Post by
MinisterofDOOM »
https://forums.nicoclub.com/ministerofdoom-u16506.html
Tue Mar 23, 2010 6:39 pm
Yep. Even guys with top-end CPUs are limited by their graphics hardware now. I have an i7 920 but it's only backed by a single gtx260 so I still only get about 50fps in Crysis at "high" (not very high) settings.
I remember back in the day when my gaming rig was a P1 200MHz with 64MB of RAM and a Diamond Stealth 2000 Pro (8MB Video RAM!!!!!) I could run games like RA2 and DII:LoD even though I didn't meet requirements processor-wise because I had so much RAM. And the Stealth took a lot of the load for D2 off the processor.
As for this "unlimited detail" thing...it sounds fishy to me. It sounds wonderful when you talk about only having to render what's visible, thus only ever the screen resolution's worth of pixels on a per-pixel basis. But it cannot be that simple. It isn't used for rendering stills or slideshows. It's used to render real-time dynamically generated 3D content. Stuff changes. Stuff EXISTS outside the FOV. You still have to keep track of that s***, you still have to render the effects of that s*** (shadows, pullets, cast light, etc.). So yeah, as a FINAL DRAW PASS that tech is grand. But the final draw isn't what consumes processing power. It's the computation of what is what does what to what and why. Just because you're not directly drawing something doesn't mean it isn't using computational power.
Really, all this sounds like to me is just neutered voxels. You can't just ignore stuff that isn't visible, ESPECIALLY occluded objects within the FOV. That stuff still "exists" and still has cause-and-effect relationships with everything else (in and out of view). So the "search engine" explanation they give is crap. The engine might be able to search for what to DRAW but everything else still exists. The draw pass is all that's seeing any improvement.
Plus, they don't say ANYTHING about how it handles shadows, light sources, dynamic lighting, etc. At least not that I could see.
In the end, you can make up lots of magic terms for stuff, but there's more to graphics than just drawing the 1440x900 (or howevermany) pixels the player sees in a given frame. If that was all that was necessary, we'd already be in a GPU-less world. Plus, that idea is not new and NOT exclusive to unlimited detail (or exclusive from traditional polygonal rendering).
Some neat ideas, but it isn't the magic solution the poshly-accented narrator fellow makes it out to be. It does not replace all that processing. The stuff the "search engine" algorithm searches for still has to come from somewhere, and THAT takes processing power.
At any rate, if you watch those videos, the framerate during the UD demo scenes is HORRENDOUS. Certainly not adequate for gaming. And those are static scenes.
With more info, it might be possible to convince me there's something to this, but right now it sounds to me just like the 100mpg carburetor does: too good to be true.