diff --git a/cmake/CMakeLists.txt b/cmake/CMakeLists.txt
index 962b42c190..c1a171fba0 100644
--- a/cmake/CMakeLists.txt
+++ b/cmake/CMakeLists.txt
@@ -1517,35 +1517,33 @@ if (onnxruntime_USE_CUDA)
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_53,code=sm_53") # TX1, Nano
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_62,code=sm_62") # TX2
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_72,code=sm_72") # AGX Xavier, NX Xavier
-      if (CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 11)
-        set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_87,code=sm_87") # AGX Orin, NX Orin
-      endif()
+      set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_87,code=sm_87") # AGX Orin, NX Orin
     else()
-      # the following compute capabilities are removed in CUDA 11 Toolkit
-      if (CMAKE_CUDA_COMPILER_VERSION VERSION_LESS 11)
-        set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_30,code=sm_30") # K series
-      endif()
       if (CMAKE_CUDA_COMPILER_VERSION VERSION_LESS 12)
         # 37, 50 still work in CUDA 11 but are marked deprecated and will be removed in future CUDA version.
         set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_37,code=sm_37") # K80
         set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_50,code=sm_50") # M series
       endif()
+      # Note that we generate SASS code for specified cuda architectures. It does not support forward compatibility.
+      # To add PTX for future GPU architectures >= XX, append -gencode=arch=compute_XX,code=compute_XX.
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_52,code=sm_52") # M60
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_60,code=sm_60") # P series
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_70,code=sm_70") # V series
       set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_75,code=sm_75") # T series
-      if (CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 11)
-        set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_80,code=sm_80") # A series
-      endif()
+      set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_80,code=sm_80") # A series
+      set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_86,code=sm_86")
+      set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_89,code=sm_89")
       if (CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12)
         set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_90,code=sm_90") # H series
+        if (CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.8)
+          set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -gencode=arch=compute_120,code=sm_120") # B series
+        endif()
       endif()
     endif()
   endif()
+  set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xfatbin=-compress-all")
   set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} --expt-relaxed-constexpr")
-  if (CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 11)
-    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} --Werror default-stream-launch")
-  endif()
+  set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} --Werror default-stream-launch")
   if (NOT WIN32)
     list(APPEND CUDA_NVCC_FLAGS --compiler-options -fPIC)
   endif()