If you are in the US you can try how quick NVidia's GRID virtual GPUs perform, this is not X windows, pretty much they encode the video in H.264 or something similar, within the GPU early enough that the latency is reduced greatly.
I haven't tried it yet because I want to make full use of my time evaluating it, I would recommend you download the manual within the page and read it before starting your 24 hour test drive.
http://test-drive-grid.com/Yup, the cloud delivering on demand GPUs for your more demanding visualization applications.
But it's too pricey for now. Specially if you pair it up with a module that NVidia is trying monitor manufacturers to adopt that will decode the stream faster eating away even more latency.
Carmack when talking about the Oculus rift a year before he decided to join them, was collaborating to some extent with them and on his GDC interview (I bet available in YouTube or at least in game developers website) stated that you can send data faster from the US to Europe than from your PC to your eye due to all the processing time TVs take (Of course monitors don't have that latency, well maybe some DHCP processing).
Say you run a software program at 60Hz, you have 16.6ms between frames, 33.333ms at 30Hz. I can send packets in less than 7ms to/from a server if the server is well connected to Tier 1 backbones. So the perceived latency will be almost non measurable unless you are a professional gamer that can tell because they trained their muscle memory to react at high frame rates and a little latency will be noticeable because their input sequences won't work due to the lag.
But this is way better than your typical windows remote desktops.
Still I would give win-sshfp a try or if you use Linux/OSx I think they have native sshfp support as described in the link I posted earlier.