Great Battles Future

[Juri_JS]

18 hours ago, AEthelraedUnraed said:

Oh I agree completely. But I'm optimistic. The performance of neural nets, the hardware and the availability of trained programmers are all rapidly improving. I think it's very likely that in the future, we'll see terrain that is at least partly generated by AI like in MSFS. Probably not in the IL2:Korea engine, but the one after that.

 

One issue that remains however is the starting cost. Even if you're using neural nets, the initial required effort of creating a map is most likely going to be much higher if you first have to program something that does the map creation for you, rather than just create the maps yourself. All the while, you don't really have anything to show for it while your existing team of map creators remains idle. A big company like MS (or Google or Meta or any of the other "usual suspects" when talking about Neural Nets) may well be able to fund such an experimental project (as we've seen with MSFS), but IL2 might not be able to raise the required revenue.

I also think that this will be the future. Actually much of the necessary technology already exists and is used in geoinformatics. For example it's already possible to automatically extract land cover data from old ww2 aerial photos. I am not an expert, just a normal GIS user, but from what I've read it's done by using machine learning technics.

 

I think it would be a huge help for the team if they had someone with a background in geoinformatics. It's really amazing what these guys can do.

[Aapje]

Perhaps, although it seems very difficult to take black & white photographs made with older equipment, during different periods, and then extract a proper map from it.

 

Modern aerial mapping has LiDAR and color, so MSFS can much more easily detect elevation and what objects are. And they can create textures from the images and put those on the ground.

 

 

[AEthelraedUnraed]

24 minutes ago, Aapje said:

Perhaps, although it seems very difficult to take black & white photographs made with older equipment, during different periods, and then extract a proper map from it.

More difficult than with modern satellite imagery, sure, but it should be possible nonetheless. The result will not be a full 100% accurate, but who cares?

 

25 minutes ago, Aapje said:

Modern aerial mapping has LiDAR and color, so MSFS can much more easily detect elevation and what objects are. And they can create textures from the images and put those on the ground.

AFAIK, lidar data isn't available for the vast majority of the planet and I don't think MSFS uses it in any of their algorithms anyhow. They do use stereoscopic imaging however to create a 3d reconstruction of a few selected cities, but this too is not available for the vast majority of the planet. To my knowledge, MSFS uses only 2D satellite images in their reconstruction of "generic" terrain.

 

Fun fact; stereoscopic imaging is available for a large part of WW2 and even WW1 aerial photography.

[Juri_JS]

3 hours ago, Aapje said:

Perhaps, although it seems very difficult to take black & white photographs made with older equipment, during different periods, and then extract a proper map from it.

Just to avoid misunderstandings - the technic does not create textures from old aerial photos, it just gives you information where the right textures need to be placed. It's very detailed data that shows you the locations of forests, water, fields, pasture, roads and positions of buildings. At the moment it's necessary to manually place such textures and buildings when creating a ww2 flight sim map, but you could automate this process when you've extracted the necessary data. Of course you still have to create the textures and objects first.

 

2 hours ago, AEthelraedUnraed said:

AFAIK, lidar data isn't available for the vast majority of the planet and I don't think MSFS uses it in any of their algorithms anyhow. They do use stereoscopic imaging however to create a 3d reconstruction of a few selected cities, but this too is not available for the vast majority of the planet. To my knowledge, MSFS uses only 2D satellite images in their reconstruction of "generic" terrain.

I think for many European countries we already have complete LIDAR data in 1 m resolution, but it's often not available for free. In my country the costs for buying it for larger areas are crazy. The situation in Russia and certain other countries is a different matter. Even if they already have such high resolution data, they will probably not make it available to the public at the moment, because it can be used militarily, for example in the navigation systems of missiles.

As far as I can tell using high resolution elevation data for a ww2 flight sim map wouldn't work very well anyway, because it will show all kinds of modern structures, for example modern road and rail embankments. When these structures have to be deleted manully, it would be an awful lot of work, if no method is found to do it automatically. Moreover the file size of a terrain mesh in 1 m resolution would be massive.

Elevation data with a resolution somewhere between 15 m and 30 m is probably much better suited for today's flight sims and it's available for most parts of the world. Moreover it's either free or available at relatively low costs.

 

Edited May 31, 2024 by Juri_JS

[Aapje]

Yeah, I'd think that you would be limited to pictures from the period. But you might not have any usable WW 2 aerial pictures for large parts of the map, outside of parts that were of particular military interest.

[AEthelraedUnraed]

But we don't need any aerial imagery, really.  As long as you've got a map (which should be available for most of Europe in the 1940s), a (combination of) neural net(s) should be able to construct a believable representation. Even ground textures, although that would require some sort of streaming and the costs of setting that up and maintaining it might be prohibitive for a relatively small company like 1CGS.

[Juri_JS]

29 minutes ago, Aapje said:

Yeah, I'd think that you would be limited to pictures from the period. But you might not have any usable WW 2 aerial pictures for large parts of the map, outside of parts that were of particular military interest.

That depends on the region. For Normandy for example we probably have full coverage by aerial photos. If you don't find any military aerial photos you could also use non-military photos created by land surveying offices in the years before and after the war. If there are still gaps you can also extract some of the data from high resolution topographic and cadastral maps using similar technics, although the results will probably be less accurate.

As AEthelraedUnraed already mentioned, you could even do it with maps alone, but I guess the results would look less natural.

[Aapje]

@Juri_JS

 

Once you start mixing data, you are probably better of first making a basic map by hand, and only then generating the more detailed map. Otherwise the generating tool will have to support a ton of data formats, types of data, qualities of data, etc.

[Jackfraser24]

On 5/31/2024 at 12:59 AM, AEthelraedUnraed said:

Oh I agree completely. But I'm optimistic. The performance of neural nets, the hardware and the availability of trained programmers are all rapidly improving. I think it's very likely that in the future, we'll see terrain that is at least partly generated by AI like in MSFS. Probably not in the IL2:Korea engine, but the one after that.

 

One issue that remains however is the starting cost. Even if you're using neural nets, the initial required effort of creating a map is most likely going to be much higher if you first have to program something that does the map creation for you, rather than just create the maps yourself. All the while, you don't really have anything to show for it while your existing team of map creators remains idle. A big company like MS (or Google or Meta or any of the other "usual suspects" when talking about Neural Nets) may well be able to fund such an experimental project (as we've seen with MSFS), but IL2 might not be able to raise the required revenue.

Do you recon they could build a Korean map covering the entirety of the Korean Peninsula?

[Juri_JS]

12 hours ago, Aapje said:

@Juri_JS

 

Once you start mixing data, you are probably better of first making a basic map by hand, and only then generating the more detailed map. Otherwise the generating tool will have to support a ton of data formats, types of data, qualities of data, etc.

Not really, the data format you get after an extraction process is rather simple and uniform raster data, that can be easily combined and adjusted when it comes from different sources. Depending on the terrain type the raster pixels would have specific values. For example 1=forest, 2=water, 3=fields. These values would be used later to automatically place textures at the correct locations of your elevation model.  As a simple example how such data looks, here's moderne CORINE Land Cover data from the Caen area. The data is extracted from Landsat satelite images.

 

 

 

 

 

[AEthelraedUnraed]

20 hours ago, Jackfraser24 said:

Do you recon they could build a Korean map covering the entirety of the Korean Peninsula?

Well, "could" in the sense that it's surely possible, yes, but I don't know if they will. The engine is in theory capable of rendering such a large map. However there are two issues with that:

- Populating the map with buildings, airfields, landmarks etc. takes time (=money). Given a certain budget, they can either build a small map with lots of detail, or a large map with less detail.

- The larger the map, the more distortion you're gonna get because of projecting spherical coordinates onto a 2D space.

 

How they weigh "detail" vs "size" and how large distortions they're willing to accept is essentially a marketing question.

[R7-S276]

It also be great to have a map which cover all the France, I don’t know why France is always cut in parts. The aerial fights between France and Italy in 1940, the south of France with the aerial operations to prepare and cover the allied operation in august 1944, the Corsica island which served to a lot of operation over France and Italy. I wish one day we could have map to cover long range flight and navigation…

[LukeFF]

3 hours ago, R7-S276 said:

It also be great to have a map which cover all the France, I don’t know why France is always cut in parts. The aerial fights between France and Italy in 1940, the south of France with the aerial operations to prepare and cover the allied operation in august 1944, the Corsica island which served to a lot of operation over France and Italy. I wish one day we could have map to cover long range flight and navigation…

 

Paris is the main problem - with today's current rendering technology even the best computer would struggle to depict it all correctly. When I was discussing the proposed map borders for the Normandy map with Jason it was a constant struggle to find a proper set of borders that would properly show enough of the map to represent the Normandy campaign but at the same time could cull Paris and not look weird. 

[AEthelraedUnraed]

7 hours ago, LukeFF said:

Paris is the main problem - with today's current rendering technology even the best computer would struggle to depict it all correctly.

Perhaps IL2's current rendering tech, but MSFS has no problems with large cities at all. For many cities it even has 3d scanned buildings, which is a graphics programmer's worst nightmare since you cannot reuse any assets. And even with IL2's engine, things can be done to mitigate the problem, e.g. better LOD'ing. It can handle a distant Moscow after all.  Since IL2 has a proprietary engine, improving the rendering tech should be possible.

 

I remain convinced the issue with large cities is more the required investment (both time and money) rather than technological barriers. Investment being needed for both the improvement of rendering tech as well as the actual building of the cities.

 

10 hours ago, R7-S276 said:

It also be great to have a map which cover all the France, I don’t know why France is always cut in parts. The aerial fights between France and Italy in 1940, the south of France with the aerial operations to prepare and cover the allied operation in august 1944, the Corsica island which served to a lot of operation over France and Italy. I wish one day we could have map to cover long range flight and navigation…

Even if the engine could be improved, there's still the very major issue of implementing it all. The Normandy map took them 2 years to build and it covers, what's it, a 10th of France or so? Extrapolate this and you're looking at 20 years to build a map of the whole of France.

 

As for what this map would add; we don't have any 1940 airplanes, no French airplanes at all and only one Italian. Operation Dragoon had barely any Luftwaffe involvement. Compared to how active the Normandy map area was, I think it's perfectly valid to cut up France.

[BMA_FlyingShark]

15 minutes ago, AEthelraedUnraed said:

Perhaps IL2's current rendering tech, but MSFS has no problems with large cities at all.

The difference is that in IL2, buildings are all separate objects, in MSFS, they're just an elevation of the ground and have no damage model or whatsoever.

That's also the reason why Italy would be so hard to model in this series, they said in one of the Q&A vids (wasn't in the latest one but I think it was the first one).

 

Have a nice day.

 

[Aapje]

@AEthelraedUnraed

 

Buildings (in games) are typically very simplistic in shape, which probably require very few polygons. I wonder what is actually the performance issue? Shadows, lighting, textures? Perhaps @LukeFF knows and can tell us what the issue is with buildings costing a lot of performance?

 

@FlyingShark

 

That's just not true, MSFS generates real objects from the satellite images. And you also have the hand-made buildings.

 

 

 

[AEthelraedUnraed]

1 hour ago, FlyingShark said:

The difference is that in IL2, buildings are all separate objects, in MSFS, they're just an elevation of the ground and have no damage model or whatsoever.

That's also the reason why Italy would be so hard to model in this series, they said in one of the Q&A vids (wasn't in the latest one but I think it was the first one).

I was about to reply to this, but Aapje was just ahead of me All I can add is that they indeed do not have a damage model. After all, why would MSFS need one? You can't bomb buildings anyway. However as long as you stick to "generic" buildings rather than photo-realistically reconstruct entire cities, it shouldn't be too hard to create damage models.

 

1 hour ago, Aapje said:

Buildings (in games) are typically very simplistic in shape, which probably require very few polygons. I wonder what is actually the performance issue? Shadows, lighting, textures? Perhaps @LukeFF knows and can tell us what the issue is with buildings costing a lot of performance?

Certainly not the shaders themselves ( =lighting), since these are only run for visible pixels. I.e. if a building is only visible in one pixel, its shader is run only once. This therefore doesn't really scale with the amount of objects.

 

Texture calls are done in the shaders so as long as various assets share textures and these fit into your VRAM, this doesn't scale with the amount of buildings either.

 

Dynamic shadows can be quite expensive since essentially the shadowmapper algorithm needs to re-render the scene (albeit without shaders). Vertex and Tri count too can be an issue even for very simplistic buildings, if you have to render hundreds of thousands of them.

 

What's usually an issue with rendering lots of objects and what I *think* is also the main issue here, is communicating everything to the GPU (a so-called "draw call"). The GPU then has to load the shader, load the 3d model, apply transformations... Each of which takes some time (especially the former two). However there are some nice shortcuts you can take. For instance, if you know beforehand that the shader and model isn't going to change, e.g. if you put 1000 copies of the same building somewhere, you can essentially skip the most expensive 2/3rds of this process for 999 of the 1000 buildings. Similarly, if you know a bunch of buildings isn't going to move, you can bake it into a single mesh per material so that you even don't have to apply transformations. Therefore, if you have 1000 buildings but each of these buildings uses the same material and the buildings don't move**, you can essentially render them using just a single draw call.

 

* Same material means the same shader and the same textures, although these textures can be large. You might dedicate the upper left corner to building 1, the upper right to building 2, etc.

** Buildings don't usually move... but if they need to have a damage model, their mesh likely changes so that you cannot pre-bake these combined meshes.

Edited June 3, 2024 by AEthelraedUnraed

[BMA_FlyingShark]

1 hour ago, Aapje said:

 

That's just not true, MSFS generates real objects from the satellite images. And you also have the hand-made buildings.

 

20 minutes ago, AEthelraedUnraed said:

However as long as you stick to "generic" buildings rather than photo-realistically reconstruct entire cities, it shouldn't be too hard to create damage models.

Ok then, I now think it shouldn't be a problem te get big cities in Il2.

 

Have a nice day.

 

[Aapje]

30 minutes ago, AEthelraedUnraed said:

However as long as you stick to "generic" buildings rather than photo-realistically reconstruct entire cities, it shouldn't be too hard to create damage models.

 

I don't really get why damage models would be a big issue, since you can replace the low poly model with a higher poly destroyed building when the bombs land. This would work until you go crazy and bomb a lot of buildings close to each other. Then system load would go up.

[LukeFF]

4 hours ago, AEthelraedUnraed said:

Perhaps IL2's current rendering tech, but MSFS has no problems with large cities at all. For many cities it even has 3d scanned buildings, which is a graphics programmer's worst nightmare since you cannot reuse any assets. And even with IL2's engine, things can be done to mitigate the problem, e.g. better LOD'ing. It can handle a distant Moscow after all.  Since IL2 has a proprietary engine, improving the rendering tech should be possible.

 

Even so, trying to run, say, a flight in the New York City area is very taxing even on the best machines. 

[AEthelraedUnraed]

44 minutes ago, Aapje said:

I don't really get why damage models would be a big issue, since you can replace the low poly model with a higher poly destroyed building when the bombs land. This would work until you go crazy and bomb a lot of buildings close to each other. Then system load would go up.

The issue with damage models and photogrammetrically created cities is that you'd either have to create an individual damage model for every single building beforehand (a waste of time and disk space), or generate one "on the fly". Even if you succeed in making the generation algorithm fast enough so that the system doesn't blink if, let's say, 100 buildings are destroyed within a couple of seconds, then each of these generated damaged buildings will still be its own mesh, precluding many of the draw call optimisations I talked about in my previous post.

 

For instance, it breaks the static batching the photogrammetric buildings likely have; i.e. many of these buildings probably have their meshes baked into a single combined mesh in order to share draw calls. In other words, if you have a city block with 100 buildings, instead of having 100 meshes (one for each building) it has one mesh containing 100 buildings. If one of these many buildings gets damaged, you'd not only have to generate a damage model but also have to re-bake the 99 remaining meshes into a new super-mesh.

 

For "generic" buildings like IL2 currently uses this is less of an issue since whatever buildings are undamaged still share the same mesh (so can be combined into a single draw call). If you had 100 of the type before, you now still have 99 left. You will need an additional draw call for the damaged one, but you still save 98 draw calls. What's more; if additional buildings get damaged, they can be combined with the "new" draw call since these share the same damaged mesh.

 

Note that none of this is based on any inside knowledge of either the MSFS or IL2 engines; it's just based on what I think are likely performance tweaks for both of them.

[LukeFF]

3 hours ago, Aapje said:

Buildings (in games) are typically very simplistic in shape, which probably require very few polygons. I wonder what is actually the performance issue? Shadows, lighting, textures? Perhaps @LukeFF knows and can tell us what the issue is with buildings costing a lot of performance?

 

I'm not really in the know about that, but I do know that one of the things with the original IL2 is that every building object has/had to "report" every second or so whether or not it was damaged. That takes a performance toll on calculations. 

[AEthelraedUnraed]

4 minutes ago, LukeFF said:

Even so, trying to run, say, a flight in the New York City area is very taxing even on the best machines. 

Mostly for photogrammetrically reconstructed cities. I flew above Mumbai (India's most populous city) the other day on my mid-range laptop, and didn't notice any severe performance issues.

 

3 minutes ago, LukeFF said:

I'm not really in the know about that, but I do know that one of the things with the original IL2 is that every building object has/had to "report" every second or so whether or not it was damaged. That takes a performance toll on calculations. 

I would be very surprised if in the current IL2 they haven't switched to a more event-based approach where these "reports" are only done when a building gets damaged.

[Aapje]

54 minutes ago, AEthelraedUnraed said:

The issue with damage models and photogrammetrically created cities is that you'd either have to create an individual damage model for every single building beforehand

 

Sure, but in the case of IL-2 we aren't expecting photogrammetric buildings, right? They can have a bunch of standard buildings and the map generator would place them in a way that makes sense for generic cities, although it won't actually be historically correct. But I don't think we are at a point where we can expect historically correct cities from smaller companies.

 

Of course, it would be very nice to have some manually made landmarks as well, especially for Paris (Eifel Tower, Arc de Triomphe).

57 minutes ago, LukeFF said:

 

I'm not really in the know about that, but I do know that one of the things with the original IL2 is that every building object has/had to "report" every second or so whether or not it was damaged. That takes a performance toll on calculations. 

 

Is this for MP, where the damage has to get synced to other systems?

 

Even then, it seems like a major coding error, since only the damaged buildings would need to report in.

[AEthelraedUnraed]

5 minutes ago, Aapje said:

Sure, but in the case of IL-2 we aren't expecting photogrammetric buildings, right?

Exactly - which is why I said that as long as you don't use photogrammetry, damage models shouldn't be a huge issue

[Jackfraser24]

Will they do another map for Great Battles once Odessa is done? If they will, I hope they do Kursk. 
 

Also, will the crew that are doing the Karelia map do another map once the one they are working on is done? 

[Jackfraser24]

I hope someone does a Battle of Crete one day. I know the Germans won this one, but the battle lasted for quite a while. If it is done, then this would a) create another purpose for the Ju-52, and other existing fighters and bombers in the game, b) satisfy many of us with an early war Mediterranean map and c) allow for some early war aircraft to be made for IL-2.

• Bf-109 E-3/4
• Bf-110 C-4/7
• Do-17 Z-2
• Ju-87 R-2
• Ju-88 A-1
• Gladiator Mk.1 and 2
• Hurricane Mk.1
• Blenheim Mk.1 and Mk.4
• Wellington Mk.1C

[LukeFF]

On 6/3/2024 at 9:52 AM, Aapje said:

Is this for MP, where the damage has to get synced to other systems?

 

Even then, it seems like a major coding error, since only the damaged buildings would need to report in.

 

Given this was for IL2 circa 2001, I dunno. 

[Jackfraser24]

Why Kiev should be considered after Odessa

• There were two very large battles that happened in Kiev during WWII. The first battle was in 1941 and went from July 7 until September 26 1941. The second battle was in 1943 and went from November 3 until December 22. Career mode would be decent.
• It was a significant battle in 1941 because A) it was a major Axis strategic victory, though it took a while to achieve; B) the battle was one of the biggest military encirclement's of all time (600,000 troops) which lead to a major Soviet defeat; and C it lead to a major boost in German morale
• It was a significant battle in 1943 because A) it was a decisive and strategic Soviet victory and morale boost; B) over a million soviet troops retook the city from 600,000 Axis defenders so it was a huge battle and C) it weakened the German forces on the Eastern Front.
• 1941 Kiev would serve as a good prelude to the Battle of Moscow and demonstrate how effective blitzkrieg was before the Soviets found a way to counter the German blitzkrieg.
• 1943 Kiev would serve to bridge chronological gaps between Kuban and possible 1944 Eastern Front late war scenarios such as Minsk, Lublin/Brest, Lvov, Balaton, Slovakia and Berlin (once computers can render large metropolitan areas smoothly) 
• A Battle of Kiev would give existing planes another map to fight on. Planes like the Bf-109 E-7, F-2 and G-6; the La-5, 5F and 5FN; the He-111 H-6 and H-16; and the Yak-9 and 9T.
• Other aircraft that are not already in the game a reason to be in the game. Planes like the Bf-109 E-4, Bf-110 C-4, Do-17 Z-2, He-111 H-3, Ju-87 B-2, IL-2 Model 1940, Ju-88 A-14, IL-4, LaGG-3 series 4, etc...
• The battle did not consist of just Kiev but also the larger area surrounding it. Would make a good multiplayer map because you wouldn't have to worry about having to fly long distances over the ocean like you do for Normandy which can be rather boring and painstakingly long to just reach enemy lines.

 

Edited June 7, 2024 by Jackfraser24

[AEthelraedUnraed]

18 hours ago, Jackfraser24 said:

Why Kiev should be considered after Odessa

Truth be told, I find it very unlikely the Devs are considering any maps at all after Odessa/Karelia (no way of knowing which will come sooner).

 

The Devs are working on the next title now. Creating a different map is a very resource intensive process so they're surely not going to spend all there resources on a game that has essentially run its course. We may see one or two maps by volunteers (similar to how the Karelia map is being created), but there will very likely not be another official module.

[Jackfraser24]

16 hours ago, AEthelraedUnraed said:

Truth be told, I find it very unlikely the Devs are considering any maps at all after Odessa/Karelia (no way of knowing which will come sooner).

 

The Devs are working on the next title now. Creating a different map is a very resource intensive process so they're surely not going to spend all there resources on a game that has essentially run its course. We may see one or two maps by volunteers (similar to how the Karelia map is being created), but there will very likely not be another official module.

Questions.

1. How long would it take for a hypothetical Kiev map to make at the rate they would normally work at?
2. Do you think a Kiev map would be profitable? 
3. Has there been talk of doing more Eastern front maps like Kiev, Smolensk, Kharkov and Kursk?
4. Could computer technology advance enough one day for pc's to render maps with cities like Berlin, London and Paris?
5. If you could ask the developers to do a map of your choice for Great Battles what would it be?  

Sorry if I have bothered you or anyone else.

Edited June 9, 2024 by Jackfraser24

[AEthelraedUnraed]

5 hours ago, Jackfraser24 said:

Questions.

1. How long would it take for a hypothetical Kiev map to make at the rate they would normally work at?
2. Do you think a Kiev map would be profitable? 
3. Has there been talk of doing more Eastern front maps like Kiev, Smolensk, Kharkov and Kursk?
4. Could computer technology advance enough one day for pc's to render maps with cities like Berlin, London and Paris?
5. If you could ask the developers to do a map of your choice for Great Battles what would it be?  

Sorry if I have bothered you or anyone else.

1. The average large and well-made map takes about 2 years to make, give or take.

2. Given how many work it is to make a map, I think it's hard to make a profit on any map if it's not also combined with additional planes. There's a reason there's many collector planes and no collector maps.

3. As I said, I think it's unlikely the Devs will release any more maps beyond Odessa. I haven't heard of any community projects besides Karelia.

4. What do you mean, "one day"? The tech is already there to render big cities like the ones you mention.

5. My personal map of choice would be Sicily/southern Italy. The planned next module when Jason was still in charge.

[Jackfraser24]

56 minutes ago, AEthelraedUnraed said:

1. The average large and well-made map takes about 2 years to make, give or take.

2. Given how many work it is to make a map, I think it's hard to make a profit on any map if it's not also combined with additional planes. There's a reason there's many collector planes and no collector maps.

3. As I said, I think it's unlikely the Devs will release any more maps beyond Odessa. I haven't heard of any community projects besides Karelia.

4. What do you mean, "one day"? The tech is already there to render big cities like the ones you mention.

5. My personal map of choice would be Sicily/southern Italy. The planned next module when Jason was still in charge.

Thank you.

[Jackfraser24]

18 hours ago, AEthelraedUnraed said:

What do you mean, "one day"? The tech is already there to render big cities like the ones you mention.

I meant like for IL-2 where they can render manually built maps, as opposed to the type of world map that MSFS uses. From my understanding, right now a computer would not be able to render a major city like Paris or London without severe lagging because computer technology has not matured enough. Maybe in the 2030s though. Do I make much sense?

Edited June 10, 2024 by Jackfraser24

[Aapje]

Current computers can render big cities. The issue is more to manage the quality/LOD to an acceptable level, so it doesn't look ugly. Even MSFS struggles with that as flying in cities tends to depress the frame rates, although there is a tool to dynamically alter the LOD (AutoFPS).

 

I personally think that Nanite would be great for flight sims, as it can dynamically alter the LOD, which fits flight sims very well, since there are so many different distances you can be to objects. But that is not something that is feasible for 1CGS to implement. Perhaps Combat Pilot will use it.

 

But without it, you can still make big cities, but you may have to accept rather bad popping.

[AEthelraedUnraed]

4 hours ago, Jackfraser24 said:

I meant like for IL-2 where they can render manually built maps, as opposed to the type of world map that MSFS uses. From my understanding, right now a computer would not be able to render a major city like Paris or London without severe lagging because computer technology has not matured enough. Maybe in the 2030s though. Do I make much sense?

Ah, right. In that case you're using the wrong terminology; computer technology has matured enough only the IL2 engine hasn't

 

But sure, it should be possible to improve IL2's engine so that large cities should render smoothly on mid-range hardware. It's mostly optimising stuff and employing a few tricks. I mentioned a couple of techniques a few posts ago. In theory, IL2 should be able to render smoother than MSFS due to IL2 using a smaller set of buildings.

 

But that still wouldn't solve the issue of actually building the place. With the current map creation technology, you're looking at possibly many months to build a large metropolis like London or Paris, and that's without any custom landmarks.

 

Hence, what's really needed is a way to automate (part of) the map creation process. However such an algorithm is quite hard to make and will take a long time. I'm not sure if 1CGS has the funds for that. The alternative is completely overhauling the engine so that automating the process would be easier, but re-writing much of the engine is an even greater task.

 

5 minutes ago, Aapje said:

I personally think that Nanite would be great for flight sims, as it can dynamically alter the LOD, which fits flight sims very well, since there are so many different distances you can be to objects. But that is not something that is feasible for 1CGS to implement. Perhaps Combat Pilot will use it.

Nanites would be great for things that you need to see both from up close in extreme detail and from far away, like an aircraft carrier. But I'd be curious to see any benchmarks where you're only looking at relatively low-poly objects to begin with (like the current IL2 buildings). My gut instinct is that traditional methods, as long as they're properly optimised, would still outperform Nanites.

[Aapje]

You can do everything that Nanite does, but it takes way more developer effort. But it doesn't have to be an issue if the game has a limited number of different buildings, that each can be optimized with different LOD versions.

[Jackfraser24]

8 hours ago, AEthelraedUnraed said:

But sure, it should be possible to improve IL2's engine so that large cities should render smoothly on mid-range hardware. It's mostly optimising stuff and employing a few tricks. I mentioned a couple of techniques a few posts ago. In theory, IL2 should be able to render smoother than MSFS due to IL2 using a smaller set of buildings.

But they won’t, will they? Is that because the engine is outdated and/or is it because it has limitations? Is that why they are  in the process of making a new game? 

[Aapje]

They are redoing the game engine, but we have no idea to what extent they are making changes and have the manpower/skill to pull off certain feats. Hopefully we get an announcement soon.

[AEthelraedUnraed]

13 hours ago, Jackfraser24 said:

But they won’t, will they? Is that because the engine is outdated and/or is it because it has limitations? Is that why they are  in the process of making a new game? 

They won't what? Build big cities? No, I don't think so, although I'm not a fortune teller.

 

As I said, I think the greater issue is with building such a map rather than rendering it. To effectively build, let's say, Berlin, they'd have to essentially create all new tools and completely change their workflow (or spend much too much time with the current tools). That takes a lot of effort. Modifiying the engine so that it can smoothly render such towns takes a lot of work as well, but my guess is that it would be much less.

 

Whether or not an engine is "outdated" is subjective. It's something usually shouted by people who've never seen the inside of a computer, whenever a particular game isn't able to do whatever they want it to do.

 

But anyway, IL2 lacks a couple of features that have become commonplace over the past years, such as PBR (another one of those over-hyped buzzwords) shaders. Some of these features, including said PRB shaders, would most likely be possible to introduce into the current IL2 without too much hassle. Some other things they're working on, such as a simpler AI for (some) AI planes or a more advanced radio system could be harder to introduce, which is probably why they're making it a new series since then they don't have to worry about backward compatibility with the current assets. But even though the current AI and radio have their issues, I'd attribute none of those to "outdated" technology but rather to poor design choices in the past.

 

1 hour ago, Aapje said:

They are redoing the game engine

To add some nuance; they are not redoing the engine but modifying it.