Explanation of severe performance issue. Attention Devs

[Underhook]

I just came back to the game after a 9 month break.  I have noticed that in some scenarios the game runs much worse than before.

Then I stumbled on a verifiable issue.

I have attached a picture of my space station.  The view is looking down onto a lower deck.  I built this in January and thought I was very clever at saving materials by using a lattice design (of every 9 blocks, 5 are hollow).  When I got into the pilot seat of the ship in the pic my frame rate dropped to about 4 FPS.  I dared not fly it.  I noticed that when walking over my lattice my jetpack kept on turning of and on very rapidly.  This never used to happen (back in January).  It seems that the game is now much more accurate in picking up these things, like its a higher resolution.  I wondered if this was causing the issue with this ship.  So, I filled in the portion bellow the ship (made it solid).  I jumped in the pilot seat and ..... no major frame rate drop.

 

I'm not a clever person so I thought I would post this in the hope the devs might find this helpful.  Also, players might find in helpful for base design etc.

[joaocordeiro]

The devs gave some explanation a few weeks ago. 

They said they were replacing the current mining system with automining because the current mining system forces players to dig around. 

 

Digging around changes the procedurally generated world and this adds data to the database that has to be streamed to players when they enter rendering distance. 

 

They will preform a full terrain reset and remove current mining mechanism soon. (check announcements) 

 

In theory this will greatly reduce costs but also greatly reduce the data the client downloads and computes upon. 

Dramatically increasing performance. 

 

So, try again after the update happens. 

[Underhook]

Thanks joaocorderio,

I am aware of the desire to fix the hollow planets and that it should improve performance around those planets.  I'm pretty sure the issue I'm showing here isnt related.  So, I'm guessing the issue I'm talking about here will persist.  The fact that it is an issue but it didnt used to be ...... might help the devs identify performance issues with the game (at least thats what I thought).

[CptLoRes]

There are two problems here.

 

1. That lattice design is close to a worst case scenario for voxels and the mesh server conversion of voxels into static 3D meshes.

2. The 3D performance in this game is still terrible, like WTF levels of bad. In the screenshot you are alone on a space station, and even with the lattice I see nothing in that picture that should results in a 4fps performance on any modern computer. Not even close..

[Bobbie]

.

[joaocordeiro]

10 hours ago, Underhook said:

I am aware of the desire to fix the hollow planets and that it should improve performance around those planets.  I'm pretty sure the issue I'm showing here isnt related.

Its not that unrelated. 

Im sure that you found a bug that needs fixing. 

 

But its the same database, if the database is already over the limit it can have performance issues with unrelated zones in map. 

It is also unclear of what will happen to the new spherical reference system and how that will affect space near a planet. 

 

But also, NQ usually accumulates fixes. Its possible that the bug you are reporting is already fixed in this new version. 

 

Also, at this time you have a very low chance of being noticed by devs. They all have the new patch occupying 100% of their attention and even the ones that notice you will have the points above i mention as a justification to not pay atention to your report. 

 

The best thing you can do is to test this exact same scenario after the new patch and fill up a proper bug report in the support site. 

[Zarcata]

I doubt that this relatively small grid alone can provide for such intrusions. In terms of pattern, it is certainly enormous, but there are entire space stations with elaborate voxel patterns that are problem-free. Similarly, the planet's problems with Earth tunnels in space should do nothing at all. It might be interesting to know what hardware you have available.

[CptLoRes]

Once the voxels/meshes has been streamed to the local cache on the player computers, there is no longer any potential server bottleneck that could limit client fps.

So it is almost certain that low fps on a space station that is far from any other players, is an performance issue with the DU client software.

[Bobbie]

.

[Underhook]

6 hours ago, Zarcata said:

I doubt that this relatively small grid alone can provide for such intrusions. In terms of pattern, it is certainly enormous, but there are entire space stations with elaborate voxel patterns that are problem-free. Similarly, the planet's problems with Earth tunnels in space should do nothing at all. It might be interesting to know what hardware you have available.

i9 850, 16gb ram, 1660 super, pcie ssd.

Yes, I thought it would be doubted and therefore overlooked.  Not sure what you mean by pattern.  I have not seen any other space stations with thousands of holes in them.  I think it has something do do with the "ship" trying to deal with multiple simultaneous inputs saying opposite things.  For example ... your floating in space, your on a hard surface.

 

[Underhook]

6 hours ago, joaocordeiro said:

Its not that unrelated. 

Im sure that you found a bug that needs fixing. 

 

But its the same database, if the database is already over the limit it can have performance issues with unrelated zones in map. 

It is also unclear of what will happen to the new spherical reference system and how that will affect space near a planet. 

 

But also, NQ usually accumulates fixes. Its possible that the bug you are reporting is already fixed in this new version. 

 

Also, at this time you have a very low chance of being noticed by devs. They all have the new patch occupying 100% of their attention and even the ones that notice you will have the points above i mention as a justification to not pay atention to your report. 

 

The best thing you can do is to test this exact same scenario after the new patch and fill up a proper bug report in the support site. 

Yes, I will test it again after the patch.

[Underhook]

3 hours ago, Bobbie said:

The dead giveaway is that there is only a frame rate drop when piloting the ship, i.e. when activating a dynamic rigid body from sleep state. And presumably only when flying close to the lattice.

 

So yeah, not only is it probably not a server issue but also it's not a rendering issue. If it were a rendering issue then it would only happen when you look at it (view frustum culling), and it wouldn't matter if you're piloting a ship or not.

 

Hence physics. It wouldn't surprise me if the voxel lattice would be a bad case scenario for an automatically generated collision shape. If the collision shape is like the visual shape then collision detection with so many surfaces close together would certainly be a huge performance hog. The only mystery in this case is why that's an issue now and not 9 months ago.

I'm just guessing, but maybe it was when they did the graphics update.  Or maybe it was that one for docking.  It "feels" to me like an increase in resolution has occurred....... because ..... When I run over the grid I get the message "jet pack on" then "jet pack off" constantly (every block).  9 months ago it happened maybe every 6 blocks.

[Novean-61657]

I don't know enough about voxels, but... I can easily describe a flat surface (x,y,z), but a flat surface with a ton of holes in it is a lot more difficult to describe. Either you describe each hole separately, which costs a lot of storage, or you make a formula for it that needs resources each time it's displayed or interacted with. Maybe they moved from one method to another in the last year?