Parhelia - any bandwidth-saving features?

[diewlei]

once and for all can someone tell me that?

any so-called occlusion - culling, deferred rendering, crossbar crap kind of techniques?

i know fast z-clears, hardware displacement mapping and those tesselation, advanced depth whatever stuff are supported in hardware, but are these categorised under "bandwidth-saving"?

thank you very much.

[Wombat]

once and for all can someone tell me that?

No, they can't tell you yet.

[Kruzin]

It's all in your point of view.
The features you listed are concidered by many to be bandwidth saving techniques.
nVidiots and fanATIcs would disagree (untill it's available in their cards)

[TdB]

Isn´t 16xfaa a "kind" of deferred antialiasing algorithm, that saves alot of bandwith?
I think that is the next-coolest bandwith saving tech i have heard of, tilebased rendering is still the coolest, but 16xfaa is a close second.

[diewlei]

[Pace]

I can tell you.

"Once and for all"

There, happy?

[DGhost]

funny thing is that i think it was anandtech that found that of the bandwidth saving features on the Radeon 8500, Fast Z-Clears was the one that made the largest difference, and the others were pretty minimal.

the "crossbar" memory architecture that NVidia is always talking about is nothing new... the G400 had a similar memory controller iirc.

the other thing is that it looks more like Matrox's core is designed to work more intelligently than others, eliminating extra work when possible. that in and of itself could be a bandwidth saving feature, of sorts...

why is everyone so worried about having bandwidth saving features when it has like twice the bandwidth of a GF4 anyways...

[typedef enum]

I think the reason is obvious: having twice the bandwidth may have diminishing returns *if* there isn't any sort of logic in place to try and preserve/save BW.

In other words, one might ask the question: Is it better to save transistors and go with a wider bus architecture, or better to spend a considerable amount of R&D on the memory interface/architecture and use a less-wide bus? In the end, do things, more or less, cancel themselves out?

I think we will finally see for ourselves tomorrow...Personally, I think the wider bus is more attractive, but it also has the side-effect of being a more expensive option as well...

[diewlei]

Originally posted by Pace
I can tell you.

"Once and for all"

There, happy?

thanks for proving yourself. have a nice day.