diff --git a/include/micm/cuda/util/cuda_param.hpp b/include/micm/cuda/util/cuda_param.hpp
index f33fb78fd..d415a3fa2 100644
--- a/include/micm/cuda/util/cuda_param.hpp
+++ b/include/micm/cuda/util/cuda_param.hpp
@@ -7,7 +7,7 @@
 
 // To make the NormalizedError function works properly on GPU,
 // make sure to choose the BLOCK_SIZE from [32, 64, 128, 256, 512, 1024]
-const std::size_t BLOCK_SIZE = 32;
+const std::size_t BLOCK_SIZE = 512;
 
 /// This struct holds the (1) pointer to, and (2) size of
 ///   each constatnt data member from the class "ProcessSet";