Keys, I think the issue with physics in games is that it's not very substantial.
Nope, that's not it. The problem is that Nvidia has PhysX, and uses it's GPU's to run it in games such as Mafia II. The problem is that ATI based GPU's cannot run it, nor can an Nvidia card with ATI GPU present as primary. The problems is, the reason people are saying PhysX isn't like anything they haven't seen before that could run successfully on a CPU is because ATI cards cant run it on their GPU's so the only other alternative for those using them, is the CPU. Problem is, it's a guaranteed thing that if ATI "had" jumped on and supported PhysX and can run PhysX content on their GPU's, nobody would give a whit about PhysX running on CPU's, because ATI can do it to. It's the root of all these discussions across the net.
You can take the approach in Arkham and Mafia NV has taken with physics which is withhold features that would easily be able to be done convincingly on a CPU without a physics engine (debris and weapon on environment scarring decals). This is stuff that should be added to games, and it would be done easily with no physics engine whatsoever.
The real advantage I can think off the top of my head for real immersion that physics adds is fluid dynamics. Cloth and water additions would be fantastic. A brook breaking up convincingly randomly around rocks, cloth that actually is affected in the wind (not just by waving, but imagine flags manipulated at correct angles by wind speed and direction), that sort of thing, but that really isn't hyped for some reason, probably because it's not as sexy. Exploding buildings are fine and all, but every time I see something like that, I always remember I've seen it done convincingly before without the need for so much additional cost and computing power without a physics engine.
The real immersion you're speaking of would need to come from devs.
The other addition would be things like true ballistics. It would be nice for devs/nvidia/amd/intel to help develop actual game play changes so that physics beings something substantial to the table. I want to see a sniper rifle fire bullets that are truly affected not just by elevation like today's games, but also windage, rotation, yaw, etc. On top of that, why not give characters actual skeletons so terminal ballistics can be simulated with entry and exit wounds? Maybe one day I'll actually have a .338 Laupa exit a enemy in an angle that wasn't the entrance angle, adding some skill to sniping multiple enemies in a line.
Perhaps that will come to pass.
As of now, PhysX s just doing exactly what we've been able to do for years, except limiting it to just a subset of the population. It really adds nothing that shouldn't already be there, and costs too much in performance and price to boot (again with rare exceptions to the rule like cloth effects in Sacred 2 and Mirror's Edge, but even the effects in those games really aren't big game changers.) What we are show in not so much ineffective uses of PhysX, but a technology languishing from lack of imagination. Where's the true newtonian physics based space movement enhanced by PhysX, melee combat like shields being knocked away by swords convincingly randomly or maybe give things like the plasmid powees in Bioshock realistic physical properties? What I just saw in that video has a been there, done that without the expensive equipment feel to it and can you really blame people for technology if it isn't being used to give people something actually new, but instead is an answer searching for a nonexistant problem right now?
It's a feature that add more to the game. Like it or not. Impressed or not. The extra content is there.
Would it really be asking too much for PhysX to stop doing the hey look how sexy, yet insubstantial these additions are and actually doing a killer app that changes the gameplay experience instead?
Might as well be asking for AA, Ambient Occlusion, Tesselation to stop doing what they do. But these are all things ATI GPU's can do, so no problems.
Also offtopic, but a .45 ACP round (fired from the Thompson Submachine Gun) would not tear up brick walls that badly (2nd example), the .45 is a very low power/low velocity round with a very soft jacket and core. The vast majority of .45 loads fired from your standard 5" barrel full size pistol will be subsonic speed and do damage to soft targets, not from high energy but from low energy impacting over a large surface area, making it terrible to damage hard objects like body armor or brick walls, but good from a expansion and penetration POV on soft bodies. The non physX example is actually accurate on how a .45 round would actually look against a wall.