It is not, that's what standard LOD techniques are for which switch to lower polygon models depending on distance. Unlike tessellated models they can also use GPU features for rapidly processing duplicate geometry parts which further increases performance.

Add the huge overhead of creating dedicated DX11 models for the developers which would only be useful for a small audience and it doesn't look like a good approach anymore. There's a reason why outside of graphics demos, tesselation is restricted to spicing up generic surfaces so far.

Like all new GPU features before it will take a few years and 1-2 GPU generations before this will see widespread use, especially with consoles out of the equation.

Besides the main issue with hundreds of planes is not the rendering, but AI & FM calculations. Optimizing this area would be more important than exploring new graphics technologies that so far only few people can use.

To me, that quote reads as a message to people on DX9/10 hardware to not worry, but doesn't read as a definite implementation of DX11 specific features. Reading it as devil's advocate again and seeing if a lawyer could wiggle his way out of it. :)


That last quote could indeed be the one. I just fear a lawyer could come in and say "D3D9 and 10 'calls' will look different on DX11 hardware than on DX9/10 hardware. :)

Heliocon!!!

Tesselation is not the magic wonder you seem to think it is. From your posts it's obvious that you don't know how it and other features common in computer gaming work.

Heliocon, LOD models (i.e. switching to lower poly models at a distance) have been in Il-2 since 2001, for crying out loud. Never mind SoW. Tessellation is just like normal/parallax mapping in its effects, if not in its inner workings.

Like Oleg said, tesselation can be useful for human body, uniforms, wheels, etc.
I think it would also be great for the railway ballast to look more 3d instead of flat and for the bomb craters. You can't make everything with tesselation because there is only a small number of tesselation units on the GPU.

I think what Oleg was hinting at in one of his posts was that he was going to wait and see if this intonation of tessellation sticks around for a little longer and becomes the industry standard. There are already a few games that use it, but most of those are FPS with established graphics engines that only require a bit of work to implement. SOW is starting from scratch, so it would be a phenomenal waste of time and resources on something that might be dropped next year or DX release for something better.

If SOW is going to be the sim Oleg hopes with some real longevity I'm sure they will implement it into the next release or maybe even a patch later on. I just hope that any patching to the engine they do for future titles is backwards compatible so it can also be done to BOB and people who want to still fly over the channel can benefit as well as those over Moscow or Korea.

LOD and Tesselation are different effects. LOD requires making more then one model, tesselation does not. Tesselation is mapped onto the model, so you could use 1 model and have its polygon count gradually increase as it gets closer. LOD is not gradual it switches out a low count for a high count model at distance which creates often a "popping" effect. Also it should be noted LOD means either the game has to page the harddrive, or you store the models in your ram which takes up space, which tesselation does not.

Mind to provide evidence/what I said that shows I do not know how effects in computer gaming work? Have you ever done any 3d modeling or texturing?

Just parroting - "you think tesselation is magic" people here on the forums are sad, especially because you havent yet told me why what I am saying is in any way incorrect or even advanced your own posts about it (except for the post commenting on LOD which I addressed).

Actually what you describe is using tesselation for continuous LODs in contrast to distinct LODs, which is the classic technique; but let's skip the semantics...

True but irrelevant as long as distinct LOD models are still needed for users without tesselation-capable cards. Also making a model purely out of primitives and surface maps is a nightmare, and that would be required if tesselation is supposed to provide continuous LODs.

RAM is cheap these days, even on the GPU, and tesselation does require memory for its data as well. The difference is insignificant. Popups can definitely be an issue with distinct LODs, but that's again a quality and not a performance issue.

You previously said the game would be "badly coded" when not using tesselation, and that without it all planes would have to be rendered at full quality all the time. That's why the standard LOD approach was brought up, and as it doesn't come with an additional calculation overhead, can make use of geometry instancing for much improved performance, is perfectly compatible with GPUs from several generations and vendors and doesn't need a completely new modeling approach, it certainly looks like the better solution at the moment. Nobody is against tesselation as such, but right now the technology simply isn't mature and widespread enough.

Yep you are correct, the example I used for the tesselation was used due to people saying tesselation does not give better performance for quality then non tessalated models with the same details. Thats why I phrased it the way I did, of course they use LOD but I was trying to explain why tessalation delivers great quality with less performance hits then if it was all high polygon count models. Also LOD has to store all the textures etc (minor stuff), but for high res gaming how many different models do you think they will use/need? What looks like a smooth hull at 50m looks like a octagon at 10m...

Like said though from what I have "heard (as I have never worked wit tesselation or that type of 3d graphics) it should be very easy to convert the models for tesselation (says Nvidia). But going from that you only need 1 model which auto scales. Reason I am advocating for it is because DX9 and DX10 have hit the end of their lives. Windows will no longer support xp in 2011 (I believe, dont know which month it official stops) and vista soon after. DX10 was horribly but Win 7 / DX11 is where everyone is going to, check out steams statistics on users hardware. Win7 64 bit is already the most common, and with the new second gen dx11 cards these graphical effects are the bread and butter of the coming (well actually this) generation.
I am not sure though about LOD and its interaction with lighting vs tesselation over various distances. One thing we havent touched on is the terrain (not planes or buildings). Maybe it would be better used on the ground due to the nature of the geometry (nothing than having a line in your vision where trees suddenly "appear" etc.