The past year has brought us Two long awaited titles:

Doom3 and Half-Life 2.

They were both scheduled to be released at about the same time, Doom3 managed to pull it off. Valve, on the other hand, had some issues to resolve before Half-Life 2 could reach it's market.

Court cases, code thefts, and one very unhappy fan base plagued Valve during the run-up to release. Although this is unfortunate, it did bring out a lot of information about just what is happening in the PC hardware market.

The First bit came out in early September, 2003, during ATI's ShaderDay.
ShaderDay was ATI's technology showing for the 9800pro/xt, the then, highest performing card for gaming on the market

Valve was there too, giving a demonstration of their upcoming game, and what their game engine could do. A question was asked of Valve "How are you optimizing for Nvidia?" The answer opened a can of worms!

Fans were told that optimizing for Nvidia, (ATI's chief competitor) was just not worthwhile. That Valve had spent most of their time on optimizing for Nvidia already, but that performance of Nvidia's line-up was very lacking. They sited benchmarks from other games currently on the market.

Fans knew there had to be another story. As in some cases, ATI's 9800Pro paled in comparison to the Nvidia counterpart. So what was going on? Thankfully someone had the nerve to ask.
   
ATI responded pretty much with "Well, it works a bit like this, on the gpu itself, ATI cards have 8 physical registers for accepting data, while Nvidia's only have 5. Half-Life 2 is a DirectX 9 game, and DirectX 9 requires 24-bit precision.

Naturally, 24 fits nicely into 8 registers...in 3 simple  passes. But with Nvidia's architecture, it takes 5 passes, and one of the registers never gets used. Simple, but that's  what's happening. But really, it's not that big an issue. Nvidia could solve this problem very easily in the driver."
   
 The products Nvidia were marketing as DirectX 9 capable were actually not fully
compatible...they could not do 24-bit precision efficiently, they excelled at DirectX 8, but that was last year's news.

So, in response, Nvidia quickly released a driver that solved a lot of the performance issues and shortened the performance gap between themselves and the competition.

But few were buying it since Nvidia had recently been exposed  writing optimizations into their drivers for a major 3D card benchmarking utility - optimizations that let Nvidia's card give almost identical visual results, but perform far better.

It sadly also prevented Nvidia's cards from being able to do certain types of work, unless it would really needed at a later time.. Then astute fans began to ask ..."IF Nvidia can pull this off at the driver level, and get the needed performance, then why can't Valve do something to improve Nvidia's Halflife 2's compatibility?"

Valve's answer was that it was way too much work.
   
So what WAS the huge difference that held Nvidia back, and Valve was just not willing to undertake?