GPU Caps Viewer v1.8.6 Dives Deep on OpenCL Support
One of the great problems of both CUDA and OpenCL is that Windows doesn't offer information about them in any Control Panel application. Therefore, developers don't have a simple way to know whether their installed GPU support CUDA or OpenCL. GPU Caps Viewer v1.8.6 is a simple and free utility that solve this problem.
This utility offers detailed information about the graphics cards and the GPU (short for Graphics Processing Units) installed in a computer. When you run GPU Caps Viewer, is displays a window with many tabs.
The GPU/CPU tab shows details about the graphics card and GPU type and clock, amount of dedicated video memory and speed, number of cores, SM/SIMD, and information about the installed drivers. Besides, it provides basic information about OpenGL, OpenCL and NVIDIA CUDA support, as in Figure 1.
Figure 1: GPU Caps Viewer displaying information about the GPU, including the OpenCL version and compute devices.
In this case, the system offers one OpenCL compute device, with the following information for the version:
OpenCL 1.0 ATI-Stream-v2.0.1 - CU: 8 @ 1729MHz (CPU)
It means that the CPU offers support for OpenCL 1.0 through ATI-Stream v.2.0.1 and it will work with the 8 logical cores running at 1729 MHz. This information doesn't consider the dynamic overclocking capabilities offered by modern microprocessors, it just displays the nominal frequency discovered by the detection engine. If you want to learn more details about dynamic overclocking, you can read my previous post "Measuring Speedup is Challenging with Intel Turbo Boost Technology".
The GPU Computing section summarizes the capabilities supported by your GPU and the installed drivers, it shows three checkboxes. In this case, the OpenCL(CPU) checkbox is the only one activated because OpenCL is supported at the CPU level but not at the GPU level. This happens because the video card driver isn't updated to activate OpenCL support for the GPU. Therefore, you have to make sure that you have the right video card driver because OpenCL isn't going to take advantage of your GPU parallel processing capabilities.
The OpenCL tab offers more detailed information about the OpenCL support. It offers a dropdown list with all the CL platforms available in the system. This way, you can select the desired platform and see the information about it displayed in the window, as in Figure 2.
Figure 2: GPU Caps Viewer displaying detailed information about the OpenCL support. In this case, only the CPU is going to work with OpenCL.
The number of CL devices should be at least 2, the CPU and one GPU. When you select the GPU in the second dropdown list, GPU Caps Viewer should display GPU in Type and it will provide information about the compute units, clock, version and driver, as in Figure 3.
Figure 3: GPU Caps Viewer displaying detailed information about the OpenCL support. In this case, there is a GPU that is going to work with OpenCL, providing 30 compute units running at 1,242 MHz.
Besides, GPU Caps Viewer offers many OpenCL demos in their CPU and GPU versions. You can run them to test your OpenCL GPU capabilities. They take advantage of the multiple compute units to render real-time effects, as in Figure 4.
Figure 4: Figure 4. GPU Caps Viewer running the 4D Quaternion Julia Set OpenCL GPU demo.