Here's an interview with Mark Rein that is pretty interesting.  I bolded a few parts that stood out to me.  Epic isn't the only developer saying things like this, Id software seems to be saying the same thing and both have commented in the past that the out of control piracy situation on the PC is causing them to devote more and more resources to consoles.

http://www.computerandvideogames.com/article.php?id=169116
Monday 30-Jul-2007 11:55 AM - Stuart Bishop

Microsoft might be making a song and dance about the advent of DirectX 10 and the new experience it brings to PC gaming, but Epic vice president Mark Rein reckons we won't see it adopted in earnest by developers and publishers for several years.

"To be honest, and I'm going to cast a small pall on the industry here, I don't think you're going to get much higher until the next generation of consoles," Rein told CVG when asked what we can expect from DX 10 beyond what we're seeing in Crysis.

One reason for this, he said, is that the gulf between a high-end PC and a low-end PC is bigger now than it's ever been, and the low-end has been anchored while the high end just gets higher "and it becomes less and less economically viable to do the super high-end stuff".

"And in addition, just now we're just barely coming into the sweet spot of the next-gen consoles... We're at the point where it's viable to ship games on these next-gen consoles and it's going to be a gold mine for a couple of years.

"Publishers are generally putting their money where the highest return is, and in the past that's been on consoles and I think that's still the case. So if you build the game that's a really super amazing high-end game, how do you make all the money you could make from that game, because it'll be too high end eventually to be on PS3 or Xbox 360"

Rein reckons there are developers who will dabble with DirectX 10 and a couple of PC-only games, or games that their big market is PC, will push the envelope a bit, "But I don't see people going way, way over what these consoles can do because then they can't sell the game... they have to dumb the games down for the consoles. Consoles will pretty much define what the next five years of games look like on the PC", Rein added.

However, he said that what we'll also see from PC games in the meantime is the ability to turn everything up a visual notch.

"That's the value to me of the PC, that the source content we have is still really high resolution, much higher than we can show on a 512MB machine. You can use that content on the PC, you can express it in higher resolutions and higher frame rates and turn on more effects. I think that's what you'll see, PC games where we turn it up higher."

Epic cares more about consoles than PCs, confirmed.

A lot of what he says has been pretty well known for awhile ...

If you remove gaming (and a few other applications) from the equation there is very little need to upgrade the hardware of your PC more than every 5 years; a 1.5GHz Pentium 4 is still an excellent PC for surfing the web and word processing and is over 5 years old. This has resulted in most gamers upgrading less often to much more modest hardware; personally I plan to upgrade to a laptop sometime in 2008 from my X2 3800+ and Geforce 7800GT.

At the same time, to keep hardware sales high companies like ATI and nVidia have targeted the elitest gamer and offered dual card (and 2 GPU per card) solutions; on top of that physics processors and high end CPUs continue to grow in power. As much as people talk about the epic gap between the processing power of the XBox 360/PS3 and the Wii PC developers have been forced to deal with a similar gap for years; if your game can't run on the low end hardware you can never sell enough units to break even and if you don't take advantage of the high end SLi systems you will never get the hype you need to sell your game.

But what's the point in getting a crossfire or sli setup if the super high-end games become few and far between?

Good maybe soon I can actually play more new PC games when they come out.

So what if DirectX 10 dies? It's proprietary and restricted by Microsoft. OpenGL is absolutely the way to go (in a free implementation such as Mesa 3D) because it is readily extensible and easily patched for performance or otherwise.

The reason DX10 is in "a stranglehold" is because only Microsoft Windows Vista uses it, which is about 5% of the desktop computer market, and MS has locked XP, Mac OS X, Linux and game consoles (except Xbox 360, but that's not a full implementation)

I expect that within 2 years the main technology that PC developers are going to be talking about is 'subdivision surfaces'

One of my final courses in University was a grad/undergrad mixed class on advanced modeling techniques. One of the topics in this course was  'subdivision surfaces' and was focused on converting a low resolution model into a higher resolution model by spliting polygons in different ways; a grad came into the course and displayed his research which was that a person could give a 'hint' on each edge of a model while modeling an object and convert an ultra-low polygon model (think N64) to an ultra-high polygon model (think XBox 360/PS3).

In 2 years I expect to start seeing games where the low end hardware displays the model with low resolution textures, few effects and no subdivision and the high end hardware displays a model after several stages of subdivision, high-resolution textures and full effects.

In other words they will be producing (expensive) content which can scale up to practically any level of hardware.

(BTW there are similar techniques for textures and you can easily create shaders for more advanced effects if they have greater processing power)

But how many games will take advantage of that? That requires programmer time and lots of testing just for the >5% of people that have high-end hardware. I think graphics were good enough on the GC, (think Mario Sunshine) which had old hardware relative to PCs, and it's only the trend toward ultra-realism that's needing even current midrange PC hardware.

Programmer time and testing is largely unimportant being that it would take less than 1 man-year to implement and test the technology; $50,000 may sound like a lot of money but it is a drop in the ocean of game development.

The artist's time is the important consideration because over 80% of game development costs are content creation. The reason why subdivision surfaces will be favoured is that it adds a minimal ammount of work (25%) and gives you a model which can be reused from this point on; consider how much of an advantage it would be for EA to create a couch model where they can reuse the exact same model in any game they create from this point on. Hypotetically speaking, large publishers and middleware developers will be able to create massive libraries of objects that are graphically advanced enough to last for years and could even have the physics-engine data built into them so they can be immediately plugged into a game engine.

Because I really, really like those games that are "few and far between". I like World of Warcraft, Half Life, Half Life 2 and Starcraft better than everything on the 360 and PS3 combined, and I'm not exaggerating. I have quite literally put 200+ days (or 4800 hours of game play) in to Starcraft alone. If these games continue to remain PC exclusive (Starcraft 2 certainly is) or PC centric (as Half Life is), then I would much rather pay 500 a year to keep my rig up to date than to play the 360/PS3 games I personally believe are much less fun.

But that's my opinion. I'm not saying you should agree. You asked, I answered.