Case in point: a setup where the GXV-3000 calls an X-Ten softphone. The video from the GXV-3000 is quite nice, the X-Ten softphone displays it without any problems. It even scales well.
The X-Ten softphone, however, has issues with its own video stream. On an 800 MHz machine, it has a lag which noticebly grows over time, until the application is running so slow that the program's buttons are unuseable until you hang up from the other end. (The video from the GXV-3000 keeps up during all this.)
(Note: On a 1.2 Ghz dual core, it appears to keep up.)
A N800-to-GXV call is a bit different. The N800 is able to keep up. It's just that the resolution of the camera on the N800 is just so low that the picture on the receiving end is comprised of giant pixels and overdriven colors. It's just too dang ugly to look at!
My recommendation is to try and maintain end-point parity (use the same hardware or software on both ends). That way, it may be a bit ugly but you don't end up comparing mediocre (the soft phones) with the good (the hard phones).