Testing T4 card fails

When using deepstream with T4 seeing performance degradation over time + when performing basic cuda test it fails , see below:

test failed after 280*10 iterations
step 280
terminate called after throwing an instance of ‘cv::Exception’
what(): OpenCV(3.4.5) /home/nvidia/opencv/modules/core/src/cuda/gpu_mat.cu:224: error: (-217:Gpu API call) unknown error in function ‘upload’

#include <opencv2/core/cvstd.hpp>
#include <opencv2/core/mat.hpp>
#include <opencv2/cudaimgproc.hpp>
#include <opencv2/cudabgsegm.hpp>
#include <opencv2/cudalegacy.hpp>
#include <opencv2/cudaarithm.hpp>
int main()
std::cout << "Number of gpu devices: " << cv::cuda::getCudaEnabledDeviceCount() << std::endl;

cv::Mat cpuMat_1, cpuMat_2;
cv::cuda::GpuMat gpuMat_1, gpuMat_2, gpuMat_3;

int i=0;

cpuMat_1 = cv::Mat::ones(14000, 14000, CV_8SC1);
cpuMat_2 = cv::Mat::ones(14000, 14000, CV_8SC1);


cv::cuda::add(gpuMat_1, gpuMat_2, gpuMat_3);

if((i%10) == 0){
std::cout << “step” << " " << i << “\n”;
std::cout << “done.”;
return 0;

Same tests work on V100.

Tested with two different T4 cards , is it a known issue? / Could it be a driver - cuda - ocv - deepstream versions issue?

Hi eyal.enav,

Sorry for the late response, would you please let us know if this still an issue at you side?
Have you migrated on the newly release DeepStream SDK 4.0.1?