Need advice. Strategic plans about voxels.

[Uberkreatur]

Ok guys, I need some advice. I have a large project that involves active usage of models. Until now, they were a voxel models, workable with zdoom software renderer.
I successfully implemented a huge workable WW2 voxel tanks, created a mechanism for visual synchronizing voxels on Z-coordinate. All this has been working and developing, but Randy suddenly gone, and future of voxel models is doubtful now. Of course, GZdoom supports voxels too, but I heared, that OpenGL renderer creates an extremely high load with voxels. ( It took about 20 tanks to get 12 FPS on the test. )

I am in some confusion now. It is planned to add more than 10 vehicle types. it's a lot of models and sub-models. And I do not know, what model format must I choose. On the one hand, I do not want to lose compatibility with the software renderer. On the other hand, if all the new ports and forks will be unified with the GL, it needs to adapt to them in advance. (And honestly, sometimes on software renderer large voxel model cause lags on maps with a lot of transparent middle textures.) Also, I know nothing about MD2/MD3 format. And I'm not sure that it does not contain any restrictions on the number of polygons or something like that.
Anyway, maybe it would not be so difficult to convert existing voxel models into polygonal, but, considering the situation with zdoom and Randy, I want to avoid the situation in which I will have to convert absolutely everything later.

So, should I continue to use voxels or already now forget about ZDoom and focus on polygonal models for GL-based ports? Needs to choose a paradigm.

[kodi]

I personally would use 3d models because they're easier to work with, less performance intensive, and more flexible in what they can do.

Here's a WIP video of a drivable tank I made with md3 models (base models borrowed from Blade of Agony). If you want to learn about how to use 3d models, check out BoA or Hunter's Moon in slade 3 - they both have extensive 3d model use and nicely organized archive files.

It's up to you of course, whether software renderer support is more important than the other stuff.

successfully implemented a huge workable WW2 voxel tanks, created a mechanism for visual synchronizing voxels on Z-coordinate.

That sounds really interesting. Care to elaborate?

[Uberkreatur]

I personally would use 3d models because they're easier to work with, less performance intensive, and more flexible in what they can do.

Of course, md2/md3 are much more flexible and they dont need many workaround methods I have to use.
I rather want to understand how software renderer is strategically relevant now, after the disappearance of Randy and "death" of Zdoom. And does it make sense to maintain it as "outdated technology". For example in Russia, where I live, people on average are poorer and computers on average are weaker, so GZdoom is considered a "heavy software" which tend not to be used without the need for. But here I have a feeling that most of modern mods are made for GZdoom. If zdoom is dead, now software renderer is dead too?

That sounds really interesting. Care to elaborate?

Here I described a problem.
In later versions script checks player camera position and moves the turret relative to the base 1 pixel closer or further. So, voxel model which at the moment needs to be drawn foreground is always closer to player, and software renderer puts it on top. This method works even with movable objects. (But only in single player mode.)

[kodi]

Of course, md2/md3 are much more flexible and they dont need many workaround methods I have to use.
I rather want to understand how software renderer is strategically relevant now, after the disappearance of Randy and "death" of Zdoom. And does it make sense to maintain it as "outdated technology". For example in Russia, where I live, people on average are poorer and computers on average are weaker, so GZdoom is considered a "heavy software" which tend not to be used without the need for. But here I have a feeling that most of modern mods are made for GZdoom. If zdoom is dead, now software renderer is dead too?

The software renderer is not dead, because A) it's still included in GZDoom and B) QZDoom (truecolor software renderer fork) is in development. The QZDoom software renderer will be available in GZDoom at some point the future AFAIK.

However, I'm not actually sure most old computers will have better performance in software rather than hardware unless they're really ancient, on normal maps at least.

Here I described a problem.
In later versions script checks player camera position and moves the turret relative to the base 1 pixel closer or further. So, voxel model which at the moment needs to be drawn foreground is always closer to player, and software renderer puts it on top. This method works even with movable objects. (But only in single player mode.)

That's a neat workaround