-kernel/dist/
+          merge-multiple: true
+          pattern: wheel-python3.10-cuda12.9
+
+      - name: Install dependencies
+        timeout-minutes: 10
+        run: |
+          source /etc/profile.d/sglang-ci.sh
+          CUSTOM_BUILD_SGL_KERNEL=${{needs.check-changes.outputs.sgl_kernel}} bash scripts/ci/ci_install_dependency.sh
+
+      - name: Run test
+        timeout-minutes: 30
+        run: |
+          source /etc/profile.d/sglang-ci.sh
+          cd test/
+          python3 run_suite.py --hw cuda --suite stage-b-test-small-1-gpu-5090 --auto-partition-id ${{ matrix.partition }} --auto-partition-size 8 --continue-on-error
+
   stage-b-test-large-2-gpu:
     needs: [check-changes, call-gate, stage-a-test-1, sgl-kernel-build-wheels]
     if: |
diff --git a/scripts/ci/ci_install_dependency.sh b/scripts/ci/ci_install_dependency.sh
index 4f975124c..582179428 100755
--- a/scripts/ci/ci_install_dependency.sh
+++ b/scripts/ci/ci_install_dependency.sh
@@ -136,7 +136,18 @@ if [ "${CUSTOM_BUILD_SGL_KERNEL:-}" = "true" ]; then
     fi
     $PIP_CMD install sgl-kernel/dist/sgl_kernel-${SGL_KERNEL_VERSION_FROM_KERNEL}-cp310-abi3-manylinux2014_${WHEEL_ARCH}.whl --force-reinstall $PIP_INSTALL_SUFFIX
 else
-    $PIP_CMD install sgl-kernel==${SGL_KERNEL_VERSION_FROM_SRT} --force-reinstall $PIP_INSTALL_SUFFIX
+    # On Blackwell machines, skip reinstall if correct version already installed to avoid race conditions
+    if [ "$IS_BLACKWELL" = "1" ]; then
+        INSTALLED_SGL_KERNEL=$(pip show sgl-kernel 2>/dev/null | grep "^Version:" | awk '{print $2}' || echo "")
+        if [ "$INSTALLED_SGL_KERNEL" = "$SGL_KERNEL_VERSION_FROM_SRT" ]; then
+            echo "sgl-kernel==${SGL_KERNEL_VERSION_FROM_SRT} already installed, skipping reinstall"
+        else
+            echo "Installing sgl-kernel==${SGL_KERNEL_VERSION_FROM_SRT} (current: ${INSTALLED_SGL_KERNEL:-none})"
+            $PIP_CMD install sgl-kernel==${SGL_KERNEL_VERSION_FROM_SRT} $PIP_INSTALL_SUFFIX
+        fi
+    else
+        $PIP_CMD install sgl-kernel==${SGL_KERNEL_VERSION_FROM_SRT} --force-reinstall $PIP_INSTALL_SUFFIX
+    fi
 fi
 
 # Show current packages
@@ -152,10 +163,30 @@ if [ "$IS_BLACKWELL" != "1" ]; then
 fi
 
 # DeepEP depends on nvshmem 3.4.5
-$PIP_CMD install nvidia-nvshmem-cu12==3.4.5 --force-reinstall $PIP_INSTALL_SUFFIX
+# On Blackwell machines, skip reinstall if correct version already installed to avoid race conditions
+if [ "$IS_BLACKWELL" = "1" ]; then
+    INSTALLED_NVSHMEM=$(pip show nvidia-nvshmem-cu12 2>/dev/null | grep "^Version:" | awk '{print $2}' || echo "")
+    if [ "$INSTALLED_NVSHMEM" = "3.4.5" ]; then
+        echo "nvidia-nvshmem-cu12==3.4.5 already installed, skipping reinstall"
+    else
+        $PIP_CMD install nvidia-nvshmem-cu12==3.4.5 $PIP_INSTALL_SUFFIX
+    fi
+else
+    $PIP_CMD install nvidia-nvshmem-cu12==3.4.5 --force-reinstall $PIP_INSTALL_SUFFIX
+fi
 
 # Cudnn with version less than 9.16.0.29 will cause performance regression on Conv3D kernel
-$PIP_CMD install nvidia-cudnn-cu12==9.16.0.29 --force-reinstall $PIP_INSTALL_SUFFIX
+# On Blackwell machines, skip reinstall if correct version already installed to avoid race conditions
+if [ "$IS_BLACKWELL" = "1" ]; then
+    INSTALLED_CUDNN=$(pip show nvidia-cudnn-cu12 2>/dev/null | grep "^Version:" | awk '{print $2}' || echo "")
+    if [ "$INSTALLED_CUDNN" = "9.16.0.29" ]; then
+        echo "nvidia-cudnn-cu12==9.16.0.29 already installed, skipping reinstall"
+    else
+        $PIP_CMD install nvidia-cudnn-cu12==9.16.0.29 $PIP_INSTALL_SUFFIX
+    fi
+else
+    $PIP_CMD install nvidia-cudnn-cu12==9.16.0.29 --force-reinstall $PIP_INSTALL_SUFFIX
+fi
 $PIP_CMD uninstall xformers || true
 
 # Install flashinfer-jit-cache with caching and retry logic (flashinfer.ai can have transient DNS issues)
diff --git a/scripts/ci/slash_command_handler.py b/scripts/ci/slash_command_handler.py
index 1f95baa92..a61329bd3 100644
--- a/scripts/ci/slash_command_handler.py
+++ b/scripts/ci/slash_command_handler.py
@@ -227,6 +227,7 @@ def handle_rerun_stage(
         "stage-a-cpu-only",
         "stage-b-test-small-1-gpu",
         "stage-b-test-large-1-gpu",
+        "stage-b-test-small-1-gpu-5090",
         "stage-b-test-large-2-gpu",
         "stage-c-test-large-4-gpu",
         "stage-c-test-large-4-gpu-b200",
diff --git a/test/registered/attention/test_create_kvindices.py b/test/registered/attention/test_create_kvindices.py
index 881e68d6e..0928dbfbc 100644
--- a/test/registered/attention/test_create_kvindices.py
+++ b/test/registered/attention/test_create_kvindices.py
@@ -9,6 +9,7 @@ from sglang.test.test_utils import CustomTestCase
 
 # Triton kernel unit test for KV indices creation
 register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/attention/test_mamba_unittest.py b/test/registered/attention/test_mamba_unittest.py
index 0a0f45fb8..3e9ce3160 100644
--- a/test/registered/attention/test_mamba_unittest.py
+++ b/test/registered/attention/test_mamba_unittest.py
@@ -15,6 +15,7 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestMamba(unittest.TestCase):
diff --git a/test/registered/attention/test_radix_attention.py b/test/registered/attention/test_radix_attention.py
index 5931a0921..a9871adba 100644
--- a/test/registered/attention/test_radix_attention.py
+++ b/test/registered/attention/test_radix_attention.py
@@ -15,6 +15,7 @@ from sglang.test.test_utils import (
 
 # RadixAttention server integration tests
 register_cuda_ci(est_time=100, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=79, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=100, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/attention/test_radix_cache_unit.py b/test/registered/attention/test_radix_cache_unit.py
index cb30a1e50..08282df29 100644
--- a/test/registered/attention/test_radix_cache_unit.py
+++ b/test/registered/attention/test_radix_cache_unit.py
@@ -21,6 +21,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 # CPU-based unit test, runs quickly on any GPU runner
 register_cuda_ci(est_time=5, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 
 import time
 import unittest
diff --git a/test/registered/attention/test_torch_native_attention_backend.py b/test/registered/attention/test_torch_native_attention_backend.py
index 2e6bc256d..5abbef881 100644
--- a/test/registered/attention/test_torch_native_attention_backend.py
+++ b/test/registered/attention/test_torch_native_attention_backend.py
@@ -19,6 +19,7 @@ from sglang.test.test_utils import (
 
 # Torch native attention backend integration test with MMLU eval
 register_cuda_ci(est_time=169, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=120, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=150, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/constrained_decoding/test_constrained_decoding.py b/test/registered/constrained_decoding/test_constrained_decoding.py
index 20f8e7388..899080aae 100644
--- a/test/registered/constrained_decoding/test_constrained_decoding.py
+++ b/test/registered/constrained_decoding/test_constrained_decoding.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=111, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=76, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=179, suite="stage-b-test-small-1-gpu-amd")
 
 import unittest
diff --git a/test/registered/core/test_hidden_states.py b/test/registered/core/test_hidden_states.py
index 0933df815..3f18fddf8 100644
--- a/test/registered/core/test_hidden_states.py
+++ b/test/registered/core/test_hidden_states.py
@@ -8,6 +8,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST, CustomTestCase
 
 register_cuda_ci(est_time=55, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=31, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestHiddenState(CustomTestCase):
diff --git a/test/registered/core/test_input_embeddings.py b/test/registered/core/test_input_embeddings.py
index bf80328f8..0ad440f07 100644
--- a/test/registered/core/test_input_embeddings.py
+++ b/test/registered/core/test_input_embeddings.py
@@ -17,6 +17,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=38, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=40, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=38, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/core/test_model_hooks.py b/test/registered/core/test_model_hooks.py
index ae14a2abd..5caaeceb6 100644
--- a/test/registered/core/test_model_hooks.py
+++ b/test/registered/core/test_model_hooks.py
@@ -10,6 +10,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=6, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=5, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
 
 HOOK_CALLS = []
diff --git a/test/registered/core/test_page_size.py b/test/registered/core/test_page_size.py
index b28bd3c7d..2b5c8c69e 100644
--- a/test/registered/core/test_page_size.py
+++ b/test/registered/core/test_page_size.py
@@ -14,6 +14,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=60, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=62, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=60, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/core/test_request_queue_validation.py b/test/registered/core/test_request_queue_validation.py
index c76e006ca..21a1b5d3a 100644
--- a/test/registered/core/test_request_queue_validation.py
+++ b/test/registered/core/test_request_queue_validation.py
@@ -18,6 +18,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=47, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=49, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=70, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/core/test_server_args.py b/test/registered/core/test_server_args.py
index 91e90a3e4..2df9f730a 100644
--- a/test/registered/core/test_server_args.py
+++ b/test/registered/core/test_server_args.py
@@ -7,6 +7,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=1, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/core/test_srt_endpoint.py b/test/registered/core/test_srt_endpoint.py
index 92aeff84c..4b9ee0800 100644
--- a/test/registered/core/test_srt_endpoint.py
+++ b/test/registered/core/test_srt_endpoint.py
@@ -28,6 +28,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=127, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=122, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=130, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/debug_utils/test_tensor_dump_forward_hook.py b/test/registered/debug_utils/test_tensor_dump_forward_hook.py
index 0cbc43627..c66f0a6fe 100644
--- a/test/registered/debug_utils/test_tensor_dump_forward_hook.py
+++ b/test/registered/debug_utils/test_tensor_dump_forward_hook.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=15, suite="stage-b-test-small-1-gpu-amd")
 
 import unittest
diff --git a/test/registered/hicache/test_hicache_storage.py b/test/registered/hicache/test_hicache_storage.py
index 75da6e138..801c6c2b5 100644
--- a/test/registered/hicache/test_hicache_storage.py
+++ b/test/registered/hicache/test_hicache_storage.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=96, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=96, suite="stage-b-test-small-1-gpu-5090")
 
 import time
 import unittest
diff --git a/test/registered/layers/mamba/test_causal_conv1d.py b/test/registered/layers/mamba/test_causal_conv1d.py
index a4a65b799..c50384f16 100644
--- a/test/registered/layers/mamba/test_causal_conv1d.py
+++ b/test/registered/layers/mamba/test_causal_conv1d.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=25, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=3, suite="stage-b-test-small-1-gpu-5090")
 
 # Adapted from https://github.com/vllm-project/vllm/blob/main/tests/kernels/mamba/test_causal_conv1d.py
 
diff --git a/test/registered/layers/mamba/test_mamba_ssm.py b/test/registered/layers/mamba/test_mamba_ssm.py
index a2372a16b..0eac3f597 100644
--- a/test/registered/layers/mamba/test_mamba_ssm.py
+++ b/test/registered/layers/mamba/test_mamba_ssm.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=5, suite="stage-b-test-small-1-gpu-5090")
 
 # Adapted from https://github.com/vllm-project/vllm/blob/633f943e30a4444d890d26b81850f7217736f840/tests/kernels/mamba/test_mamba_ssm_ssd.py
 
diff --git a/test/registered/layers/mamba/test_mamba_ssm_ssd.py b/test/registered/layers/mamba/test_mamba_ssm_ssd.py
index d7d42be1e..727b6ae49 100644
--- a/test/registered/layers/mamba/test_mamba_ssm_ssd.py
+++ b/test/registered/layers/mamba/test_mamba_ssm_ssd.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=13, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu-5090")
 
 # Adapted from https://github.com/vllm-project/vllm/blob/633f943e30a4444d890d26b81850f7217736f840/tests/kernels/mamba/test_mamba_ssm_ssd.py
 
diff --git a/test/registered/lora/test_lora_backend.py b/test/registered/lora/test_lora_backend.py
index 0cc2aca9b..e4447905d 100644
--- a/test/registered/lora/test_lora_backend.py
+++ b/test/registered/lora/test_lora_backend.py
@@ -30,6 +30,7 @@ from sglang.test.lora_utils import (
 from sglang.test.test_utils import CustomTestCase, is_in_ci
 
 register_cuda_ci(est_time=200, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=137, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(
     est_time=200,
     suite="stage-b-test-small-1-gpu-amd",
diff --git a/test/registered/lora/test_lora_eviction.py b/test/registered/lora/test_lora_eviction.py
index 4404d11ec..4b8505437 100644
--- a/test/registered/lora/test_lora_eviction.py
+++ b/test/registered/lora/test_lora_eviction.py
@@ -24,6 +24,7 @@ from sglang.test.runners import SRTRunner
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=224, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=215, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=224, suite="stage-b-test-small-1-gpu-amd")
 
 PROMPTS = [
diff --git a/test/registered/lora/test_lora_hf_sgl_logprob_diff.py b/test/registered/lora/test_lora_hf_sgl_logprob_diff.py
index ddb671373..e8105ff56 100644
--- a/test/registered/lora/test_lora_hf_sgl_logprob_diff.py
+++ b/test/registered/lora/test_lora_hf_sgl_logprob_diff.py
@@ -41,6 +41,10 @@ register_cuda_ci(
     est_time=150,
     suite="stage-b-test-small-1-gpu",
 )
+register_cuda_ci(
+    est_time=76,
+    suite="stage-b-test-small-1-gpu-5090",
+)
 
 from sglang.test.test_utils import DEFAULT_PORT_FOR_SRT_TEST_RUNNER, CustomTestCase
 
diff --git a/test/registered/metrics/test_metrics.py b/test/registered/metrics/test_metrics.py
index 9b0f9eee8..d1d47bee7 100644
--- a/test/registered/metrics/test_metrics.py
+++ b/test/registered/metrics/test_metrics.py
@@ -6,6 +6,7 @@ import requests
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=32, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=29, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=32, suite="stage-b-test-small-1-gpu-amd")
 from prometheus_client.parser import text_string_to_metric_families
 from prometheus_client.samples import Sample
diff --git a/test/registered/model_loading/test_external_models.py b/test/registered/model_loading/test_external_models.py
index 3dbfcecf8..7aac71fef 100644
--- a/test/registered/model_loading/test_external_models.py
+++ b/test/registered/model_loading/test_external_models.py
@@ -6,6 +6,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=30, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=26, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=45, suite="stage-b-test-small-1-gpu")
 
 
diff --git a/test/registered/model_loading/test_modelopt_loader.py b/test/registered/model_loading/test_modelopt_loader.py
index 96aaffa95..af61eeceb 100644
--- a/test/registered/model_loading/test_modelopt_loader.py
+++ b/test/registered/model_loading/test_modelopt_loader.py
@@ -26,6 +26,7 @@ CALIBRATION_NUM_SAMPLES = 512
 DEFAULT_DEVICE = "cuda:0"
 
 register_cuda_ci(est_time=11, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestModelOptModelLoader(CustomTestCase):
diff --git a/test/registered/models/test_cross_encoder_models.py b/test/registered/models/test_cross_encoder_models.py
index cc60d5d4a..5b8ac32ef 100644
--- a/test/registered/models/test_cross_encoder_models.py
+++ b/test/registered/models/test_cross_encoder_models.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Cross encoder model tests
 register_cuda_ci(est_time=100, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=105, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=150, suite="stage-b-test-small-1-gpu-amd")
 
 import multiprocessing as mp
diff --git a/test/registered/models/test_embedding_models.py b/test/registered/models/test_embedding_models.py
index 03ebe8a12..9541d2cc0 100644
--- a/test/registered/models/test_embedding_models.py
+++ b/test/registered/models/test_embedding_models.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Embedding model tests
 register_cuda_ci(est_time=73, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=58, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(
     est_time=73,
     suite="stage-b-test-small-1-gpu-amd",
diff --git a/test/registered/models/test_encoder_embedding_models.py b/test/registered/models/test_encoder_embedding_models.py
index 1bfb7c7aa..d922c47be 100644
--- a/test/registered/models/test_encoder_embedding_models.py
+++ b/test/registered/models/test_encoder_embedding_models.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 
 # Encoder embedding model tests (CUDA only)
 register_cuda_ci(est_time=270, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=144, suite="stage-b-test-small-1-gpu-5090")
 
 # Copyright 2023-2024 SGLang Team
 # Licensed under the Apache License, Version 2.0 (the "License");
diff --git a/test/registered/models/test_qwen_models.py b/test/registered/models/test_qwen_models.py
index cdfa4c06f..1bcb4914d 100644
--- a/test/registered/models/test_qwen_models.py
+++ b/test/registered/models/test_qwen_models.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Qwen model tests
 register_cuda_ci(est_time=90, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=84, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=130, suite="stage-b-test-small-1-gpu-amd")
 
 import unittest
diff --git a/test/registered/models/test_reward_models.py b/test/registered/models/test_reward_models.py
index 2f615d272..5c357b2ff 100644
--- a/test/registered/models/test_reward_models.py
+++ b/test/registered/models/test_reward_models.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Reward model tests
 register_cuda_ci(est_time=103, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=69, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=132, suite="stage-b-test-small-1-gpu-amd")
 
 # Copyright 2023-2024 SGLang Team
diff --git a/test/registered/models/test_transformers_models.py b/test/registered/models/test_transformers_models.py
index cf570a14b..5358ebaf8 100644
--- a/test/registered/models/test_transformers_models.py
+++ b/test/registered/models/test_transformers_models.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Transformers fallback model tests
 register_cuda_ci(est_time=245, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=210, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=320, suite="stage-b-test-small-1-gpu-amd")
 
 import dataclasses
diff --git a/test/registered/openai_server/basic/test_openai_embedding.py b/test/registered/openai_server/basic/test_openai_embedding.py
index 1432a5307..0b49a2d54 100644
--- a/test/registered/openai_server/basic/test_openai_embedding.py
+++ b/test/registered/openai_server/basic/test_openai_embedding.py
@@ -14,6 +14,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=70, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=48, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=141, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_openai_server.py b/test/registered/openai_server/basic/test_openai_server.py
index 3f27fb500..35d628c77 100644
--- a/test/registered/openai_server/basic/test_openai_server.py
+++ b/test/registered/openai_server/basic/test_openai_server.py
@@ -29,6 +29,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=184, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=130, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=149, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_protocol.py b/test/registered/openai_server/basic/test_protocol.py
index 47bf56381..561ebbdfa 100644
--- a/test/registered/openai_server/basic/test_protocol.py
+++ b/test/registered/openai_server/basic/test_protocol.py
@@ -31,6 +31,7 @@ from sglang.srt.entrypoints.openai.protocol import (
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=3, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=2, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_serving_chat.py b/test/registered/openai_server/basic/test_serving_chat.py
index d81f2efb0..4764edb6b 100644
--- a/test/registered/openai_server/basic/test_serving_chat.py
+++ b/test/registered/openai_server/basic/test_serving_chat.py
@@ -24,6 +24,7 @@ from sglang.srt.utils import get_or_create_event_loop
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/basic/test_serving_completions.py b/test/registered/openai_server/basic/test_serving_completions.py
index 34c262e60..bbc3c5c6e 100644
--- a/test/registered/openai_server/basic/test_serving_completions.py
+++ b/test/registered/openai_server/basic/test_serving_completions.py
@@ -14,6 +14,7 @@ from sglang.srt.managers.tokenizer_manager import TokenizerManager
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/features/test_json_mode.py b/test/registered/openai_server/features/test_json_mode.py
index 946d8e0ff..112b08b3c 100644
--- a/test/registered/openai_server/features/test_json_mode.py
+++ b/test/registered/openai_server/features/test_json_mode.py
@@ -14,6 +14,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=109, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=84, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=180, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/features/test_openai_server_ebnf.py b/test/registered/openai_server/features/test_openai_server_ebnf.py
index 165bbcf78..f024d0455 100644
--- a/test/registered/openai_server/features/test_openai_server_ebnf.py
+++ b/test/registered/openai_server/features/test_openai_server_ebnf.py
@@ -14,6 +14,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=27, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=20, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/features/test_openai_server_hidden_states.py b/test/registered/openai_server/features/test_openai_server_hidden_states.py
index 0ac3d3265..8218fedfd 100644
--- a/test/registered/openai_server/features/test_openai_server_hidden_states.py
+++ b/test/registered/openai_server/features/test_openai_server_hidden_states.py
@@ -17,6 +17,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=186, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=134, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(
     est_time=186,
     suite="stage-b-test-small-1-gpu-amd",
diff --git a/test/registered/openai_server/features/test_reasoning_content.py b/test/registered/openai_server/features/test_reasoning_content.py
index 2c59921fe..ad80d02a0 100644
--- a/test/registered/openai_server/features/test_reasoning_content.py
+++ b/test/registered/openai_server/features/test_reasoning_content.py
@@ -24,6 +24,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=89, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=77, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=89, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/function_call/test_tool_choice.py b/test/registered/openai_server/function_call/test_tool_choice.py
index b12cd70d0..77e3825f1 100644
--- a/test/registered/openai_server/function_call/test_tool_choice.py
+++ b/test/registered/openai_server/function_call/test_tool_choice.py
@@ -23,6 +23,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=120, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=109, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=258, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/validation/test_matched_stop.py b/test/registered/openai_server/validation/test_matched_stop.py
index aa890e363..88f264d99 100644
--- a/test/registered/openai_server/validation/test_matched_stop.py
+++ b/test/registered/openai_server/validation/test_matched_stop.py
@@ -12,6 +12,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=40, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=39, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=60, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/openai_server/validation/test_openai_server_ignore_eos.py b/test/registered/openai_server/validation/test_openai_server_ignore_eos.py
index bbf696d41..891f67855 100644
--- a/test/registered/openai_server/validation/test_openai_server_ignore_eos.py
+++ b/test/registered/openai_server/validation/test_openai_server_ignore_eos.py
@@ -12,6 +12,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=6, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=37, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=47, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/ops/test_repeat_interleave.py b/test/registered/ops/test_repeat_interleave.py
index 5337281a1..8a9b97445 100644
--- a/test/registered/ops/test_repeat_interleave.py
+++ b/test/registered/ops/test_repeat_interleave.py
@@ -2,6 +2,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # Ops - Repeat Interleave tests (1-GPU)
 register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=75, suite="stage-b-test-small-1-gpu-amd")
 
 import time
diff --git a/test/registered/profiling/test_profile_merger.py b/test/registered/profiling/test_profile_merger.py
index 412ae8046..4f52dcfe9 100644
--- a/test/registered/profiling/test_profile_merger.py
+++ b/test/registered/profiling/test_profile_merger.py
@@ -18,6 +18,7 @@ from sglang.srt.utils.profile_merger import ProfileMerger
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=8, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/profiling/test_profile_merger_http_api.py b/test/registered/profiling/test_profile_merger_http_api.py
index 9bf656f26..bf9d38c32 100644
--- a/test/registered/profiling/test_profile_merger_http_api.py
+++ b/test/registered/profiling/test_profile_merger_http_api.py
@@ -5,6 +5,7 @@ from sglang.srt.managers.io_struct import ProfileReqInput
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=6, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=9, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/profiling/test_start_profile.py b/test/registered/profiling/test_start_profile.py
index ddd19b6bb..a59316374 100644
--- a/test/registered/profiling/test_start_profile.py
+++ b/test/registered/profiling/test_start_profile.py
@@ -30,6 +30,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=41, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=39, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=60, suite="stage-b-test-small-1-gpu-amd")
 
 OUTPUT_DIR = "./profiler_dir"
diff --git a/test/registered/quant/test_autoround.py b/test/registered/quant/test_autoround.py
index df69f1e5c..d7000b702 100644
--- a/test/registered/quant/test_autoround.py
+++ b/test/registered/quant/test_autoround.py
@@ -18,6 +18,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=77, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=76, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestAutoRound(CustomTestCase):
diff --git a/test/registered/quant/test_block_int8.py b/test/registered/quant/test_block_int8.py
index 91f2cd474..d090e5443 100644
--- a/test/registered/quant/test_block_int8.py
+++ b/test/registered/quant/test_block_int8.py
@@ -11,6 +11,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=44, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=25, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=22, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/quant/test_int8_kernel.py b/test/registered/quant/test_int8_kernel.py
index d45f82efb..6e2726a72 100644
--- a/test/registered/quant/test_int8_kernel.py
+++ b/test/registered/quant/test_int8_kernel.py
@@ -12,6 +12,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu-5090")
 
 
 def native_w8a8_per_token_matmul(A, B, As, Bs, output_dtype=torch.float16):
diff --git a/test/registered/quant/test_torchao.py b/test/registered/quant/test_torchao.py
index bc2c93884..30480f894 100644
--- a/test/registered/quant/test_torchao.py
+++ b/test/registered/quant/test_torchao.py
@@ -7,6 +7,7 @@ from sglang import Engine
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=103, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=88, suite="stage-b-test-small-1-gpu-5090")
 from sglang.lang.chat_template import get_chat_template_by_model_path
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
diff --git a/test/registered/quant/test_triton_scaled_mm.py b/test/registered/quant/test_triton_scaled_mm.py
index 35a30d710..512191ee6 100644
--- a/test/registered/quant/test_triton_scaled_mm.py
+++ b/test/registered/quant/test_triton_scaled_mm.py
@@ -9,6 +9,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=12, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/rl/test_fp32_lm_head.py b/test/registered/rl/test_fp32_lm_head.py
index 4d4acee9a..b59698a6c 100644
--- a/test/registered/rl/test_fp32_lm_head.py
+++ b/test/registered/rl/test_fp32_lm_head.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=15, suite="stage-b-test-small-1-gpu-amd")
 
 import unittest
diff --git a/test/registered/rl/test_lora_load_from_tensor.py b/test/registered/rl/test_lora_load_from_tensor.py
index c656e2f8c..a5633a73a 100644
--- a/test/registered/rl/test_lora_load_from_tensor.py
+++ b/test/registered/rl/test_lora_load_from_tensor.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=90, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=54, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=90, suite="stage-b-test-small-1-gpu")
 
 import json
diff --git a/test/registered/rl/test_update_weights_from_tensor.py b/test/registered/rl/test_update_weights_from_tensor.py
index 9faff8e26..7a37c9195 100644
--- a/test/registered/rl/test_update_weights_from_tensor.py
+++ b/test/registered/rl/test_update_weights_from_tensor.py
@@ -1,6 +1,7 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=195, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=164, suite="stage-b-test-small-1-gpu-5090")
 
 import gc
 import json
diff --git a/test/registered/sampling/test_original_logprobs.py b/test/registered/sampling/test_original_logprobs.py
index 9093dea51..3056f4a2a 100644
--- a/test/registered/sampling/test_original_logprobs.py
+++ b/test/registered/sampling/test_original_logprobs.py
@@ -18,6 +18,7 @@ import os
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=41, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=43, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=60, suite="stage-b-test-small-1-gpu-amd")
 import random
 import unittest
diff --git a/test/registered/sampling/test_penalty.py b/test/registered/sampling/test_penalty.py
index d73b5d62a..d4f84cd69 100644
--- a/test/registered/sampling/test_penalty.py
+++ b/test/registered/sampling/test_penalty.py
@@ -10,6 +10,7 @@ from sglang.srt.utils import kill_process_tree
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=82, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=44, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=82, suite="stage-b-test-small-1-gpu-amd")
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
diff --git a/test/registered/sampling/test_pytorch_sampling_backend.py b/test/registered/sampling/test_pytorch_sampling_backend.py
index 51bf254a6..e98e8d9bd 100644
--- a/test/registered/sampling/test_pytorch_sampling_backend.py
+++ b/test/registered/sampling/test_pytorch_sampling_backend.py
@@ -7,6 +7,7 @@ from sglang.srt.utils import kill_process_tree
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 register_cuda_ci(est_time=66, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=64, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=66, suite="stage-b-test-small-1-gpu-amd")
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
diff --git a/test/registered/scheduler/test_abort.py b/test/registered/scheduler/test_abort.py
index b8e6b691d..c2b3e8c2d 100644
--- a/test/registered/scheduler/test_abort.py
+++ b/test/registered/scheduler/test_abort.py
@@ -18,6 +18,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=131, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=119, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=51, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/scheduler/test_chunked_prefill.py b/test/registered/scheduler/test_chunked_prefill.py
index 951b09f74..336200a20 100644
--- a/test/registered/scheduler/test_chunked_prefill.py
+++ b/test/registered/scheduler/test_chunked_prefill.py
@@ -8,6 +8,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase, run_mmlu_test, run_mulit_request_test
 
 register_cuda_ci(est_time=312, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=323, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=312, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/scheduler/test_prefill_adder.py b/test/registered/scheduler/test_prefill_adder.py
index f11b0b16f..587ddb8e0 100644
--- a/test/registered/scheduler/test_prefill_adder.py
+++ b/test/registered/scheduler/test_prefill_adder.py
@@ -8,6 +8,7 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=1, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=6, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=2, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/scheduler/test_priority_scheduling.py b/test/registered/scheduler/test_priority_scheduling.py
index b3ba23ea0..fb998fd71 100644
--- a/test/registered/scheduler/test_priority_scheduling.py
+++ b/test/registered/scheduler/test_priority_scheduling.py
@@ -18,6 +18,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=130, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=146, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=195, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/scheduler/test_retract_decode.py b/test/registered/scheduler/test_retract_decode.py
index 80dc74cec..c1df809f8 100644
--- a/test/registered/scheduler/test_retract_decode.py
+++ b/test/registered/scheduler/test_retract_decode.py
@@ -18,6 +18,7 @@ from sglang.test.test_utils import (
 from sglang.utils import is_in_ci
 
 register_cuda_ci(est_time=311, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=400, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=450, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/spec/eagle/test_eagle3_basic.py b/test/registered/spec/eagle/test_eagle3_basic.py
index b60b167d0..a346214ad 100644
--- a/test/registered/spec/eagle/test_eagle3_basic.py
+++ b/test/registered/spec/eagle/test_eagle3_basic.py
@@ -12,6 +12,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=50, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=87, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestEagle3Basic(EagleServerBase):
diff --git a/test/registered/spec/eagle/test_eagle_infer_beta.py b/test/registered/spec/eagle/test_eagle_infer_beta.py
index c67ff334a..fe488e99a 100644
--- a/test/registered/spec/eagle/test_eagle_infer_beta.py
+++ b/test/registered/spec/eagle/test_eagle_infer_beta.py
@@ -17,6 +17,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=283, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=404, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestEagleServerBase(CustomTestCase, MatchedStopMixin):
diff --git a/test/registered/spec/utils/test_build_eagle_tree.py b/test/registered/spec/utils/test_build_eagle_tree.py
index 3f5c71125..32dc413c9 100644
--- a/test/registered/spec/utils/test_build_eagle_tree.py
+++ b/test/registered/spec/utils/test_build_eagle_tree.py
@@ -9,6 +9,7 @@ from sglang.srt.speculative.eagle_utils import (
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=3, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=2, suite="stage-b-test-small-1-gpu-5090")
 
 
 class TestBuildEagleTree(unittest.TestCase):
diff --git a/test/registered/tokenizer/test_skip_tokenizer_init.py b/test/registered/tokenizer/test_skip_tokenizer_init.py
index 7d95c19cf..c00a655cf 100644
--- a/test/registered/tokenizer/test_skip_tokenizer_init.py
+++ b/test/registered/tokenizer/test_skip_tokenizer_init.py
@@ -24,6 +24,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=77, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=52, suite="stage-b-test-small-1-gpu-5090")
 register_amd_ci(est_time=117, suite="stage-b-test-small-1-gpu-amd")
 
 
diff --git a/test/registered/vlm/test_evs.py b/test/registered/vlm/test_evs.py
index 0b6742a71..78592245e 100644
--- a/test/registered/vlm/test_evs.py
+++ b/test/registered/vlm/test_evs.py
@@ -7,6 +7,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import run_doctests
 
 register_cuda_ci(est_time=20, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu-5090")
 
 
 def test_resolve_evs_config():
diff --git a/test/run_suite.py b/test/run_suite.py
index 1a687b0f8..4cd75678d 100644
--- a/test/run_suite.py
+++ b/test/run_suite.py
@@ -27,6 +27,7 @@ PER_COMMIT_SUITES = {
     HWBackend.CUDA: [
         "stage-a-test-1",
         "stage-b-test-small-1-gpu",
+        "stage-b-test-small-1-gpu-5090",
         "stage-b-test-large-1-gpu",
         "stage-b-test-large-2-gpu",
         "stage-c-test-large-4-gpu",