Well Anand said it didn't look as good as what Nvidia have shown so far, and you can bet your ass this was an absolute current best case scenario from AMD.
Which is better bet. A more expensive but better solution, or cheaper but worse? You pay your money and make your choice. Much like 3dvision mentioned in the OP. Was more expensive and locked down by Nvidia, most certainly.
Software support for 3dvision and nothing to do with the technology difference, it's a 120hz screen and a pair of glasses that sync to the 120hz, nothing more or less, being locked out of 3d mode based on the screen device id, there was nothing different. The software stack doesn't change the hardware having zero difference, either it paid the 3dvision branding charge, or it didn't.
Likewise ANandtech did NOT say g-sync was better, it said Nvidia had a better demo, nothing more or less. They also said specifically the biggest downside to g-sync was lack of monitor support and freesync appears to be down the path to fixing that.
This was also two notebooks with Kabini's in, not a 780gtx being used. It was existing retail bought laptops that supported the feature.
I'm sure Nvidia and AMD will improve their drivers for it in the future, but they didn't say the feature was worse than g-sync only that the demo wasn't as good. It was run on vastly different hardware and very different screens.
Nvidia won't have an extra cost for hardware, or better hardware, they won't continue to add a $150 cost through a FPGA the monitor makers certainly don't want to use long term.
FPGA is essentially a slower custom chip you can program to act(to a degree) in any way you like. It will use vastly more power, hugely more transistors than a fixed function silicon solution. The difference is using an existing FPGA and programming it is just stupidly faster than waiting for a full product tape out and chips to be ready. My opinion then, and even more so now(because there is essentially no other reason to both use a FPGA here nor stick it onto and replace an existing chip in the existing monitor) than time to market, they did this to get monitors out first and to be able to lock them in.
In the future when given the choice of that final silicon at a fraction of the price and power usage being available vs the expensive FPGA, it's a non contest.
Future screens will have this feature as standard in the normal controller that costs say $5 a screen and won't use a $100 FPGA instead. In the future Nvidia will be using standard hardware to use this feature. They could, possibly not lock Nvidia users to g-sync branded screens, more chance of it if their users kick up a fuss. As with 3dvision/sli mobo's/other stuff, they will likely continue to charge the likes of Asus say $10 a screen to be branded compatible and not locked out via drivers.
