diff --git a/src/backends/neon/NeonLayerSupport.cpp b/src/backends/neon/NeonLayerSupport.cpp
index 165e067..28c4b75 100644
--- a/src/backends/neon/NeonLayerSupport.cpp
+++ b/src/backends/neon/NeonLayerSupport.cpp
@@ -453,38 +453,4 @@
                                    output);
 }
 
-bool IsNeonDirectConvolutionPreferred(const TensorInfo& weightInfo, const Convolution2dDescriptor& desc)
-{
-    // See arm_compute::NEDirectConvolutionLayer documentation for the supported cases,
-    // and complement with NEDirectConvolutionLayerKernel::configure() implementation.
-
-    // Only 1x1 is using direct convolution. Performance results and details are in:
-    //    https://jira.arm.com/browse/IVGCVSW-1003
-    // Measurements were taken as of clframework: f105ab972135bcd21304883eff040d7e587099bc
-
-    const bool dataTypeSupported = (weightInfo.GetDataType() == armnn::DataType::Float32);
-
-    // Strides: 1|2|3
-    const bool strideSupported = (desc.m_StrideX == 1 || desc.m_StrideX == 2 || desc.m_StrideX == 3) &&
-                                 (desc.m_StrideY == 1 || desc.m_StrideY == 2 || desc.m_StrideY == 3);
-
-    auto paddingLargerThan = [](const Convolution2dDescriptor& conv2ddesc, unsigned int value)
-    {
-        return conv2ddesc.m_PadLeft > value || conv2ddesc.m_PadRight > value ||
-               conv2ddesc.m_PadTop > value || conv2ddesc.m_PadBottom > value;
-    };
-
-    // Supported sizes and padding.
-    const bool sizeAndPaddingSupported =
-        // Pad > 0 not supported for 1x1 weights.
-        (weightInfo.GetShape()[2] == 1 && weightInfo.GetShape()[3] == 1 && !paddingLargerThan(desc, 0u));
-
-    const bool preferDirectConvolution = dataTypeSupported &&
-                                         strideSupported &&
-                                         sizeAndPaddingSupported &&
-                                         // NEDirectConvolutionLayerKernel doesn't support NULL bias.
-                                         desc.m_BiasEnabled;
-    return preferDirectConvolution;
-}
-
 } // namespace armnn
diff --git a/src/backends/neon/NeonLayerSupport.hpp b/src/backends/neon/NeonLayerSupport.hpp
index 5e80ab8..fb1567c 100644
--- a/src/backends/neon/NeonLayerSupport.hpp
+++ b/src/backends/neon/NeonLayerSupport.hpp
@@ -169,6 +169,4 @@
                                 Optional<std::string&> reasonIfUnsupported = EmptyOptional()) const override;
 }; // class NeonLayerSupport
 
-bool IsNeonDirectConvolutionPreferred(const TensorInfo& weightInfo, const Convolution2dDescriptor& desc);
-
 } // namespace armnn
diff --git a/src/backends/neon/test/NeonLayerTests.cpp b/src/backends/neon/test/NeonLayerTests.cpp
index d4bfa17..f0410f2 100644
--- a/src/backends/neon/test/NeonLayerTests.cpp
+++ b/src/backends/neon/test/NeonLayerTests.cpp
@@ -44,54 +44,6 @@
                      armnn::DataLayout::NHWC)
 
 ARMNN_AUTO_TEST_CASE(SimpleConvolution2dSquareNhwc, SimpleConvolution2d3x3NhwcTest, false)
-namespace
-{
-
-armnn::Convolution2dDescriptor MakeConv2dDesc(uint32_t strideX, uint32_t strideY,
-    uint32_t padLeft = 0, uint32_t padRight = 0, uint32_t padTop = 0, uint32_t padBottom = 0)
-{
-    armnn::Convolution2dDescriptor result;
-    result.m_StrideX = strideX;
-    result.m_StrideY = strideY;
-    result.m_PadLeft = padLeft;
-    result.m_PadRight = padRight;
-    result.m_PadTop = padTop;
-    result.m_PadBottom = padBottom;
-    result.m_BiasEnabled = true;
-    return result;
-}
-
-}
-
-BOOST_AUTO_TEST_CASE(Conv2dUtils)
-{
-    // The only preferred Neon convolution is 1x1 with padding=0 and stride size {1,2,3}.
-    armnn::TensorShape shape1x1({ 1,1,1,1 });
-    armnn::TensorInfo info1x1(shape1x1, armnn::DataType::Float32);
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(1, 1)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(1, 2)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(1, 3)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(2, 1)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(2, 2)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(2, 3)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(3, 1)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(3, 2)));
-    BOOST_TEST(armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(3, 3)));
-
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(4, 1)));
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(4, 5)));
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(3, 6)));
-
-    // non zero padding is not preferred for direct convolution
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(1, 1, 1, 0)));
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(1, 1, 0, 1)));
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info1x1, MakeConv2dDesc(1, 1, 1, 1)));
-
-    // 2x2 filter not preferred for direct convolution
-    armnn::TensorShape shape2x2({ 1,1,2,2 });
-    armnn::TensorInfo info2x2(shape2x2, armnn::DataType::Float32);
-    BOOST_TEST(!armnn::IsNeonDirectConvolutionPreferred(info2x2, MakeConv2dDesc(1, 1)));
-}
 
 // Depthwise Convolution
 ARMNN_AUTO_TEST_CASE(DepthwiseConvolution2dDepthMul1,
diff --git a/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp b/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp
index 781c91a..151132f 100644
--- a/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp
+++ b/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp
@@ -82,30 +82,14 @@
                                              m_Data.m_Parameters.m_PadBottom,
                                              arm_compute::DimensionRoundingType::FLOOR);
 
-    const bool preferDirectConvolution =
-        IsNeonDirectConvolutionPreferred(m_Data.m_Weight->GetTensorInfo(),
-                                         m_Data.m_Parameters);
+    auto convolutionLayer = std::make_unique<arm_compute::NEConvolutionLayer>(memoryManager);
+    convolutionLayer->configure(&input,
+                                m_KernelTensor.get(),
+                                m_BiasTensor.get(),
+                                &output,
+                                padStrideInfo);
+    m_ConvolutionLayer.reset(convolutionLayer.release());
 
-    if (preferDirectConvolution)
-    {
-        auto directConvolutionLayer = std::make_unique<arm_compute::NEDirectConvolutionLayer>(memoryManager);
-        directConvolutionLayer->configure(&input,
-                                          m_KernelTensor.get(),
-                                          m_BiasTensor.get(),
-                                          &output,
-                                          padStrideInfo);
-        m_ConvolutionLayer.reset(directConvolutionLayer.release());
-    }
-    else
-    {
-        auto convolutionLayer = std::make_unique<arm_compute::NEConvolutionLayer>(memoryManager);
-        convolutionLayer->configure(&input,
-                                    m_KernelTensor.get(),
-                                    m_BiasTensor.get(),
-                                    &output,
-                                    padStrideInfo);
-        m_ConvolutionLayer.reset(convolutionLayer.release());
-    }
     BOOST_ASSERT(m_ConvolutionLayer);
 
     InitializeArmComputeTensorData(*m_KernelTensor, m_Data.m_Weight);