We need competition -- as most of us agree. While I might like AMD and the next guy/gal likes Intel, we need them both -- or new products at lower prices will not be released.

An article that surmises that Intel release schedule is being slowed because of inadequate competition from Bulldozer:

Intel Delays Product

Same would apply to AMD vs Nvidia -- we need them both to do well -- so we can get more for our money.

Competition is good

It seems like PC hardware advancement has slowed down in the last four years. Most games these days being multiplatform releases largely geared toward 5/6-year-old consoles might have something to do with that.

I still remember the days of CPU performance DOUBLING every year (and new games developed with this sort of expectation if you want to keep a smooth 60 FPS), new DirectX and OpenGL revisions at a breakneck pace, the GeForce 6800 Ultra coming out of nowhere and utterly trouncing the Radeon 9800 XT while releasing at the same price point...you get the picture.

In fact, it's why I'm still using the same CPU, mobo, and graphics card for four years. I don't want to be one of those suckers who gets caught buying a 9800 XT right before the 6800 Ultra, or a 7950GX2 before the 8800 GTX, or...you get the picture. But will we ever see that sort of thing happen again?

Between now and 2020, not likely. Because you run into fundamental quantum mechanical problems at around 10 nm. So, after this is is more clever, not smaller that will rule.

Playing devils advocate:
If one of the graphic card manufacturers were to quit the gaming GPU market, game designers could bypass DirectX D3D. You would get an effect similar to the one you see on the consoles: the hardware might stay the same for 3 to 6 years, but the games would visually improve over time. It would also drop the damned 3000(ish) moving object draw call limit in DX3D and up that to 30.000. Take that number to a flight sim and you can have 3000 aircraft(*) moving in view. Or 2900 aircraft and 10.000 bombs and parachutes. Add ground objects, remove 3 object calls per vehicle.

(*) Aircraft moving parts:
Main body = 1
Propeller
Ailerons
Rudder
Elevator
Flaps
Cowling
Oil cooler
Canopy = 10

Add a pilot, add at least two calls if the head tracks. Moving arms? Add 5 or 6 (with stick yoke). Gunner? 6 calls for each too. A fully crewed B-17 would therefore require around 70 calls.

No wonder a game with detailed aircraft and crew can only draw a maximum of 30 bombers, 30 escorts and 30 bandits in an otherwise empty world. With tricks the number can be made a bit higher, but you get the idea.

A simulation on a console can handle a 10 fold number. Time for a sim developed on PS4 and XBOX720?

One side effect of a slower hardware market JAMF, could be higher prices due to lower volume. I'm not talking about monopolistic behavior either. But rather than a 5 year old GPU that is still the "latest" does not grow old on the store shelf.

Thus with narrower selection on the shelf, the drive for consumers to pick and choose favorites is gone. That kind of consumer "frenzy" increases sales (upgrades, sidegrades, brand switches) which directly allows greater economy of scale at manufacturing.

And yes, perhaps you could see improvements in technology as you mention. But what would motivate the manufacturers to invest the R&D and plants to actually realize this? If there is only one graphic option on the shelf, the consumer really doesn't have any to choice, no matter what the manufacturer puts in the box. The envelope is pushed by companies trying to stay alive and ahead in the market. With one product alone, we'd see exactly what Allen points out: A new product held for later release because they can profit on the same old stale retail boxes today as yesterday. Consumers have nowhere else to go.

Voting with the wallet in such a case would mean leaving the hobby.

I know the other side of the coin.