Quote Originally Posted by Shioban View Post
But essentially each texture has its own memory footprint. A diffuse and sometimes a normal/specular map. Normal/Specular maps don't eat up as much memory as a diffuse texture for the most-part as they generally hold less colour information (i.e. less data). So one of the other concerns is the memory impact of the textures in the scene over-all i'd wager.
I personally don't know the impact of the textures in FFXIV, it could already be fairly high so I might be blowing a trumpet on 'Give me better textures' when the option simply isn't viable.

The rendering impact of ray-casting the lighting for the bump and specular maps generally isn't that heavy in deffered rendering, obviously the quality of this changes the rendering impact.
Yep, I may make bad assumptions from time to time, but while I don't specialize in CG, I am a fairly competent (if you'll excuse the hubris) computer scientist that dabbles in CG, so I can at least extrapolate the concepts. Is it possible the number of draw calls would cause an issue with older 512MB GPUs? FFXIV has some of the most non-static objects per scene I've seen in any game I can think of, even counting MMOs. That could be one of the things limiting their choices for texture size.

Quote Originally Posted by Sumii View Post
Not necessarily on topic, but I saw this and wanted to make mention of something.

Minimum specs requirement should have nothing to do with the DX11 client. They are not changing the DX9 client over to DX11 and dropping it, but rather giving customers the option to run a more graphic intensive client. If someone wants to complain about not being able to use the DX11 client when they have the DX9 client, that is all on the consumer to upgrade their hardware. If SE was replacing the DX9 client with DX11 that'd be something else entirely, but Yoshi has said that it will be a whole other release client.
I suspect you aren't appreciated the complexity change involved here. When I made that comment, I had forgotten they were using deferred lighting. If, as I thought, they were using some form of vertex lighting, even just doubling the size of the textures would result in 1/64 the performance of the lighting. That would cause even the best GPUs a little concern. I said minimum to be generous, truthfully, that kind of naive change would throttle virtually any system. But it's a moot point, because fortunately, they are using deferred lighting, for all of it's foibles, it is very fast.