I'm Jon Olick. I make shiny things. I simplify.

I presented Sparse Voxel Octrees at Siggraph 2008.

Wednesday, July 8, 2009

Voxel Performance and Requirements

I read lots of papers on voxel rendering techniques and such describing that they achieved 60fps without specifying some of the most important factors.

1) Image Resolution -- This is obviously important, yet some papers seem to magically forget it. It has to run at 60fps or more at HD (720p and 1080p) resolutions in order to be considered for video games.
2) Voxel Resolution -- IE, how deep is the tree? This is very important as each level of the tree adds one more iteration to your traversal loop for every pixel and as such significantly affects performance.
3) Data Transfer Amount / Speed -- This may not be obvious, but it does play an important role. The GPU can do behind the scenes transfers from system ram to video ram, but the bandwidth is only good if you do it in large chunks (many sequential MB). This plays an important role in performance.
4) Cache Responsiveness -- This one shows its ugly head when you look at a scene that you have never seen before. If the cache is not responsive enough, the coarseness of the unloaded oct-tree will show its ugly head. This is primarily more of a quality issue.

Also, I think its important to mention what the system requirements were to run smoothly at those resolutions.
1) How much Video RAM does it consume? If you require more than 2gb of vram SVO/Voxel data, its not realistic for next gen consoles (720, PS4). Even 2gb may be pushing it. Keeping the requirements under 1.5gb is preferable.
2) How much System RAM does it consume? If you have to store the entire volume set in RAM. This is not realistic as most volume sets of a real game will be in the many gigabytes range.
3) What video card were the results computed on? dual GTX 295's may be a sign that the technology is unreasonably expensive.

I would really like it if all voxel papers in the future would mention these things, because its very hard to evaluate otherwise.

Friday, July 3, 2009

4/8/2008 - The Future of Piracy

When somebody mentions 3D printers, some may think this is the first step towards a star trek replicator. They are probably right. 3D printers will change the world as we know it. Mostly they will usher in a new era of open source hardware and piracy.

Imagine a world where you can download a design for a new ipod, and print it out within seconds.

Imagine a world where there is a thriving open hardware movement to complement today's open software movement.

A bit farther off, imagine downloading a recipe for an entree from a five star restaurant, and then print it in front of you.

What would happen to all the jobs built around the concept of design once, sell many... especially when a hardware design can easily be pirated?
Food and Clothing jobs will eventually change significantly, but chefs and fashion designers will remain for the foreseeable future.
Entertainment will not change a whole lot and still be complaining about piracy.

What would happen to FedEX and UPS?
Shipping services would most likely be relegated to raw materials and perhaps goods made with volatile materials.

How far off is this?
It is hard to tell... but a decent guess would be within 50 years, but it could happen as early as 30 years. You will start seeing the signs of these trends in less than 15 years, although you can see them right now if you look closely enough.

So lets look at the printing of food for a bit and take a few logical guesses about the future here. The thought of whipping out your HP printer and printing out a nice juicy New York Steak and Eggs is probably not going to happen, at least not in the form it exists today. However, the machines could build "flavor" molecules which act as spices or when heated and with water could produce various textures of certain types of food. These machines would more be of the simulation of foods rather than the production of Angus Beef.

Why would anybody want a simulation of a steak rather than the steak itself?
When humanity expands out into space, those space dwellers aren't going to have the amount of land necessary to support cows. Its simply too expensive. However, having one of these fancy simulation machines can trick the person into believing he is eating a Kobe Steak.

So in conclusion, would this technology change the world for the better? I think so. It allows people to experience the entire range of life, and for very little money. Anything that improves the lives of the entire human race, I think is a good technology to have. Even though it can and surely will be misused, its potential for good is too great to ignore.

Thursday, July 2, 2009