diff --git a/src/plugins/intel_gpu/src/graph/convolution.cpp b/src/plugins/intel_gpu/src/graph/convolution.cpp
index 9fee2996bca..8684bb9287e 100644
--- a/src/plugins/intel_gpu/src/graph/convolution.cpp
+++ b/src/plugins/intel_gpu/src/graph/convolution.cpp
@@ -19,6 +19,46 @@ primitive_type_id convolution::type_id() {
     return &instance;
 }
 
+static format get_recommended_format(layout input_layout, data_types output_type) {
+    if (data_type_traits::is_i8_u8(output_type)) {
+        switch (input_layout.format) {
+            case format::b_fs_yx_fsv16:         return format::b_fs_yx_fsv32;
+            case format::bs_fs_yx_bsv32_fsv16:  return format::bs_fs_yx_bsv32_fsv32;
+            case format::b_fs_zyx_fsv16:        return format::b_fs_zyx_fsv32;
+            case format::bs_fs_zyx_bsv32_fsv16: return format::bs_fs_zyx_bsv32_fsv32;
+            case format::b_fs_yx_fsv2:
+            case format::b_fs_yx_fsv4:          return format::b_fs_yx_fsv32;
+            case format::b_fs_zyx_fsv2:
+            case format::b_fs_zyx_fsv4:         return format::b_fs_zyx_fsv32;
+            case format::bs_fs_yx_bsv8_fsv2:
+            case format::bs_fs_yx_bsv8_fsv4:    return input_layout.batch() > 16 ? format::bs_fs_yx_bsv32_fsv32 : format::b_fs_yx_fsv32;
+            case format::bs_fs_zyx_bsv8_fsv2:
+            case format::bs_fs_zyx_bsv8_fsv4:   return input_layout.batch() > 16 ? format::bs_fs_zyx_bsv32_fsv32 : format::b_fs_zyx_fsv32;
+            default:
+                break;
+        }
+    } else if (data_type_traits::is_floating_point(output_type)) {
+        switch (input_layout.format) {
+            case format::b_fs_yx_fsv32:         return format::b_fs_yx_fsv16;
+            case format::bs_fs_yx_bsv32_fsv32:  return format::bs_fs_yx_bsv32_fsv16;
+            case format::b_fs_zyx_fsv32:        return format::b_fs_zyx_fsv16;
+            case format::bs_fs_zyx_bsv32_fsv32: return format::bs_fs_zyx_bsv32_fsv16;
+            case format::b_fs_yx_fsv2:
+            case format::b_fs_yx_fsv4:          return format::b_fs_yx_fsv16;
+            case format::b_fs_zyx_fsv2:
+            case format::b_fs_zyx_fsv4:         return format::b_fs_zyx_fsv16;
+            case format::bs_fs_yx_bsv8_fsv2:
+            case format::bs_fs_yx_bsv8_fsv4:    return input_layout.batch() > 16 ? format::bs_fs_yx_bsv32_fsv16 : format::b_fs_yx_fsv16;
+            case format::bs_fs_zyx_bsv8_fsv2:
+            case format::bs_fs_zyx_bsv8_fsv4:   return input_layout.batch() > 16 ? format::bs_fs_zyx_bsv32_fsv16 : format::b_fs_zyx_fsv16;
+            default:
+                break;
+        }
+    }
+
+    return format::any;
+}
+
 layout convolution_inst::calc_output_layout(convolution_node const& node) {
     auto desc = node.get_primitive();
 
@@ -173,40 +213,10 @@ layout convolution_inst::calc_output_layout(convolution_node const& node) {
 
     // Adjust output format for mixed precision case in onednn
     auto out_fmt = input_layout.format;
-    bool is_2d = (input_layout.format.spatial_num() == 2);
-    bool is_3d = (input_layout.format.spatial_num() == 3);
     if (node.get_preferred_impl_type() == impl_types::onednn) {
-        if (data_type_traits::is_i8_u8(output_type)) {
-            if (is_2d) {
-                if (input_layout.format == format::b_fs_yx_fsv16)
-                    out_fmt = format::b_fs_yx_fsv32;
-                else if (input_layout.format == format::bs_fs_yx_bsv32_fsv16)
-                    out_fmt = format::bs_fs_yx_bsv32_fsv32;
-                else if (input_layout.format == format::b_fs_yx_fsv2)
-                    out_fmt = format::b_fs_yx_fsv32;
-            } else if (is_3d) {
-                if (input_layout.format == format::b_fs_zyx_fsv16)
-                    out_fmt = format::b_fs_zyx_fsv32;
-                else if (input_layout.format == format::bs_fs_zyx_bsv32_fsv16)
-                    out_fmt = format::bs_fs_zyx_bsv32_fsv32;
-            }
-        } else if (data_type_traits::is_floating_point(output_type)) {
-            if (is_2d) {
-                if (input_layout.format == format::b_fs_yx_fsv32)
-                    out_fmt = format::b_fs_yx_fsv16;
-                else if (input_layout.format == format::bs_fs_yx_bsv32_fsv32)
-                    out_fmt = format::bs_fs_yx_bsv32_fsv16;
-            } else if (is_3d) {
-                if (input_layout.format == format::b_fs_zyx_fsv32)
-                    out_fmt = format::b_fs_zyx_fsv16;
-                else if (input_layout.format == format::bs_fs_zyx_bsv32_fsv32)
-                    out_fmt = format::bs_fs_zyx_bsv32_fsv16;
-                else if (input_layout.format == format::b_fs_zyx_fsv2)
-                    out_fmt = format::b_fs_zyx_fsv16;
-                else if (input_layout.format == format::bs_fs_zyx_bsv8_fsv2)
-                    out_fmt = input_layout.batch() > 16 ? format::bs_fs_zyx_bsv32_fsv16 : format::b_fs_zyx_fsv16;
-            }
-        }
+        format recommended_fmt = get_recommended_format(input_layout, output_type);
+        if (recommended_fmt != format::any)
+            out_fmt = recommended_fmt;
     }
 
     // get output feature map from weights. It should be the same as number of biases. Will be verifed in
diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp b/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp
index a91b3c0c8b7..4da00b1d2ad 100644
--- a/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp
+++ b/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp
@@ -608,12 +608,10 @@ void reorder_inputs::run(program& p, layout_optimizer& lo, reorder_factory& rf)
             if (new_layout == input_layout)
                 return;
 
-            if (!input.is_type<reorder>() || input.get_users().size() > 1) {
-                auto new_input = rf.get_reorder(input.id(), input_layout, new_layout);
-                if (new_input.first)
-                    p.add_intermediate(new_input.first, conv_node, 0, !new_input.second);
+            auto new_input = rf.get_reorder(input.id(), input_layout, new_layout);
+            if (new_input.first) {
+                p.add_intermediate(new_input.first, conv_node, 0, !new_input.second);
             }
-
             conv_node.get_dependencies().front()->set_output_layout(new_layout, false);
         }
 
diff --git a/src/plugins/intel_gpu/src/graph/reorder.cpp b/src/plugins/intel_gpu/src/graph/reorder.cpp
index 79988395970..16524b607cc 100644
--- a/src/plugins/intel_gpu/src/graph/reorder.cpp
+++ b/src/plugins/intel_gpu/src/graph/reorder.cpp
@@ -32,16 +32,6 @@ layout reorder_inst::calc_output_layout(reorder_node const& node) {
         ofmt = ifmt;
     }
 
-    if (node.is_valid_output_layout() && input_layout.feature() <= 4) {
-        auto users = node.get_users();
-        if (users.size() > 0 && users.front()->is_type<convolution>()) {
-            auto expected_fmt = node.get_output_layout().format;
-            if (expected_fmt == format::b_fs_zyx_fsv2 || expected_fmt == format::bs_fs_zyx_bsv8_fsv2) {
-                ofmt = expected_fmt;
-            }
-        }
-    }
-
     if (ifmt.is_nv12()) {
         auto data_size = tensor{ input_layout.batch(), input_layout.feature() * 3,
                                  input_layout.spatial(0), input_layout.spatial(1) };