Jump to content

Voxels, voxels everywhere..


Recommended Posts

The list of impressive voxel games/engines is getting longer each day.

 

And all I can conclude is that for a game studio being 100% focused on making a voxel game, NQ is really bad at voxels..

 

Link to comment
Share on other sites

  

2 hours ago, CptLoRes said:

The list of impressive voxel games/engines is getting longer each day.

 

And all I can conclude is that for a game studio being 100% focused on making a voxel game, NQ is really bad at voxels..

 

 

 

how would something like this work if construct voxels turn into a mesh when you aren't in build mode?

would elements work on an engine like this?

how much data is transferred to the client to load the voxels?

can low spec computers handle this engine? how well does this engine work with moving constructs?

does it work with terrain the size of planets?

what happens to the square voxels if there are ships which aren't aligned to the grid and are slightly rotated?

do you have to re-render the entire model if you modify it? how do collisions work?

how much developer time is wasted by using this voxel system, compared to a much simpler one?

Link to comment
Share on other sites

The point is that (by their own admission) NQ is using 20 year old voxel technology and has to really jump through hoops to make it hold up.

You can't take what NQ does as the benchmark when their engine is using what really is ancient if not archaic technology.

 

The developer is not claiming this is a game engine(yet), that is not the point. It does show a modern, current application of voxel technology. 

 

Frankly, for what NQ seems to want to achieve from a game perspective, using freeform voxels is probably the worst choice you can make and while I'd commend NQ for being bold enough to try, it's becoming more obvious every day that the hurdles to overcome are high and many.

 

Link to comment
Share on other sites

Another core feature for most modern voxel engines is GPU acceleration (the voxel matrix math is almost a perfect match), and again this is something NQ does not have.

 

It is amazing how much more powerful a GPU is compared to CPU, provided the problem is a good fit for the GPU architecture. And using modern GPU's the only practical restriction on the amount of voxels shown in a mmo, should be the streaming bandwidth. And using GPU's there should not be a need to meshify constructs for performance (other then bandwidth) reasons.

 

And in my book this is just another one of those "let's just make it work using the quickest and easiest to develop method first, and make it better later" features in DU becoming the permanent solution. And it is disturbing how little actual progress NQ is showing, even using quick and dirty methods. And while NQ is fumbling around, a single person made the tear down game from scratch including the physics based voxel engine in a 2-3 years time frame. Imagine if NQ (68 employees at the moment) had 20-30 people working at that pace.

Link to comment
Share on other sites

1 hour ago, CptLoRes said:

Another core feature for most modern voxel engines is GPU acceleration (the voxel matrix math is almost a perfect match), and again this is something NQ does not have.

 

It is amazing how much more powerful a GPU is compared to CPU, provided the problem is a good fit for the GPU architecture. And using modern GPU's the only practical restriction on the amount of voxels shown in a mmo, should be the streaming bandwidth. And using GPU's there should not be a need to meshify constructs for performance (other then bandwidth) reasons.

 

And in my book this is just another one of those "let's just make it work using the quickest and easiest to develop method first, and make it better later" features in DU becoming the permanent solution. And it is disturbing how little actual progress NQ is showing, even using quick and dirty methods. And while NQ is fumbling around, a single person made the tear down game from scratch including the physics based voxel engine in a 2-3 years time frame. Imagine if NQ (68 employees at the moment) had 20-30 people working at that pace.

Why do you think DU doesn't use the GPU in its voxel engine? I seem to recall a dev blog from a year ago that indicated that it does, but that devblog seems to be no longer available. I did find this link which supports my recollection of GPU use: https://www.mmorpg.com/news/dual-universe-dev-blog-looks-at-tech-optimization-2000117365

Link to comment
Share on other sites

On 5/13/2021 at 5:47 PM, blazemonger said:

The point is that (by their own admission) NQ is using 20 year old voxel technology and has to really jump through hoops to make it hold up.

You can't take what NQ does as the benchmark when their engine is using what really is ancient if not archaic technology.

 

The developer is not claiming this is a game engine(yet), that is not the point. It does show a modern, current application of voxel technology. 

 

Frankly, for what NQ seems to want to achieve from a game perspective, using freeform voxels is probably the worst choice you can make and while I'd commend NQ for being bold enough to try, it's becoming more obvious every day that the hurdles to overcome are high and many.

 

 

 

I'm not sure if i understand how this would have been a better choice for DU.

 

Those look like much smaller voxels.  And it looks like the entire world is built from them.  So basically Minecraft with smaller cubes.

 

Would making the voxels smaller but immutable be better for DU?  It doesn't seem like that would be more efficient for creating an entire solar system/galaxy.

 

It's clearly better for creating trees and flowers.  but would trees and flowers that are made of voxels improve DU?

 

It would be cool to mess with all those voxel physics.  I'm sure ship damage could be more realistic using that tech.  And if someone makes a game out of whatever that is in the video, i would love to check it out. 

 

But i don't see how it would improve DU.

 

 

Link to comment
Share on other sites

 

Have you noticed how terraforming work here in comparison to DU? 

 

Have you considered what DU _could_ look like if Voxels were given properties like lift, drag, being un-obstructive to thrust, reactive to thrust (by glowing), being translucent, transparent or fluid? 

 

Different voxel types combined see their properties make then become engines, tanks, desks, chairs, etc.. 

 

I could see a lot of ways in which this tech could improve the experience and the uniqueness of DU

Link to comment
Share on other sites

I think there is a better chance of Mr Lin reproducing DU in his engine than NQ incorporating this tech. 

 

That said, it really would be amazing. More natural terrain deformation and mining, trees that aren't invincible etc... 

Link to comment
Share on other sites

Another example from 2012, demonstrating the lackluster voxel performance in DU. This entire demo is shown using a GTX480 which is something like maybe 60% as fast as a GTX1050 would be today.

 

And again this shows that the only bottleneck for voxel performance in DU should have been streaming bandwidth. And what you currently see rendered on the screen visiting District 6 etc should be trivial for any modern GPU.

 

 

Link to comment
Share on other sites

we'll that's not entirely true. You see the road from voxel grid to raster output is quite straightforward when working with plain cube primitive voxels, but that pretty much always results in that lego like midcraft-esque output. Now with marching cubes algorythms things can get much prettier but also much more complicated very fast. Especially if you need the end result to have reasonable number of verts and workable topology.  Even otherwise straightfward things like texture stitching can easily become a sourse of great headache.  That's why you seldom see it in indie tech demos.  Now that doesn't mean it can't be done right and done efficiently. There are quite a few examples of games that succeeded to the point that you may not even realize their terrain data is based on voxels. 

 

And as the saying goes, there's always more than one way to skin a cat.

One thing i could never undestand about DU in particular is why do the whole planet in voxels.  I mean it was fun digging  X km down for the fun of it, but there's no ore to dig beyond certain depth so why bother. It just makes the dataset insanely large.  I don't think anyone would compain if max depth to dig was let's  say 1,5 km.  

 

The rest of the way down could have been a huge spherical cube, a simple efficient and easy to map to primitive.  Planets could have been same size as now but the data density would have been orders of magnitude smaller. You could even partition entire voxel grid arrays onto its topology.  It's usually the hybrid solutions that win in the end. 

 

 

 

 

 

 

 

Link to comment
Share on other sites

  • 3 weeks later...
On 5/14/2021 at 2:55 PM, JayleBreak said:

Why do you think DU doesn't use the GPU in its voxel engine? I seem to recall a dev blog from a year ago that indicated that it does, but that devblog seems to be no longer available. I did find this link which supports my recollection of GPU use: https://www.mmorpg.com/news/dual-universe-dev-blog-looks-at-tech-optimization-2000117365

A copy of the dev blog without the images : https://fasrtraveler696.weebly.com/blog/dual-universe-trailers

 

Quote

GPU culling
Dual Universe features highly-detailed environments, filled with non-interactive items we call ‘decors’. These include things like grass, trees, small rocks, and more. The engine we use for the game, Unigine, calculates the geometry of these items—usually in the tens of thousands at any given time—using the computer’s CPU, and decides which are more important. For example, items that are behind the player or less visible (i.e. somewhere in the far-off background) will not be rendered. This is particularly important in more dense environments, such as forests. This process is known as ‘culling’, and our team of developers have coded the engine to utilize the computer’s GPU to do these calculations instead of the CPU. Considering it’s already responsible for rendering these decors, this will significantly improve how fast these calculations are produced. What this means for the gamer is improved framerates and more detailed environments.

 

edit:

NQ uses Unigine to process the voxels procedurally generated, there is no bigger dataset for bigger planet, at least for the core terrain.

The additionnal data for voxels are all additions that NQ did on top of the core generation (here is a video about their editor), and all players modifications to the terrain. Add to that all the constructs with elements. All of that is stored on the RAM, the more you have the more detailed the world is.

To faster this process, the game client stores data locally in a cache folder. It checks for deletions or additions of the voxels with the server and avoid redownloading every voxels data for the place where the player is.

As DU is not a single player game, all of this data must be synchronized between players, but not only voxels. All positions and orientations of constructs, all elements states (broken elements, lights, engines, force fields, etc), must be known to be correctly displayed and in sync with each player.

The CPU generates optimized meshes with the desired LOD for the terrain and constructs, then the relevent data is sent to the GPU to display everything. As said above, the culling is done on the GPU instead of the CPU.

 

I might be not right on all aspects of the game, but at the end, it needs so much hardware to work decently that I'm concerned by the targeted player base. Not everyone can run this game decently. Add to that the niche game design, at a time where fortnite and short game sessions and fast consumption of games dominate the gaming. But the potential of DU is immence, so believe in it :D

 

To speak of NQ as a compagny where not every employees are devs, I wouldn't compare a prototype voxel engine (as good as it is) or other voxel experiments to a functionning (as far as it is) multiplayer game, with an immence single shard world and all of its new techs (client and server).

Link to comment
Share on other sites

  • 2 weeks later...
  • 4 weeks later...
On 5/28/2021 at 4:37 PM, Bazzy_505 said:

we'll that's not entirely true. You see the road from voxel grid to raster output is quite straightforward when working with plain cube primitive voxels, but that pretty much always results in that lego like midcraft-esque output. Now with marching cubes algorythms things can get much prettier but also much more complicated very fast. Especially if you need the end result to have reasonable number of verts and workable topology.  Even otherwise straightfward things like texture stitching can easily become a sourse of great headache.  That's why you seldom see it in indie tech demos.  Now that doesn't mean it can't be done right and done efficiently. There are quite a few examples of games that succeeded to the point that you may not even realize their terrain data is based on voxels. 

 

And as the saying goes, there's always more than one way to skin a cat.

One thing i could never undestand about DU in particular is why do the whole planet in voxels.  I mean it was fun digging  X km down for the fun of it, but there's no ore to dig beyond certain depth so why bother. It just makes the dataset insanely large.  I don't think anyone would compain if max depth to dig was let's  say 1,5 km.  

 

The rest of the way down could have been a huge spherical cube, a simple efficient and easy to map to primitive.  Planets could have been same size as now but the data density would have been orders of magnitude smaller. You could even partition entire voxel grid arrays onto its topology.  It's usually the hybrid solutions that win in the end. 

 

 

 

 

 

 

 

That I also don't understand. Maybe they re even removin planetary mining now. I agree a limit of 1,5-2 km wouldn't have really limited the experience. If it helped the performance I think everyone would have supported it.

Link to comment
Share on other sites

  • 3 weeks later...

I own teardown and it's cool but actually not that impressive once you play it. Smashing things with hammer allways feel the same, only metals feel like more than a voxel you can smash into smaller voxels... hell some structures will even levitate in the air instead of breaking if you remove everything that touches the floor.

 

It's really really cool but it gets old very quick. And I don't see it ever running in multiplayer, even on my highend pc its trivial to drop the framerates, just put on a fire and throw a handful of bombs in quick succession.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...