Hi dmcavalcante As I understand OpenCL and directX are just APIs to utilize the same hardware for some rather different purposes. GPUs are primarily made for rendering purpose, so there maybe some parts which are exposed in Directx but not in OpenCL. But this certainly does not mean, that GPU will be underutilized from OpenCL. At the end, GPU has lots of Compute units and keeping them busy for long time, is not very easy always. You can try running MatrixMultiplication samples for large matices (eg. 4K X 4K)and run it for thousands of iterations. Also make sure you are not using CPU in between, which gives GPU some time to rest. It will be helpful if you can give details as to what OpenCL application and DirectX applications you used for the above behaviour. Also please mention your GPU, Driver, SDK and OS details.