Jump to content

DEVBLOG: PRECISION IN BUILDING - discussion thread


 Share

Recommended Posts

16 minutes ago, Tordan said:

I am unconcerned. even the BIG wedge maxes out at 95% complexity.  my stairs, ladders, max out at 85% and for those litterally NONE of the voxels are actually wherethey look like they are.

A ship that I've built, which isn't even really pushing the limits of detail, has a few sections that are beyond 100%. https://imgur.com/a/4Klkmas I predict that it will be a significant limiting concern once we have the VPT. I definitely think I'll have problems with it. Why should we be limited in our creative projects for the sake of a few edge cases that have extreme voxel complexity? IMO they either need to raise the complexity limit or get rid of it.

If NQ thinks it's absolutely necessary, I sure would like to hear the technical reasoning. This is a big deal for a detail-oriented ship builder.

Link to comment
Share on other sites

13 minutes ago, TobiwanKenobi said:

Yes - it is absolutely essential that we be allowed to extended vertices 1.5 voxel away from zeroed. If they limit us to 1 voxel in the future, it will break so many amazing builds and make many feats of voxelmancy impossible. And for what? Why would we need to limit this?

 

Im actually all for full complexity im just pointing out the flood of ore Post Demeter as well as piling on more ore with Panacea.

 

IMO as long as the Vertex Complexity tool is an addition to the voxel building arsnel and not replacing any of the other existing tools im cool with that. While it would be nice for it to paper thin micro voxels im cool if the VCTool helps streamline more of the building process and still has a degree of voxelmancy required to add to the voxel complexity. Personally I hope that anything built with the VCT is non voxel reactive so that it can be paired easily with cutting and splicing microvoxel work into build to where it can be made more seamless in nature like non reactio.ns between landscape voxels and voxels as a third class of voxels so that it streamlines and creates things quickly with the VCT and the real complexity is done still through voxelmancy as the other half of it even if the VCT makes reactive voxels its still worth using.

Link to comment
Share on other sites

18 minutes ago, eviltek2099 said:

The blog said nothing about reducing the move limit of a point! you will still be able to move a point 1.5 vx in any direction

"And, in the example of these four cubes, if we were to go to a value above 84, it would result in an ill-formed shape because the blue shape would have a negative volume. This would probably create visual artifacts, and we may prevent this situation in the future."

They're saying that if you go beyond one voxel away from vertex origin, it can overlap the rendered faces of the resulting voxels, leading to an ugly inside-out mess. Which is bad of course, but easily dealt with and not worth limiting our voxel positions in such a hugely destructive way.

Link to comment
Share on other sites

33 minutes ago, TobiwanKenobi said:

A ship that I've built, which isn't even really pushing the limits of detail, has a few sections that are beyond 100%. https://imgur.com/a/4Klkmas I predict that it will be a significant limiting concern once we have the VPT. I definitely think I'll have problems with it. Why should we be limited in our creative projects for the sake of a few edge cases that have extreme voxel complexity? IMO they either need to raise the complexity limit or get rid of it.

If NQ thinks it's absolutely necessary, I sure would like to hear the technical reasoning. This is a big deal for a detail-oriented ship builder.

I know we're far off topic, But I haven't talked to a single ship builder that hasn't surpassed 100% on at least one build yet. It doesn't seem to be limited to voxel builders. Even primarily basic tool users get over 100%. What we have decided is it's material changes, and air gaps that push that number up. So it isn't at all limited to a few constructs as they said. Nearly every ship builder has gone over 100% somewhere. I wish they gave more than a 20 second commentary 20 minutes into a 2 hour video on the subject. Or just not included it in the game without any real explanation other than, "don't worry about it. It only effects a hand full of people and constructs"

 

Oh I didn't realize who I was talking to. HA. Yep you are one of the many builders I have talked to. I don't want to drop other names. But even people that don't call themselves ship builders get this. Industry guys making non sale industry ships have gone over 100%.

Link to comment
Share on other sites

19 minutes ago, TobiwanKenobi said:

ship that I've built, which isn't even really pushing the limits of detail, has a few sections that are beyond 100%

I have noticed that too many different textures (or overlapping textures) will cause this. be carefull not to overlap textures into the same space.

Link to comment
Share on other sites

@NQ-Ligo

Will Panacea be going to the PTS? I feel such major changes need a bit of testing first.

 

I also hope that the team is considering adding edge, face, and full-voxel dragging to the VPT in the future. Dragging individual vertices is nice, but being able to drag multiple at the same time would really speed up some processes.

Link to comment
Share on other sites

2 hours ago, eviltek2099 said:

The blog said nothing about reducing the move limit of a point! you will still be able to move a point 1.5 vx in any direction

 

I think this is what they are talking about.

 

Quote

And, in the example of these four cubes, if we were to go to a value above 84, it would result in an ill-formed shape because the blue shape would have a negative volume. This would probably create visual artifacts, and we may prevent this situation in the future.

 

I think they are saying they might not allow us to apply a texture to a voxel that has a negative volume.

 

I hope they forget about that though.  It's not really hurting anyone as far as i know.

 

And there are a lot of situations where a shape has a very slight negative portion.  Like the inside of very small curves.  It would be a shame to have a limit like that get in the way of using those shapes.

 

 

 

 

Link to comment
Share on other sites

6 hours ago, eviltek2099 said:

The blog said nothing about reducing the move limit of a point! you will still be able to move a point 1.5 vx in any direction

They are like Edgar Allen Poe's monkey paw for each whish 3 bad things are going to happen. 

Link to comment
Share on other sites

7 hours ago, Warlander said:

There will come a time when the costs of people filling cores with mass amounts of voxels will start to escalate in sever costs since instead of mining players are just filling up the surface/subsurface with voxels and various degrees of voxel complexity that some bean counter at NQ will impose it or force players to simplify their builds. You can count on that.

Remember when NQ used to say DU was a open world civ-rebuild game and they wanted us to build as much and as big as possible. And when we said are you sure you can handle that? They said either nothing or "don't worry guys" for 4+ years..

 

So while what you say is probably true and likely to happen, it will also means the main premise for DU will be null and void and that NQ never managed to deliver on the tech needed for the game they sold us..

 

 

Link to comment
Share on other sites

My little feedback:

I'm not a creative person, but I'm really happy with the arrival of this tool. This will save considerable time. Too bad if everyone will be able to do like me now. It's a good thing for the whole community that the mastery of voxelomancy is no longer reserved for the few people who can spend hundreds of hours on it.

For the complexity, I worked on an L core with a lot of detail, I could easily correct the places that exceed 100% on the chunks. I'm not worried about this limitation. I already did what I wanted with my voxels, with the tool there won't be any more.

The new steps will require a bit of adaptation, we'll see what we can do with them.

Link to comment
Share on other sites

1 hour ago, CptLoRes said:

Remember when NQ used to say DU was a open world civ-rebuild game and they wanted us to build as much and as big as possible. And when we said are you sure you can handle that? They said either nothing or "don't worry guys" for 4+ years..


 

The topic is overlapping with other topics, but yes, it is sad that there are and will be more and more restrictions, although there was talk of great freedom at the beginning.

I also miss a clear indication of this on the part of NQ, so that players don't run after a dream that can no longer be fulfilled.

 


 ->> Everything in the game is player-made: ships, cities, orbital stations,... with no limit in size. The world is fully editable.
 ->> Build a giant space station the size of a moon with your friends
-> > Gather along with thousends of others in player-made cities

Link to comment
Share on other sites

17 hours ago, Ving said:

The tool in action looks great, but the new numbering system seems like a solution to a technical problem at the expense of logic and usability.

 

Using steps of 2-4-8-16-32-64 etc (which we can do with the current reactors) makes it very easy to plan shapes and just double the precision each time you need a finer line for instance.

 

Having 3 different scales 2-4 steps, 6-12 steps, and 21-42-84 steps within the same system seems like it will be a pain to use when scaling shapes up / down. At the moment I draw everything on graph paper and just select the precision that will allow me to create the shapes I want. Usually 16 is enough.

 

I can't imagine how I would be able to do this when it would need to show 3 different scales that aren't divisible by each other. 

 

Is there no way you can make it work with a geometric progression? It works with reactors now, so what's the problem continuing this way?

 

 

 

So this is an interesting question which I will attempt to answer to the best of my ability...

 

I'm really fighting the urge to make the "It never was" meme here, but I'm sure one of you will do that for me soon enough.
The reality here is that you never actually were making 1/8th or 1/16th slopes, you've been creating what is the closest approximation of that.

 

  • In the old system, we used 253 points.
  • In the new system, we use 252 points.

This means that in the old system, a single voxel was: 84.3333333333333 (recurring) points.
84.333 also does not divide by 8, 16, 32, or 64.


In fact, in the old system you couldn't really reliably cut a voxel in half to an exact precision, and even a single voxel was not precise. As for example:

  • 84.333 / 2 = 42.166 (in reality this would have been 42 because we don't store decimals)
  • 84.333 / 4 = 21.083 (in reality this would have been 21 because we don't store decimals)
  • 84.333 / 8 = 10.541 (in reality this would have been 11 because we don't store decimals)
  • 84.333 / 16 = 5.270 (in reality this would have been 5 because we don't store decimals)
  • 84.333 / 32 = 2.635 (in reality this would have been 3 because we don't store decimals)
  • 84.333 / 64 = 1.317 (in reality this would have been 1 because we don't store decimals)

Sure, the difference is so negligible that you can't see it by eye. But that's essentially the same in the new system as the new pattern looks like this:

  • 84 / 2 = 42 exactly
  • 84 / 4 = 21 exactly
  • 84 / 8 = 10.5 (in reality this would be either 10 or 11 again because we don't store decimals)
  • 84 / 16 = 5.25 (in reality this would be 5 again because we don't store decimals)
  • 84 / 32 = 2.625 (in reality this would be 3 again because we don't store decimals)
  • 84 / 64 = 1.315 (in reality this would be 1 again because we don't store decimals)

Now, if we had changed the division to 64 instead of 84.333 you could expect the following to happen to all currently existing constructs:

  • A loss of precision around 25%
  • Every existing voxel would have lost around 25% of its available detail.
  • You would see huge changes in your designs and most existing designs would likely loose a lot of their detail.
  • Curves would be less curvy, more blocky.
  • But you would have access to a 1/8 slope.

With the new division of 84 instead of 84.333, you can expect the following:

  • The precision loss is only 0.395%
  • Every voxel will look near enough exactly the same, except for a few edge case ones.
  • You likely not see any noticeable change in your existing designs.
  • Curves are still curvy.
  • But your 1/8 slope might be a bit wonky, and its probably better to adjust to 1/7. 

In short, the precision cost of changing to 1/64 is not worth it. It really isn't. Trust me, we've looked. It's ugly.


I can already see the new question brewing in your minds: Why didn't you increase it to 128 per voxel?
Sure, this could increase the detail and be more divisible, however it also doesn't fit inside a single byte. So now we are talking about every single construct in the game taking up twice as much in terms of data. And if you feel your cache is big now, you really don't want to know what its like with double the resolution of voxels.  


We could perhaps consider introducing a pseudo 1/64 grid mode further down the road, which would give you a 1/64 grid. However it will still not actually place a vertex at a 1/8, 1/16, 1/32, 1/64 position. It would place it at its closest available position.


Also, to answer the question about the scale at which the tool works. No, it will always be 1.5vx in each direction from the vertices point of origin.
I thought maximum adjustment range on this image made that quite clear, but perhaps that was an error on my part:

image2.png


I highly recommend you try it before you cast to much judgement on it, as someone who's tinkered with voxels for a long time. I absolutely love using the tool.

I find myself mostly using Grid 2 and Grid 7, using the control key to make bigger jumps.

 

I hope this answers some of the burning questions you all have.
I wish you all a wonderful day, and look forward to seeing what you will all create with it.
- Deckard

Link to comment
Share on other sites

3 hours ago, CptLoRes said:

Remember when NQ used to say DU was a open world civ-rebuild game and they wanted us to build as much and as big as possible. And when we said are you sure you can handle that? They said either nothing or "don't worry guys" for 4+ years..

 

So while what you say is probably true and likely to happen, it will also means the main premise for DU will be null and void and that NQ never managed to deliver on the tech needed for the game they sold us..

 

 

Like these from the E3 announcement? 

Dual Universe

 

Holiday road.

 

Or the loading screen that reminds you every 5th login what was meant to be possible? 

 

DUAL.png

Edited by Megabosslord
Link to comment
Share on other sites

1 hour ago, NQ-Deckard said:

 

So this is an interesting question which I will attempt to answer to the best of my ability...

 

Thank you, Deckard, for the detailed response.

 

I do look forward to trying it out, and seeing how existing voxel libraries are affected.

 

The pseudo 1/64 grid sounds like a good idea ^^

Link to comment
Share on other sites

You know I love you all(in a very safe, not threatening platonic way.)

and yes, but DAMN were our approximations as close as we could get them. /grin

and i know I'm not supposed to mix my metaverses....

but...

Fear is the mindkiller, it is the little death....

 

 

we did so very much with such clumsy tools for so long, that there isn't a dev team alive that isn't deathly afraid of what we will all do with POWERFULL TOOLS.

but they should stop being afraid. just do it, and let the magic happen!

Link to comment
Share on other sites

7 hours ago, NQ-Deckard said:

Also, to answer the question about the scale at which the tool works. No, it will always be 1.5vx in each direction from the vertices point of origin.

Good. I was really hoping I was misreading the post. But what does this mean? "And, in the example of these four cubes, if we were to go to a value above 84, it would result in an ill-formed shape because the blue shape would have a negative volume. This would probably create visual artifacts, and we may prevent this situation in the future."

 

And do you have any plans to add edge, face, and full voxel moving to the VPT? It would be nice to move multiple vertices at once.

 

Appreciate your detailed response regarding the new encoding. My 1/8th slopes were all a lie...

Edited by TobiwanKenobi
Link to comment
Share on other sites

It was hard to tell from the video,  but is there any kind of numerical position indicator  for the vertex in relation to the true position ? IE (x.5, y.2, z.125). 

Link to comment
Share on other sites

18 hours ago, Warlander said:

Personally im kinda hoping that much that was showcased in this vid was used with the VPT

 

 

 

It would be ironic and unsurprising if showcased ships can no longer be built post complexity limits. LOL!

Link to comment
Share on other sites

37 minutes ago, Rahzig said:

It was hard to tell from the video,  but is there any kind of numerical position indicator  for the vertex in relation to the true position ? IE (x.5, y.2, z.125). 

damn good question. I mean, if there isn't, then they will not achive the objectives. total epic fail to level up.

Link to comment
Share on other sites

"We could perhaps consider introducing a pseudo 1/64 grid mode further down the road, which would give you a 1/64 grid. However it will still not actually place a vertex at a 1/8, 1/16, 1/32, 1/64 position. It would place it at its closest available position."

 

Please this. This would make this tool so much more usable.  In fact I would gladly take a pseudo 1/16 grid over the 1/42. A 1/64 would be heaven. 

Link to comment
Share on other sites

9 hours ago, NQ-Deckard said:

 

So this is an interesting question which I will attempt to answer to the best of my ability...

 

I'm really fighting the urge to make the "It never was" meme here, but I'm sure one of you will do that for me soon enough.
The reality here is that you never actually were making 1/8th or 1/16th slopes, you've been creating what is the closest approximation of that.

 

  • In the old system, we used 253 points.
  • In the new system, we use 252 points.

This means that in the old system, a single voxel was: 84.3333333333333 (recurring) points.
84.333 also does not divide by 8, 16, 32, or 64.


In fact, in the old system you couldn't really reliably cut a voxel in half to an exact precision, and even a single voxel was not precise. As for example:

  • 84.333 / 2 = 42.166 (in reality this would have been 42 because we don't store decimals)
  • 84.333 / 4 = 21.083 (in reality this would have been 21 because we don't store decimals)
  • 84.333 / 8 = 10.541 (in reality this would have been 11 because we don't store decimals)
  • 84.333 / 16 = 5.270 (in reality this would have been 5 because we don't store decimals)
  • 84.333 / 32 = 2.635 (in reality this would have been 3 because we don't store decimals)
  • 84.333 / 64 = 1.317 (in reality this would have been 1 because we don't store decimals)

Sure, the difference is so negligible that you can't see it by eye. But that's essentially the same in the new system as the new pattern looks like this:

  • 84 / 2 = 42 exactly
  • 84 / 4 = 21 exactly
  • 84 / 8 = 10.5 (in reality this would be either 10 or 11 again because we don't store decimals)
  • 84 / 16 = 5.25 (in reality this would be 5 again because we don't store decimals)
  • 84 / 32 = 2.625 (in reality this would be 3 again because we don't store decimals)
  • 84 / 64 = 1.315 (in reality this would be 1 again because we don't store decimals)

Now, if we had changed the division to 64 instead of 84.333 you could expect the following to happen to all currently existing constructs:

  • A loss of precision around 25%
  • Every existing voxel would have lost around 25% of its available detail.
  • You would see huge changes in your designs and most existing designs would likely loose a lot of their detail.
  • Curves would be less curvy, more blocky.
  • But you would have access to a 1/8 slope.

With the new division of 84 instead of 84.333, you can expect the following:

  • The precision loss is only 0.395%
  • Every voxel will look near enough exactly the same, except for a few edge case ones.
  • You likely not see any noticeable change in your existing designs.
  • Curves are still curvy.
  • But your 1/8 slope might be a bit wonky, and its probably better to adjust to 1/7. 

In short, the precision cost of changing to 1/64 is not worth it. It really isn't. Trust me, we've looked. It's ugly.


I can already see the new question brewing in your minds: Why didn't you increase it to 128 per voxel?
Sure, this could increase the detail and be more divisible, however it also doesn't fit inside a single byte. So now we are talking about every single construct in the game taking up twice as much in terms of data. And if you feel your cache is big now, you really don't want to know what its like with double the resolution of voxels.  


We could perhaps consider introducing a pseudo 1/64 grid mode further down the road, which would give you a 1/64 grid. However it will still not actually place a vertex at a 1/8, 1/16, 1/32, 1/64 position. It would place it at its closest available position.


Also, to answer the question about the scale at which the tool works. No, it will always be 1.5vx in each direction from the vertices point of origin.
I thought maximum adjustment range on this image made that quite clear, but perhaps that was an error on my part:

image2.png


I highly recommend you try it before you cast to much judgement on it, as someone who's tinkered with voxels for a long time. I absolutely love using the tool.

I find myself mostly using Grid 2 and Grid 7, using the control key to make bigger jumps.

 

I hope this answers some of the burning questions you all have.
I wish you all a wonderful day, and look forward to seeing what you will all create with it.
- Deckard

thanks for that detailed reply, i was always wondering why my carefully calculated edges weren´t as strait as expected.

 

im absolutely fine with the new divisions, since we only thought we would have 1/8 1/16 /32 1/64 1/128 precision i dont se the difference where this actually matters.

most people wont calculate ther vertex points and just shape by the eye. everyone else can get almost as close to the expected points as possible just like before with the difference of knowing they aren´t... ?

 

only thing im wondering is if a 0.5 result in a grid calculation would end up rounded up or down.

Link to comment
Share on other sites

3 hours ago, TobiwanKenobi said:

No, it will always be 1.5vx in each direction from the vertices point of origin.

recording this for posterity!!!!!!!!!!

Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
 Share

×
×
  • Create New...