diff --git a/ci/matrix.yaml b/ci/matrix.yaml
index 881f553f65d..618e6832856 100644
--- a/ci/matrix.yaml
+++ b/ci/matrix.yaml
@@ -8,7 +8,7 @@ workflows:
   #   - {jobs: ['test'], project: 'thrust', std: 17, ctk: 'curr', cxx: ['gcc12', 'clang16']}
   #
   override:
-
+    - {jobs: ['build'], std: [17, 20], ctk: '12.5', cxx: 'nvhpc'}
   pull_request:
     # Old CTK/compiler
     - {jobs: ['build'], std: 'minmax', ctk: '12.0', cxx: ['gcc7', 'gcc9', 'clang14', 'msvc2019']}
diff --git a/cub/cub/detail/detect_cuda_runtime.cuh b/cub/cub/detail/detect_cuda_runtime.cuh
index 7666f9b2d23..cc00f847d50 100644
--- a/cub/cub/detail/detect_cuda_runtime.cuh
+++ b/cub/cub/detail/detect_cuda_runtime.cuh
@@ -46,6 +46,12 @@
 
 // CUDA headers might not be present when using NVRTC, see NVIDIA/cccl#2095 for detail
 #if !_CCCL_COMPILER(NVRTC)
+_CCCL_SUPPRESS_DEPRECATED_PUSH
+#  ifdef __CUDA_RUNTIME_H__
+#    error cuda_runtime.h was already included
+#  endif
+#  include <cuda_runtime.h>
+_CCCL_SUPPRESS_DEPRECATED_POP
 #  include <cuda_runtime_api.h>
 #endif // !_CCCL_COMPILER(NVRTC)