I think they did stick with their formula, which was a mistake. Besides some tiny free roaming and driving, the game is basically a linear corridor shooter like Doom.
The engine itself is pretty remarkable, with the very large megatextures used in the landscape; the reason so many other textures look like shit is so those big textures would fit on a console disc. It was originally supposed to spew out code for PC and all consoles with one codebase too; but apparently that got dropped.
Great engine, mediocre game.
The Engine contains all 3 engines but uses the single interface, so create it once, it spits out 3 versions. Compression was the problem, if Megatexture was uncompressed, it wieghs in some 1 Terabytes or more. That is the major downfall of virtual texturing, All of these different compression algorithms are lossy, meaning the data must eventually be decimated to reduce the size. As a result, texture quality is reduced.
Another problem, if they decreased the compression, console size is 22gb's. You reach closer to the peak bandwidth or seek time...now you have stuttering. ID chose to select lower mip levels which makes it look muddy but reduces stuttering.
Basically, for both console and PC, they maxed out what they were able to do within reason of the fact that most PC gamers don't have the most powerful rig with TB's of data.
Virtual texturing is a great theory, it just isn't the most practical which is why no one else ever used it. Some have done so in tech demo's, but the stuttering is too horrible.
Other issue is it's not Modder friendly.
Here is a list i found with the benefits and problems of this method
Benefits
Unique texel for every inch (or smaller) of your world!
Parallel distributable art pipeline (at least for painting surfaces).
Smaller, close quarter environments are sure to shine with this approach. I am eager and hopeful to see what Doom 4 could be in the confines of a space station.
Raw rendering performance is governed by just one parameter (triangle count) instead of balancing two (triangles and texture usage). Textures are a fixed overhead.
This technique can be used for all typical surfaces, animated or otherwise.
Disadvantages
Extensive tools required to work with this type of database. Photoshop is not a Megatexture editor.
Terabytes of shared storage space is required during development, typically on a networked drive and potential VPN access for outside contractors.
Getting the run-time implementation to work "right" is potentially years of effort.
The more sprawled an environment becomes the more surfaces you need. This means lower texel density or more aggressive compression will be required to maintain storage space criteria. Overall it will begin to degrade the quality and textures will become muddy. I am hopeful that the muddy bits I've seen in trailers for Rage are an artifact of Flash video and not the game but I have a feeling my suspicions are correct.
Unknown
The PC front is experiencing some changes. It's hard to say if this approach is feasible in the next generation of games. I suppose most game technology has issues surviving more than one or two generations anyhow, so this isn't too big of a surprise. There is also the hope that this type of content can be stored in the Cloud and streamed over a live connection for the players local area.
How much surface area can we cover before this technology explodes?
Will gamers be forgiving of a 100GB installation, or will it be a giant target when that gamer is looking to free up some space on their hard drives?