diff --git a/CMakeLists.txt b/CMakeLists.txt
index dfb5159ea12179b127d3780c8affdcfe5978f6db..0c40e3c6f7d1170443abe2a93c07f75527b18b49 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -51,13 +51,7 @@ option(ON_TRAVIS "Running test on travis-ci or not." OFF)
 option(ON_COVERALLS "Generating code coverage data on coveralls or not." OFF)
 option(COVERALLS_UPLOAD "Uploading the generated coveralls json." ON)
 
-if(NOT CMAKE_BUILD_TYPE)
-    set(CMAKE_BUILD_TYPE "RelWithDebInfo" CACHE STRING
-        "Choose the type of build, options are: Debug Release RelWithDebInfo MinSizeRel"
-        FORCE)
-endif()
 
-include(enableCXX11)
 include(cpplint)
 include(ccache)
 if(WITH_RDMA)
@@ -83,18 +77,13 @@ if(NOT WITH_GPU)
 
     list(APPEND CMAKE_CXX_SOURCE_FILE_EXTENSIONS cu)
 else()
-    if(${CUDA_VERSION_MAJOR} GREATER 6)
-        if(COMPILER_SUPPORT_CXX11)
-            LIST(APPEND CUDA_NVCC_FLAGS -std=c++11)
-        endif()
+    if(${CUDA_VERSION_MAJOR} VERSION_LESS 7)
+        message(FATAL_ERROR "Paddle need CUDA >= 7.0 to compile")
     endif()
 
-    # TODO(yuyang18): Change it to remove std=c++11 in cuda compile.
-    set(CUDA_PROPAGATE_HOST_FLAGS OFF)
     if(NOT CUDNN_FOUND)
         message(FATAL_ERROR "Paddle need cudnn to compile")
     endif()
-    set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-g -O3 --use_fast_math")
 
     if(WITH_AVX)
         set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler ${AVX_FLAG}")
diff --git a/cmake/enableCXX11.cmake b/cmake/enableCXX11.cmake
deleted file mode 100644
index dc8cc3371aa6e577676289750cc525ec4f1fb6b4..0000000000000000000000000000000000000000
--- a/cmake/enableCXX11.cmake
+++ /dev/null
@@ -1,13 +0,0 @@
-# Enable C++ 11 for GCC.
-# NOTE: It's only tested for gcc.
-include(CheckCXXCompilerFlag)
-CHECK_CXX_COMPILER_FLAG("-std=c++11" COMPILER_SUPPORT_CXX11)
-CHECK_CXX_COMPILER_FLAG("-std=c++0x" COMPILER_SUPPORT_CXX0X)
-
-if(COMPILER_SUPPORT_CXX11)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
-elseif(COMPILER_SUPPORT_CXX0X)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++0x")
-else()
-    message(FATAL_ERROR "Your compiler must support c++11")
-endif()
\ No newline at end of file
diff --git a/cmake/flags.cmake b/cmake/flags.cmake
index e087770991aefc17535d50c0539c50f6316520d7..0983d83b73a32d0615170155759d45001cc6ff54 100644
--- a/cmake/flags.cmake
+++ b/cmake/flags.cmake
@@ -2,6 +2,37 @@
 include(CheckCXXCompilerFlag)
 include(CheckCCompilerFlag)
 include(CheckCXXSymbolExists)
+
+if(NOT CMAKE_BUILD_TYPE)
+    set(CMAKE_BUILD_TYPE "RelWithDebInfo" CACHE STRING 
+        "Choose the type of build, options are: Debug Release RelWithDebInfo MinSizeRel"
+        FORCE)
+endif()
+
+function(CheckCompilerCXX11Flag)
+    if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+        if(${CMAKE_CXX_COMPILER_VERSION} VERSION_LESS 4.8)
+            message(FATAL_ERROR "Unsupported GCC version. GCC >= 4.8 required.")
+        endif()
+    elseif(CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang" OR CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+        # cmake >= 3.0 compiler id "AppleClang" on Mac OS X, otherwise "Clang"
+        # Apple Clang is a different compiler than upstream Clang which havs different version numbers.
+        # https://gist.github.com/yamaya/2924292
+        if(APPLE)  # cmake < 3.0 compiler id "Clang" on Mac OS X
+            if(${CMAKE_CXX_COMPILER_VERSION} VERSION_LESS 5.1)
+                message(FATAL_ERROR "Unsupported AppleClang version. AppleClang >= 5.1 required.")
+            endif()
+        else()
+            if (${CMAKE_CXX_COMPILER_VERSION} VERSION_LESS 3.3)
+                message(FATAL_ERROR "Unsupported Clang version. Clang >= 3.3 required.")
+            endif()
+        endif()   
+    endif()
+endfunction()
+
+CheckCompilerCXX11Flag()
+LIST(APPEND CMAKE_CXX_FLAGS -std=c++11)
+
 # safe_set_flag
 #
 # Set a compile flag only if compiler is support
@@ -41,9 +72,7 @@ macro(safe_set_nvflag flag_name)
     CHECK_C_COMPILER_FLAG(${flag_name} C_COMPILER_SUPPORT_FLAG_${safe_name})
     set(safe_name C_COMPILER_SUPPORT_FLAG_${safe_name})
     if(${safe_name})
-        set(CUDA_NVCC_FLAGS
-            --compiler-options;${flag_name}
-            ${CUDA_NVCC_FLAGS})
+        LIST(APPEND CUDA_NVCC_FLAGS -Xcompiler ${flag_name})
     endif()
 endmacro()
 
@@ -109,8 +138,22 @@ foreach(flag ${GPU_COMMON_FLAGS})
 endforeach()
 
 
+set(CUDA_PROPAGATE_HOST_FLAGS OFF)
+
 # Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
 # So, don't set these flags here.
+LIST(APPEND CUDA_NVCC_FLAGS -std=c++11)
+LIST(APPEND CUDA_NVCC_FLAGS --use_fast_math)
+
+if(CMAKE_BUILD_TYPE  STREQUAL "Debug")
+    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_DEBUG})
+elseif(CMAKE_BUILD_TYPE  STREQUAL "Release")
+    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELEASE})
+elseif(CMAKE_BUILD_TYPE  STREQUAL "RelWithDebInfo")
+    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELWITHDEBINFO})
+elseif(CMAKE_BUILD_TYPE  STREQUAL "MinSizeRel")
+    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_MINSIZEREL})
+endif()
 
 function(specify_cuda_arch cuda_version cuda_arch)
     if(${cuda_version} VERSION_GREATER "8.0")
diff --git a/doc/getstarted/build_and_install/build_from_source.md b/doc/getstarted/build_and_install/build_from_source.md
index a2a96f6f48369bf80c8f7b5040121fd7dba5a77b..150d7fc43720314462ac5c5b72f6a93b18e6d735 100644
--- a/doc/getstarted/build_and_install/build_from_source.md
+++ b/doc/getstarted/build_and_install/build_from_source.md
@@ -16,13 +16,16 @@ git submodule update --init --recursive
 
 ## <span id="requirements">Requirements</span>
 
-To compile the source code, your computer must be equipped with GCC >=4.6 or Clang compiler.
-### Dependencies
+To compile the source code, your computer must be equipped with the following dependencies.
 
+- **Compiler**: GCC >= 4.8 or Clang >= 3.3 (AppleClang >= 5.1)
 - **CMake**: version >= 2.8
 - **BLAS**: MKL, OpenBlas or ATLAS
-- **protobuf**: version >= 2.4, **Note: 3.x is not supported**
-- **python**: only python 2.7 is supported currently
+- **Protocol Buffers**: version >= 2.4, **Note: 3.x is not supported**
+- **Python**: only python 2.7 is supported currently
+
+**Note:** For CUDA 7.0 and CUDA 7.5, GCC 5.0 and up are not supported!
+For CUDA 8.0, GCC versions later than 5.3 are not supported!
 
 ### Options
 
@@ -50,8 +53,8 @@ PaddlePaddle supports some build options. To enable it, first you need to instal
 </html>
 
 **Note:**
-  - The GPU version works best with Cuda Toolkit 7.5 and cuDNN v5.
-  - Other versions like Cuda Toolkit 6.5, 7.0, 8.0 and cuDNN v2, v3, v4 are also supported.
+  - The GPU version works best with Cuda Toolkit 8.0 and cuDNN v5.
+  - Other versions like Cuda Toolkit 7.0, 7.5 and cuDNN v3, v4 are also supported.
   - **To utilize cuDNN v5, Cuda Toolkit 7.5 is prerequisite and vice versa.**
 
 As a simple example, consider the following:  
diff --git a/paddle/cuda/CMakeLists.txt b/paddle/cuda/CMakeLists.txt
index 10fa34b92727b03f8219a721a60b623f74582ffa..aa1ff4a771c4a1c64be86893e7b2261ae65f0f94 100755
--- a/paddle/cuda/CMakeLists.txt
+++ b/paddle/cuda/CMakeLists.txt
@@ -34,9 +34,6 @@ else()
         src/hl_warpctc_wrap.cc)
 endif()
 
-set_source_files_properties(${AVX_SOURCES}
-                            PROPERTIES COMPILE_FLAGS "-mavx")
-
 set(CUDA_CU_SOURCES
     src/hl_perturbation_util.cu
     src/hl_cuda_aggregate.cu