CUDA with Terminal Server

Otherwise known as Remote Desktop, the problem being that CUDA cannot initialize when run through terminal services.
Any ideas on a work-around?

use VNC.

(this is not a snarky response, I don’t think there’s any workaround for the session 0 isolation with RD because RD uses its own display driver)

Is this something that has any likelihood of being fixed? We just bought a Tesla board and put in in a server 2008 box for multiple people to log in to and use, and using VNC is not a great solution for our environment.

This will be fixed in the not-too-distant future for Tesla cards through a special driver.

Install cygwin, configure SSH server and allow multiple users to test their code via “puTTY” or a suitable SSH client. We use this for our projects.

We had a big discussion on the forum on this before and this was the solution recommended. It works…as long as u dont require GUI

Hi Sarnath,

Can you please shed more lights on how to configure SSH server with Cygwin for remote ssh to access CUDA? I have Cygwin ssh server running as a windows service and it’s a no go…




If you want a quick solution that also gives you a GUI, try TeamViewer. I’ve been using it for years now to run CUDA programs remotely on Windows with no issues –

As for the ssh server, you probably need to configure the usernames allowed to connect to the SSH server using the sshd.conf file, I believe. You might also need to allow access to pot 22 (or whatever other port you run the SSH server) on Windows Firewall.