diff --git a/src/layer/x86/convolution_im2col_gemm_int8.h b/src/layer/x86/convolution_im2col_gemm_int8.h
index 40d5e4453a2..6420c92103c 100644
--- a/src/layer/x86/convolution_im2col_gemm_int8.h
+++ b/src/layer/x86/convolution_im2col_gemm_int8.h
@@ -13,29 +13,23 @@
 // specific language governing permissions and limitations under the License.
 
 #if NCNN_RUNTIME_CPU && NCNN_AVX512VNNI && __AVX512F__ && !__AVX512VNNI__
-// int convolution_im2col_gemm_int8_avx512vnni(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt);
-
 void convolution_im2col_input_tile_int8_avx512vnni(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h);
 #endif
 
-#if NCNN_RUNTIME_CPU && NCNN_AVXVNNI && __AVX__ && !__AVXVNNI__ && !__AVX512VNNI__
-// int convolution_im2col_gemm_int8_avxvnni(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt);
+#if NCNN_RUNTIME_CPU && NCNN_AVXVNNIINT8 && __AVX__ && !__AVXVNNIINT8__ && !__AVX512VNNI__
+void convolution_im2col_input_tile_int8_avxvnniint8(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h);
+#endif
 
+#if NCNN_RUNTIME_CPU && NCNN_AVXVNNI && __AVX__ && !__AVXVNNI__ && !__AVX512VNNI__
 void convolution_im2col_input_tile_int8_avxvnni(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h);
 #endif
 
 #if NCNN_RUNTIME_CPU && NCNN_AVX2 && __AVX__ && !__AVX2__ && !__AVXVNNI__ && !__AVX512VNNI__
-// void convolution_im2col_gemm_transform_kernel_int8_avx2(const Mat& kernel, Mat& AT, int inch, int outch, int kernel_w, int kernel_h, const Option& opt);
-// int convolution_im2col_gemm_int8_avx2(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt);
-
 void convolution_im2col_input_tile_int8_avx2(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h);
-
 void unpack_output_tile_int32_avx2(const Mat& topT, Mat& top_blob, int i, int max_ii, int j, int max_jj);
-
 #endif
 
 #if NCNN_RUNTIME_CPU && NCNN_XOP && __SSE2__ && !__XOP__ && !__AVX2__ && !__AVXVNNI__ && !__AVX512VNNI__
-// int convolution_im2col_gemm_int8_xop(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt);
 #endif
 
 // gemm_x86.h
@@ -2612,13 +2606,13 @@ static void convolution_im2col_input_tile_int8(const Mat& bottom_blob, Mat& B, i
     }
 #endif
 
-    // #if NCNN_RUNTIME_CPU && NCNN_AVXVNNIINT8 && __AVX__ && !__AVXVNNIINT8__ && !__AVX512VNNI__
-    //     if (ncnn::cpu_support_x86_avx_vnni_int8())
-    //     {
-    //         convolution_im2col_input_tile_int8_avxvnniint8(bottom_blob, B, j, max_jj, k, max_kk, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h);
-    //         return;
-    //     }
-    // #endif
+#if NCNN_RUNTIME_CPU && NCNN_AVXVNNIINT8 && __AVX__ && !__AVXVNNIINT8__ && !__AVX512VNNI__
+    if (ncnn::cpu_support_x86_avx_vnni_int8())
+    {
+        convolution_im2col_input_tile_int8_avxvnniint8(bottom_blob, B, j, max_jj, k, max_kk, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h);
+        return;
+    }
+#endif
 
 #if NCNN_RUNTIME_CPU && NCNN_AVXVNNI && __AVX__ && !__AVXVNNI__ && !__AVXVNNIINT8__ && !__AVX512VNNI__
     if (ncnn::cpu_support_x86_avx_vnni())
@@ -2707,14 +2701,6 @@ static void convolution_im2col_input_tile_int8(const Mat& bottom_blob, Mat& B, i
 
 static void convolution_im2col_gemm_transform_kernel_int8(const Mat& kernel, Mat& AT, int inch, int outch, int kernel_w, int kernel_h, const Option& opt)
 {
-    // #if NCNN_RUNTIME_CPU && NCNN_AVX2 && __AVX__ && !__AVX2__ && !__AVXVNNI__ && !__AVX512VNNI__
-    //     if (ncnn::cpu_support_x86_avx2())
-    //     {
-    //         convolution_im2col_gemm_transform_kernel_int8_avx2(kernel, AT, inch, outch, kernel_w, kernel_h, opt);
-    //         return;
-    //     }
-    // #endif
-
     // NCNN_LOGE("convolution_im2col_gemm_transform_kernel");
     const int maxk = kernel_w * kernel_h;
 
@@ -2765,12 +2751,12 @@ static void convolution_im2col_gemm_transform_kernel_int8(const Mat& kernel, Mat
         }
     }
 
-#if 1 //NCNN_AVX512VNNI || NCNN_AVXVNNI
+#if NCNN_AVX512VNNI || NCNN_AVXVNNI
     bool has_w_shift = false;
     if (TILE_K >= 4)
     {
         has_w_shift = ncnn::cpu_support_x86_avx512_vnni() || ncnn::cpu_support_x86_avx_vnni();
-#if 0  //NCNN_AVXVNNIINT8
+#if NCNN_AVXVNNIINT8
         if (ncnn::cpu_support_x86_avx_vnni_int8())
             has_w_shift = false;
 #endif // NCNN_AVXVNNIINT8
@@ -4585,7 +4571,7 @@ static void unpack_output_tile_int32(const Mat& topT, Mat& top_blob, int i, int
     }
 }
 
-static int convolution_im2col_gemm_int8(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, const Mat& output_descales, const Mat& bias_data, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt)
+static int convolution_im2col_gemm_int8(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt)
 {
     const int maxk = kernel_w * kernel_h;
 
diff --git a/src/layer/x86/convolution_x86.cpp b/src/layer/x86/convolution_x86.cpp
index 1de8c1640c9..65ee41b332d 100644
--- a/src/layer/x86/convolution_x86.cpp
+++ b/src/layer/x86/convolution_x86.cpp
@@ -1022,8 +1022,7 @@ int Convolution_x86::forward_int8_x86(const Mat& bottom_blob, Mat& top_blob, con
     }
     else if (opt.use_sgemm_convolution)
     {
-        // ret = convolution_im2col_gemm_int8(bottom_blob_bordered, top_blob_int32, weight_sgemm_data, scale_in_data, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, _nT, opt);
-        ret = convolution_im2col_gemm_int8(bottom_blob_bordered, top_blob_int32, weight_sgemm_data, scale_in_data, bias_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, _nT, opt);
+        ret = convolution_im2col_gemm_int8(bottom_blob_bordered, top_blob_int32, weight_sgemm_data, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, _nT, opt);
     }
     else
     {
diff --git a/src/layer/x86/convolution_x86_avx2.cpp b/src/layer/x86/convolution_x86_avx2.cpp
index 927d3e143e6..c70c2c6a39b 100644
--- a/src/layer/x86/convolution_x86_avx2.cpp
+++ b/src/layer/x86/convolution_x86_avx2.cpp
@@ -34,16 +34,6 @@ void convolution_packed_int8_avx2(const Mat& bottom_blob, Mat& top_blob, const M
 }
 
 // gemm
-// void convolution_im2col_gemm_transform_kernel_int8_avx2(const Mat& kernel, Mat& AT, int inch, int outch, int kernel_w, int kernel_h, const Option& opt)
-// {
-//     convolution_im2col_gemm_transform_kernel_int8(kernel, AT, inch, outch, kernel_w, kernel_h, opt);
-// }
-//
-// int convolution_im2col_gemm_int8_avx2(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt)
-// {
-//     return convolution_im2col_gemm_int8(bottom_blob, top_blob, AT, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, nT, opt);
-// }
-
 void convolution_im2col_input_tile_int8_avx2(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h)
 {
     convolution_im2col_input_tile_int8(bottom_blob, B, j, max_jj, k, max_kk, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h);
diff --git a/src/layer/x86/convolution_x86_avx512vnni.cpp b/src/layer/x86/convolution_x86_avx512vnni.cpp
index e3a5262194d..1ab42a541b2 100644
--- a/src/layer/x86/convolution_x86_avx512vnni.cpp
+++ b/src/layer/x86/convolution_x86_avx512vnni.cpp
@@ -29,11 +29,6 @@ void convolution_packed_int8_avx512vnni(const Mat& bottom_blob, Mat& top_blob, c
 }
 
 // gemm
-// int convolution_im2col_gemm_int8_avx512vnni(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt)
-// {
-//     return convolution_im2col_gemm_int8(bottom_blob, top_blob, AT, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, nT, opt);
-// }
-
 void convolution_im2col_input_tile_int8_avx512vnni(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h)
 {
     convolution_im2col_input_tile_int8(bottom_blob, B, j, max_jj, k, max_kk, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h);
diff --git a/src/layer/x86/convolution_x86_avxvnni.cpp b/src/layer/x86/convolution_x86_avxvnni.cpp
index 9a64d2735b6..b612f711b95 100644
--- a/src/layer/x86/convolution_x86_avxvnni.cpp
+++ b/src/layer/x86/convolution_x86_avxvnni.cpp
@@ -29,11 +29,6 @@ void convolution_packed_int8_avxvnni(const Mat& bottom_blob, Mat& top_blob, cons
 }
 
 // gemm
-// int convolution_im2col_gemm_int8_avxvnni(const Mat& bottom_blob, Mat& top_blob, const Mat& AT, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h, int nT, const Option& opt)
-// {
-//     return convolution_im2col_gemm_int8(bottom_blob, top_blob, AT, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h, nT, opt);
-// }
-
 void convolution_im2col_input_tile_int8_avxvnni(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h)
 {
     convolution_im2col_input_tile_int8(bottom_blob, B, j, max_jj, k, max_kk, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h);
diff --git a/src/layer/x86/convolution_x86_avxvnniint8.cpp b/src/layer/x86/convolution_x86_avxvnniint8.cpp
new file mode 100644
index 00000000000..3a2beceadb8
--- /dev/null
+++ b/src/layer/x86/convolution_x86_avxvnniint8.cpp
@@ -0,0 +1,29 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2024 THL A29 Limited, a Tencent company. All rights reserved.
+//
+// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
+// in compliance with the License. You may obtain a copy of the License at
+//
+// https://opensource.org/licenses/BSD-3-Clause
+//
+// Unless required by applicable law or agreed to in writing, software distributed
+// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+// CONDITIONS OF ANY KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations under the License.
+
+#include "cpu.h"
+#include "mat.h"
+#include "x86_usability.h"
+
+namespace ncnn {
+
+#include "convolution_im2col_gemm_int8.h"
+
+// gemm
+void convolution_im2col_input_tile_int8_avxvnniint8(const Mat& bottom_blob, Mat& B, int j, int max_jj, int k, int max_kk, int kernel_w, int kernel_h, int dilation_w, int dilation_h, int stride_w, int stride_h)
+{
+    convolution_im2col_input_tile_int8(bottom_blob, B, j, max_jj, k, max_kk, kernel_w, kernel_h, dilation_w, dilation_h, stride_w, stride_h);
+}
+
+} // namespace ncnn