Does Windows absolutely not expose OpenGL when connecting through Remote Desktop?

This problem has puzzled me quite a long time. When I connect to a Windows system through Remote Desktop, the OpenGL disappeared (to be accurate, the only OpenGL available is version 1.1, which is equivalent to no OpenGL at all). It is the case when I used Amazon EC2 p2 instance yesterday. The graphics card in this type of instance is Tesla K80. I installed its driver and CUDA 9.2 inside the p2 instance. CUDA works fine, but OpenGL has the same problem when I connect through Remote Desktop (you know I have to connect remotely). The system can detect the presence of K80 hardware, but it thinks the graphics adapter used is a basic VGA card, so cannot expose the OpenGL 4.6 to the user. Please see some screenshots of AIDA64 below:

As you can see from the picture above, the system think the graphics device is only “Basic Display Adapter”

As you can see from the two pictures above, the system correctly detects the presence of K80 graphics card hardware.


the Windows system only exposes OpenGL 1.1 to the user. Is it absolutely true that Windows does not expose OpenGL when connecting through Remote Desktop? Must we use Linux system in order to have correct version of OpenGL? Has anybody ever successfully configured a Windows system to have OpenGL 4.5 or 4.6 available through remote desktop? The OS used is Windows server 2016. I am looking forward to a definitive answer from some experts and, if possible, some instructions to solve this issue. Thanks a lot!

This forum is for using NGC on Amazon Web Services. Please try posting on a Windows-centric forum