diff --git a/onnxruntime/core/providers/cuda/cuda_execution_provider_info.h b/onnxruntime/core/providers/cuda/cuda_execution_provider_info.h
index cd4d27f78f..534d3354ec 100644
--- a/onnxruntime/core/providers/cuda/cuda_execution_provider_info.h
+++ b/onnxruntime/core/providers/cuda/cuda_execution_provider_info.h
@@ -56,9 +56,10 @@ struct CUDAExecutionProviderInfo {
   // arena config.
   OrtArenaCfg* default_memory_arena_cfg{nullptr};
   CUDAExecutionProviderExternalAllocatorInfo external_allocator_info{};
-  // By default use fix workspace size (32M) for Conv algo search, the final algo might not be the best.
-  // If set to true, try to use as much as possible memory for algo search.
-  bool cudnn_conv_use_max_workspace{false};
+
+  // By default, try to use as much as possible memory for algo search.
+  // If set to false, use fix workspace size (32M) for Conv algo search, the final algo might not be the best.
+  bool cudnn_conv_use_max_workspace{true};
 
   bool enable_cuda_graph{false};
 
diff --git a/onnxruntime/core/session/provider_bridge_ort.cc b/onnxruntime/core/session/provider_bridge_ort.cc
index 1198a7a3ef..319cb5f7b6 100644
--- a/onnxruntime/core/session/provider_bridge_ort.cc
+++ b/onnxruntime/core/session/provider_bridge_ort.cc
@@ -1182,7 +1182,7 @@ OrtCUDAProviderOptionsV2 OrtCUDAProviderOptionsToOrtCUDAProviderOptionsV2(const
   cuda_options_converted.user_compute_stream = legacy_cuda_options->user_compute_stream;
   cuda_options_converted.default_memory_arena_cfg = legacy_cuda_options->default_memory_arena_cfg;
   // Use default value as this field is not available in OrtCUDAProviderOptions
-  cuda_options_converted.cudnn_conv_use_max_workspace = 0;
+  cuda_options_converted.cudnn_conv_use_max_workspace = 1;
   cuda_options_converted.enable_cuda_graph = 0;
   cuda_options_converted.cudnn_conv1d_pad_to_nc1d = 0;
 
@@ -1668,7 +1668,7 @@ ORT_API_STATUS_IMPL(OrtApis::CreateCUDAProviderOptions, _Outptr_ OrtCUDAProvider
   (*out)->has_user_compute_stream = 0;
   (*out)->user_compute_stream = nullptr;
   (*out)->default_memory_arena_cfg = nullptr;
-  (*out)->cudnn_conv_use_max_workspace = 0;
+  (*out)->cudnn_conv_use_max_workspace = 1;
   (*out)->enable_cuda_graph = 0;
   (*out)->cudnn_conv1d_pad_to_nc1d = 0;
   return nullptr;