I have been reading up on this, and outside of two areas, there is little current benefit to this card.
Those two areas are nice-to-haves, but they don’t justify the cost.
Those areas of improvement come from tweaks and tricks within the GPU itself to get around the memory bandwidth bottleneck.
First, it uses video data compression to handle items like anti-aliasing. This means that at least in theory, there will much, much less of a performance hit using full anti-aliasing. This appears to actually mean something at fairly high resolutions and color depth.
Second, it uses more intelligent means to reducing overdrawing, which makes the overall graphics engine more efficient, at least in theory to the tune of 20-30%.
How well these things work in real life remains to be seen, but even assuming these do work as suggested, the current cost overrides the benefits for most.
The Ray of Hope
The most intriguing development of the GeForce GPU is the ability for game programmers to program new effects for the CPU. Eventually, this could be a major advance in gaming technology.
However, this doesn’t help current games at all, and it remains to be seen just when games begin using such features, how many games will take advantage of this, and what they’ll be able to do with it.
Remember that games are usually put together to work at least reasonably well on older platforms. Any programmable features will be useless for a long time for the vast majority of gaming platforms, and will certainly remain that way so long as the price remains stratospheric.
Unless there’s a sizable user base, programmers won’t program for it. So long as you charge an arm and a leg for the product; you’re never going to get that sizable user base. For competitive reasons, NVidia needs to get a sizable user base to get programmers to program as quickly as possible.
Therefore, unless Nvidia is downright suicidal, we should see a Geforce3 MX fairly shortly. Because NVidia really needs game programmers to go out of their accustomed way and use the programmability of the GPU, it has much more need for a reasonably priced GeForce3 MX than it ever did for a GeForce2 MX. It needs to make its programmable GPU the standard (not only for itself, but to really thwart an ATI), and to do that, it’s going to have to make it relatively cheap.
It took NVidia about two months after the GeForce2 GTS came out to come up with the GeForce2 MX. I would bet we’ll see about the same timeframe for the GeForce3 MX. Matter of fact, from what I can gather, unless NVidia really tries to cripple it with slow memory, it ought to work very well with relatively cheap 5ns DDR, given some of the tweaks and tricks built into the GPU.
I would roughly guess that a GeForce3 MX card should get roughly 80% of the benefits of the megabucker for a third or less of its price.
Do I have a shred of proof for this? No, but it makes more sense to do that to establish a standard quickly than to wait around for a couple years before all those circuits get used.
Look at it this way, if you wait and I’m wrong, you’ve still saved some money. If I’m right, you’ve saved a lot more money.:)