Little Australian company develop way to replace polygons with atoms while making graphical objects.
Look's great. I like the idea of scanning in real world objects.
I turned off as his voice is getting right on my nerves
and amazed this is truely incredible! Now that is a huge leap in visuals!
I wonder if they would consider releasing a the island level as a demo we could fly though and see running in real time. I will send them an email and ask.
Wow, that was awesome. Would need to see a real-time demo though, but if they are actually getting this stuff accomplished, Next-Gen will be REALLY interesting.
Unfortunately, just like their previous videos, they fail to mention the things the engine CAN'T do. Because obviously they would like to get investors. This system won't be used anytime soon, due to these things he doesn't talk about. Like shadows for instance. You cannot do dynamic shadows/lighting in real-time when dealing with such massive amounts of point cloud data. It's too much point data to search through on each frame to update lighting. Which brings me to the second point, animation/physics. As with the lighting, it's too much point data to crunch each frame for an object to be animated, which is why you never see anything moving in the video, it's all camera movement. And thus my final point, file size. The amount of point cloud data that would be needed for a retail game with many levels would require many blu-ray discs. Games wouldn't be able to have anywhere near the amount of "unlimited detail" he shows off in the video. But it would still be better detail than todays console games. But then you are taking a massive step back in lighting and gameplay quality merely for more geometric detail. But luckily DX11's hardware accelerated tessellation allows for polygons own type of "unlimited detail", since the tessellation can be controlled dynamically, having the amount of detail increase the closer you get to the object. Geometry is subdivided the closer you get to it, allowing a displacement map to bring out more detail. And since it's all done with a few special tessellation cores built into DX11 supporting GPU's, the geometry is created on the fly within memory, so the performance hit isn't ANYWHERE near what it would be if you actually imported a mesh of that high tessellation. Plus the tessellation has no problem with the objects being animated or having physics, as well as lighting/shadows. Since it's still the same geometry, merely subdivided/tessellated.
When I have emailed them in the past about some of the issues they replied with some vague answers last year. They claim they can do animation with their system. I don't understand how but they claim they can. Linked to http://www.mediafire.com/fi... As for file size "MEMORY STORAGE: The Unlimited Detail rendering technology does not require significant system memory to run. The geometry may be compressed, instanced and streamed to allow the best mix of runtime and disk based storage as well as support unique geometry." And also "SHADOWS: Unlimited Detail will have both real-time shadows and pre-calculated lighting as well as a variety of advanced lighting options." More bold claims with little actual details.
Yeah haha. The only thing the memory storage says is that it's working as any octree structured pointcloud should, streaming in only the detail that needs to be... That data has to be decompressed when brought into memory anyway. They will still be facing the same problems similar implementations face. The Atomontage engine seems to be far more advanced than theirs already. http://www.youtube.com/user...
N4G is a community of gamers posting and discussing the latest game news. It’s part of NewsBoiler, a network of social news sites covering today’s pop culture.