Can NVIDIA extend this feature to GeForce cards if they are exclusively used for computation (no video output) as a secondary card? Remote desktop is orders of magnitude faster than any version of VNC I tried. Right now I’m using RD to program and then I stop RD, start VNC and try the program … I know that RD is disabling the normal driver and installing another graphic driver, so probably we need a computing driver only for Geforce & CUDA. Or maybe is easier for Microsoft to solve this problem by not deactivating the drivers for non-displaying cards?
I agree that NVIDIA needs to sell TESLAs, but I think they are for companies, not for universities where the money for research are always limited. We have 6 GTX 400s. A single TESLA card has a similar price. When we will deploy the application (handwritten text recognition) to a company, we will advice them to buy TESLA cards because they are more reliable.