sorry to ask. From the concise description of the webpage http://test-drive-grid.com/?content=TestDriveCTA&medium=email&source=driver-download, it seems that the NVIDIA GRID is a cluster service that provide a virtual but powerful (and up-to-date) nVidia graphics card. So, even if a user doesn’t install a physical nVidia graphics card, s/he can run applications built for nVidia graphics card through sending some requests to the remote computers in the cluster equipped with nVidia graphics card (using, say, OpenCL). Is my understanding correct? Thanks.

No apologies, good question. The Test Drive is simply a demo environment to see GRID in action, to experience it, test it, confirm it works. GRID is the virtualization of NVIDIA GPUs installed in a host server and then leveraging the use of a hypervisor to distribute those GPUs to guest workloads. So yes there are physical NVIDIA GPUs installed in the host, and depending on the delivery mechanism and stack those GPUs can be shared, pinned, or virtualized to bring an amazing visual experience to a VDI desktop user on almost any device.

Thanks for that, was exactly the answer I came here to find. For another basic question, what are the limitations around the number and types of GPUs that can be installed in the host system?

Thank you!