diff --git a/bioem_cuda.cu b/bioem_cuda.cu
index 4860c28e599b352f3b45bab1372a13797b7f35e5..d9d804ed7a1c3c9e3bd35dc4578f0c9e4009b6a9 100644
--- a/bioem_cuda.cu
+++ b/bioem_cuda.cu
@@ -302,7 +302,7 @@ int bioem_cuda::selectCudaDevice()
 		cuCtxDestroy(tmpContext);
 		checkCudaErrors(cudaGetDeviceProperties(&deviceProp, i));
 
-		if (DebugOutput >= 2) printf("CUDA Device %2d: %s (Rev: %d.%d - Mem Avail %lld / %lld)\n", i, deviceProp.name, deviceProp.major, deviceProp.minor, (long long int) free, (long long int) deviceProp.totalGlobalMem);
+		if (DebugOutput >= 2 && mpi_rank == 0) printf("CUDA Device %2d: %s (Rev: %d.%d - Mem Avail %lld / %lld)\n", i, deviceProp.name, deviceProp.major, deviceProp.minor, (long long int) free, (long long int) deviceProp.totalGlobalMem);
 		long long int deviceSpeed = (long long int) deviceProp.multiProcessorCount * (long long int) deviceProp.clockRate * (long long int) deviceProp.warpSize;
 		if (deviceSpeed > bestDeviceSpeed)
 		{
@@ -356,7 +356,7 @@ int bioem_cuda::selectCudaDevice()
 	
 	if (DebugOutput >= 1)
 	{
-		printf("BioEM for CUDA initialized, %d GPUs found, using GPU %d\n", count, bestDevice);
+		printf("BioEM for CUDA initialized (MPI Rank %d), %d GPUs found, using GPU %d\n", mpi_rank, count, bestDevice);
 	}
 	
 	return(0);