Holy shit ! Rendering with atoms !

meh not bying it.

Quote from Rick_D

interesting comment on youtube, that these points have to be assigned locations, so each point's position along with colour information, will take up hundreds of gigabytes. unless they have a way around this then that is probably one of the biggest problems.

they're probably not loading all those points at once, since it would be the same as loading all half life 2 levels at once. i think so far the biggest problem is that the initialization will take ages if they are to sort the visual information tree, but then again they seem to be doing ok so far. i don't doubt for a second that this could work in the future.

Quote from Bunglo

Let me know when an animator tries working with this...

As far as i know, you can't animate it

i still cant decide if this is "real" or somebody making fun of "them" or if "they" are making fun of "us"

Professional developer trolls?

Meanwhile at Naughty dog, people are making wonderfull ans real game with a PS3.

I'm pretty sure they're trolling.

let's just watch their next video instead of debating whether it's possible or not. if it's a scam, it sure seems like an elaborate one, and given that it's government funded... it just doesn't seem very likely.

Quote from Sentura

if it's a scam, it sure seems like an elaborate one, and given that it's government funded... it just doesn't seem very likely.

Ignorance or troll? You decide.

Quote from Sentura

let's just watch their next video

see you in a year then

Notch brings up a similar thought that I had on watching the video: Even with each dot simply being coordinates and color, the "atom" density they're claiming would lead to absolutely enormous data sizes. Look at how much a drag on performance IO is for Minecraft, and then imagine every block is one "atom" and there are trillions of them instead.

Maybe they could claim to be saving on these costs by making everything hollow, idk. But then the point of WHY you'd want this rendering style over polygons and textures and shaders, and all the wonderful things we've come up with to fake detail using those elements comes immediately to mind...

it's quite simple, make an actual game that works on new technology and people will start to take interest!

Quote from Jetsetlemming

Notch brings up a similar thought that I had on watching the video: Even with each dot simply being coordinates and color, the "atom" density they're claiming would lead to absolutely enormous data sizes.

I think that's kind of the point of this technology isn't it? Floating point data is far from a new concept, it's their increasingly efficient algorithms for using that data without slowdown and huge amounts of overhead that're causing a stir — plus an proper SDK for pulling it all together.

I personally think this has a very good chance of being the future. At the moment there're naturally going to be countless naysayers who don't like the idea of having to change the status quo, but this technology is quite clearly better than the current approach. Things like LODing, pop-up, and having to waste huge amounts of time optimising for crappy hardware are suddenly no longer an issue. Detail can now be as sophisticated as you're capable of modelling.

Obviously the tools are nowhere near usable for making games, but the sheer amount of money going into this project tells me they're very serious about doing everything required to make it something developers can quite readily start experimenting with and ultimately embracing. Even since last year they've managed to get direct importing from 3D Studio Max or Maya implemented, meaning modellers hardly have to change what they do at all — apart from not having to spend ages trimming polies off that is. It's inevitable that similar bridges will be built for animation, audio, shaders, etc. They just need time.

The real question is: why fake it when you don't have to? That's a question developers will seriously start asking themselves in the near future IMO, especially as the prospect of jumping ahead five+ generations of graphics starts sinking in.

^ word

I remain skeptical until they demonstrate the technology in a real-time animated game situation instead of a static environment with millions of copy pasted meshes.

Quote from Thrik

I think that's kind of the point of this technology isn't it? Floating point data is far from a new concept, it's their increasingly efficient algorithms for using that data without slowdown and huge amounts of overhead that're causing a stir — plus an proper SDK for pulling it all together.

I personally think this has a very good chance of being the future. At the moment there're naturally going to be countless naysayers who don't like the idea of having to change the status quo, but this technology is quite clearly better than the current approach. Things like LODing, pop-up, and having to waste huge amounts of time optimising for crappy hardware are suddenly no longer an issue. Detail can now be as sophisticated as you're capable of modelling.

Obviously the tools are nowhere near usable for making games, but the sheer amount of money going into this project tells me they're very serious about doing everything required to make it something developers can quite readily start experimenting with and ultimately embracing. Even since last year they've managed to get direct importing from 3D Studio Max or Maya implemented, meaning modellers hardly have to change what they do at all — apart from not having to spend ages trimming polies off that is. It's inevitable that similar bridges will be built for animation, audio, shaders, etc. They just need time.

The real question is: why fake it when you don't have to? That's a question developers will seriously start asking themselves in the near future IMO, especially as the prospect of jumping ahead five+ generations of graphics starts sinking in.

Some of the ideas presented, like having each pixel on the display being specifically looked up and rendered, rather than the entire world being rendered and then translated into a display, seem really smart, and related to raytracing concepts I've heard before. But that could possibly work with polygons...

thats essentially what deferred rendering is

The question is :

At the maximum level of detail required for a game - the point where the artists no longer have to think/care about poly count - which technology will perform better? Polygon or this?

I'd be very surprised if current-gen hardware can handle this technology and look as good as "traditional" tech. But maybe in a few years this will become more feasible.