Following its original reveal a year ago, Brisbane-based technology company Euclideon has released a new video of its engine, which purports to show off progress it has made - and the "largest breakthrough since 3D graphics began".
The results are remarkable. The camera zooms into the company's logo, revealing it to be made of a one kilometre square world of trees, rock and ruined ancient buildings, and then it shows the ground to be made from individual clods of dirt. It consists, the video claims, of 21,062,352,435,000 polygons, and it runs at 20 frames a second.
Quite how it works isn't yet clear. CEO Bruce Robert Dell refers to polygon shapes having to be converted into point cloud data and says that instead of using polygons the engine builds the world from 'little 3D atoms', a process that previously took up a lot of processing power but his tech, he claims, is able to run unlimited numbers of them in real time.
"If what we've said is true, then it is the largest breakthrough since 3D graphics began," Dell says, suggesting that realtime graphics companies are increasing poly counts by about 25 per cent a year at present.
"If any of these large companies were to suddenly come out with ten times more polygons than their competitors, it would be enormous news. But we didn't increase the geometry count by ten times, or 100 times, or 1000 times. We increased it so far that we could abandon polygons altogether and move to little atoms, and run them in unlimited quantities."
The video is presumably designed to bait investor interest in Euclideon, which was founded last year. According to its website, it has a staff of nine and is to hire more "soon". In March the Australian government granted it nearly $2 million Australian dollars (£1.3 million; US$2.2 million).
Is it real? Id's John Carmack seems to think so, according to his Twitter account. But that doesn't mean games will use it any time soon:
"Re Euclideon, no chance of a game on current gen systems, but maybe several years from now. Production issues will be challenging."
Euclideon's video, however, claims that artists can easily import their 3D Studio Max models into the engine, and can also scan real-world objects into it - with no reduction in fidelity because of the engine's "unlimited detail" nature. But certainly, a transfer from one technology paradigm - polygons - to another would represent a major shift for the industry.
Next, Dell says, his team will work on lighting, which is very flat in the new video. It will also release a software development kit "some months from now" and hand it to game developers.
What do you think? Real or fantasy? Vapourware or just impractical? Let us know in the comments.
Comments
14Cheesus Christ! Gimme!
Infinite Polygon Engine waffle.
All credit to them for thinking outside the box. I hope it works out. The approach seems not dissimilar to the old voxel engines.
Loyd Grossman sure talks the talk...
It is the old voxel, but in the future. Imagine China full of sweat shops producing high-res COD forests and Crysis buildings.
From what I've read around the net about this tech, it seems animation is the greatest challenge here, and supposedly very difficult to do right.
Anyone with some insight on this?
Okay, so by the time we got into the part where the compare ground in this "unlimited" engine to current floors-- that's all good and great. WE CAN make beautiful THINGS. Wow. We can do that using existing engines, too. WHAT WE CAN'T DO IS WASTE THE TIME AND MONEY MAKING THOSE TINY THINGS NO ONE LOOKS AT.
Not trying to be a downer but how many hours and money should invested in a game to make a complex 3D atom world if the inventor of the engine alone only made some repetitive trees, rocks and cactous to promote it?
I have no idea how or if this could work, but if it does as a gamer this could be one very large step towards photorealism
I disbelieve.
I mean I believe what I'm seeing, but I am also seeing tons of replicated structures and I don't think that's because 'we're not artists'. It's because at their claimed voxel (sorry, 'atom') density you'd need petabytes of data to hold those landscapes. Now perhaps you can go back to the tiling solution here and have 12 ground squares you mix and match, for instance. But this has a lot more tradeoffs than they're letting on, and again not just because 'we're not artists.' How much movement did you see, for instance?
Edit: In addition to 3d replication (tiling), everything could be procedural - which brings its own set of issues but would mean lower memory requirements. We need to see something with more terrain/object variety, even if it's still ugly.
That voice over is extremely irritating.
It does look incredibly detalied, but it also looks incredibly static. How movement will be incorporated, such as real-time lighting and physics, could be a deal breaker.
Polygons were wearing a little thin. Win.
Could somebody combine this with a current engine technology, and use it to show only the part of the world that is static to the player?
That's awesome, can't wait to explore those worlds!
Meanwhile a downloadable demo along this line of thinking: http://albertoven.com/2011/09/08/colorado-interactive-raytraced-unlimited-geometry/