Jump to content

The Outerra engine


Snave

Recommended Posts

OK, so let's try a GROWN-UP discussion about the possibilities of some new engine that mysteriously suddenly seems to be being plugged mercilessly across simulation forums by previously-unheard of posters <_< - or those with the aptitude of a seven year old. :beanie boy_s:

First, go HERE and learn first-hand about the Outerra terrain engine.

What relevance does a ground based simulation engine have to an airborne simulation product? What the developer claims as a `planetary` engine is nothing of the sort unless it ALSO replicates air masses, weather dynamics and of course, water. The blogger mentions climate data. Where is it, or where does it come from?

While this might have potential does no-one consider the fps impact of fractal alogorithms running entirely on the GPU, when that GPU is also tasked with aircraft and weather display..? The load transients could cause massive fps fluctuations. Centimetric resolution implies both the very highest quality real-world data (which does not exist) and load-focus on the capabilities of the GPU. ACES could easily have re-targetted focus on the GPU but resisted. Wonder why?

Next, visit the forum in the link above to read what Aleksandr says about tile propogation and LOD problems. The Outerra developer is using 76m resolution as a baseline (same as FSX on much of the populated world) but there is no direct comparison with file sizes nor rendering capacity with existing global terrain engines such as the FSX-derived ESP. My interpretation of the data intperolation suggests a logarithmic increase in rendering requirement as the LOD distance shortens and I don't see how this marries with the need for smooth fps and minimal fluctuation during landing and take-off.

Adam Szofran of ACES wrote in 2006 about the limitations of rendering exclusively on the CPU, and his comments apply equally to passing the load exclusively on to the GPU:

Because Flight Simulator must do much more than just render the Earth, the terrain engine can't monopolize the CPU resources of the host system. Adequate resources must be reserved for the other aspects of the game such as aircraft and vehicle simulation, artificial intelligence, multiplayer communication, and rendering of the aircraft, buildings, and other non-terrain entities in the scene.

Lest anyone think I am quoting out of context, the full article is HERE

While this might be an alternative it would need to convince by showing exponential – but not necessarily logarithmic – increases in performance with increasingly powerful modern GPUs to be even considered for a future FS product engine. Outerra has a superficial appeal, but there is as yet nothing of substance to go all gaga over. FS has yet to see full integration of Shader model engineering, the current and near-term future industry standard, much less fractal algorithms based on non-existent data streams.

Discuss.

  • Upvote 4
Link to comment
Share on other sites

I agree with you that this engine in this state is not very suitable yet for a flight simulator.

But everything that it does not have can be programmed and developed. I can only speak for myself, but I think that what most people find so interesting about this engine is that it has a very solid base. It renders the planet "from space down to the surface", it displays "detail ranging from thousands of kilometers down to centimeters", and apparently it is suitable for simulation of vehicles. Look at the truck and the Cessna in the screenshots. And last but not least, it looks very good! This is a project worth following in my opinion.

Link to comment
Share on other sites

OK, so let's try a GROWN-UP discussion about the possibilities of some new engine that mysteriously suddenly seems to be being plugged mercilessly across simulation forums by previously-unheard of posters <_< - or those with the aptitude of a seven year old. :beanie boy_s:

I wish you could post in a manner different to that because I'm sure you have a lot to offer apart from gratuitous insults ... what do you expect with an opening sentence like that? You really have poor social skills, dont you.

I stopped reading right there, if you want a decent and proper discussion, then you need to look in the mirror ... in the words of Pink Floyd ... Good Manners Cost Nothing.

You honestly need therapy, go seek it.

  • Upvote 4
Link to comment
Share on other sites

Ok, let's have a grown up discussion. As a grown-up, I disregard your first paragraph (which isn't grown up at all) and focus on the matter.

As I understand it, the graphics cards of today are very bored with what FSX throws at them. For example, having AA at 2x or combined 8x on my 285 GTX doesn't make a lot of difference in framerates at the resolution I use (3360x1050). So it seems a good idea to have it take care of more things in a flight sim than nowadays.

I know too little about the Outerra engine to judge if it could be a good base for a flight simulator or how much it could be adapted to be one. Knowing from other areas of gaming, I know that some engines are very flexible and the guys that licensed them did amazing things with it.

But the point to be made is another one: Personally, I think that anything looking less impressive visually in 2012 than the Outerra engine looks now would be an epic fail for a flight sim. When I saw the Cessna video for the first time, I was blown away by the seamless transition between wide shot and close up, and by the way the terrain was rendered in the distance. That's something I expect from a next gen flight sim.

  • Upvote 2
Link to comment
Share on other sites

I wish you could post in a manner different to that because I'm sure you have a lot to offer apart from gratuitous insults ... what do you expect with an opening sentence like that? You really have poor social skills, dont you.

I stopped reading right there, if you want a decent and proper discussion, then you need to look in the mirror ... in the words of Pink Floyd ... Good Manners Cost Nothing.

You honestly need therapy, go seek it.

Thank you for your (non-)contribution to the debate. If the best you have to offer to a technical discussion about a flight sim engine is a quote from an obsolete popular beat combo rather than relevant quotes from an essayer, philosopher or incisive commentator then this thread is probably not for you.

  • Upvote 2
Link to comment
Share on other sites

Ok, let's have a grown up discussion. As a grown-up, I disregard your first paragraph (which isn't grown up at all) and focus on the matter.

As I understand it, the graphics cards of today are very bored with what FSX throws at them. For example, having AA at 2x or combined 8x on my 285 GTX doesn't make a lot of difference in framerates at the resolution I use (3360x1050). So it seems a good idea to have it take care of more things in a flight sim than nowadays.

I know too little about the Outerra engine to judge if it could be a good base for a flight simulator or how much it could be adapted to be one. Knowing from other areas of gaming, I know that some engines are very flexible and the guys that licensed them did amazing things with it.

But the point to be made is another one: Personally, I think that anything looking less impressive visually in 2012 than the Outerra engine looks now would be an epic fail for a flight sim. When I saw the Cessna video for the first time, I was blown away by the seamless transition between wide shot and close up, and by the way the terrain was rendered in the distance. That's something I expect from a next gen flight sim.

I concur. Yet nothing I have seen on the Outerra engine thus far shows that it is even capable of matching the rendering qualities of a FSX at high settings. Also modern rendering is fundamentally tied on current and next-generation cards to Shader Modelling, rather than any alternative suggestions. The norm in game design is to make the game fit the hardware, not the other way around. I don't see any specifics of relative performance on a mid-range or better current generation card. And that kind of worries me as developing a game engine in isolation is simply NOT going to `cut the mustard` in modern gaming design.

This, after all, was the basic failing of FSX - it reached too far into hardware of the future, got the guesses wrong and in so doing created the dinosaur that needs replacement today.

  • Upvote 4
Link to comment
Share on other sites

I think many people are responding to the potentials of the technology rather than specifically to the Outerra engine itself. Recently, more and more programs have been demonstrating the power of procedural generation to automate tasks that seem to require painstaking manual effort with current technology.

Hills, for instance. We have all seen buildings hanging partially in mid air because they were placed on a sloped surface, yet with proper procedural generation, buildings could automatically compensate for that to fit perfectly with the contours of the scenery.

That could also be applied to water and roads, and there are already numerous demonstrations of procedural engines doing just that to create very complex cities based on fairly available real-life grid and road data. This is happening at the same time that Gpus are being designed to take more and more of the burden from the Cpu for mathematical calculation as well as simple graphics rendering.

Your concerns about the overhead required, and about what would remain for aircraft physics and etc may point to a problem, but I am not certain that those concerns have been validated yet. Even if the Outerra engine is not the answer, it might point in the directions of an answer.

  • Upvote 3
Link to comment
Share on other sites

I think many people are responding to the potentials of the technology rather than specifically to the Outerra engine itself. Recently, more and more programs have been demonstrating the power of procedural generation to automate tasks that seem to require painstaking manual effort with current technology.

Actually fractal techniques for flight simming/terrain is nothing new.. in fact before the GPU it was fairly common technique for having variety in terrain for a lot of games. Just think about the lowered disc and memory requirements alone, when it is all replaced by a simple iterative function. And I think it's fair to say that Aerosoft won't be using the current autogen model. As humans and as sim pilots we've only come to expect accuarcy in certain details like mesh/terrain.. the rest I think is open game for change.

  • Upvote 1
Link to comment
Share on other sites

Actually fractal techniques for flight simming/terrain is nothing new.. in fact before the GPU it was fairly common technique for having variety in terrain for a lot of games. Just think about the lowered disc and memory requirements alone, when it is all replaced by a simple iterative function. And I think it's fair to say that Aerosoft won't be using the current autogen model. As humans and as sim pilots we've only come to expect accuarcy in certain details like mesh/terrain.. the rest I think is open game for change.

I was trying not to be pedantic! :P

I actually remember some of the very early fractal use in games like the Eidolon, Rescue on Fractalus and even Koronis Rift. I have also kept pretty close eye on developments since then, especially the disk and memory requirements, compression and other fun goodies. Given Aerosofts tight development time and the ability of procedural generation to pump out huge swaths of super-detailed territory at record speed, it seemed to me from the beginning to be something they would look at with at least mild interest.

  • Upvote 1
Link to comment
Share on other sites

Hi fellow Simmers,

to be honest, I don't understand the hype concerning the Outerra engine. For me there is nothing "overwhelming" to it. Personally there is no difference to my FSX setup with FS Global 2010! The terrain in my sim is just fine for me. Flying over the Alps makes me feel like I am there! What could be better with the hardware I own?

What we must understand is as ACES developed FSX they had to think into the future concerning hardware performance. I can still rememder all the bad critics about poor performance, and how the ACES Team explained to the people that they should not set the sliders to max if they don't have the hardware power.

Looking now in 2010 any standard PC can run FSX with good FPS. We don't know what will be standard in 2012, but with upcoming 6-Core CPUs, GPUs with tremendous rendering capabilities I see no problem, and by then the Outerra engine will be outdated, as I assume that 3-D presentations will be the standard.

Serious simmulators such as Flight, Road, Space, Ships will always need the best hardware, and its a hobby that cost money if you are dedicated to it.

I know this might be off topic, but it seems that most simmers forget how things were a few years back.

Cheers to all......

  • Upvote 1
Link to comment
Share on other sites

As far as what I can tell the Outerra engine is entirely devoid of atmosphere. Since we are discussing a flight simulator and not land based vehicle simulators I would think that what is more relevant is the fluid dynamic model as well as the weather model.

Link to comment
Share on other sites

Hi fellow Simmers,

to be honest, I don't understand the hype concerning the Outerra engine. For me there is nothing "overwhelming" to it. Personally there is no difference to my FSX setup with FS Global 2010! The terrain in my sim is just fine for me. Flying over the Alps makes me feel like I am there! What could be better with the hardware I own?

What we must understand is as ACES developed FSX they had to think into the future concerning hardware performance. I can still rememder all the bad critics about poor performance, and how the ACES Team explained to the people that they should not set the sliders to max if they don't have the hardware power.

Looking now in 2010 any standard PC can run FSX with good FPS. We don't know what will be standard in 2012, but with upcoming 6-Core CPUs, GPUs with tremendous rendering capabilities I see no problem, and by then the Outerra engine will be outdated, as I assume that 3-D presentations will be the standard.

Serious simmulators such as Flight, Road, Space, Ships will always need the best hardware, and its a hobby that cost money if you are dedicated to it.

I know this might be off topic, but it seems that most simmers forget how things were a few years back.

Cheers to all......

I think the point was that the performance doesn't scale with the architecture because of the fundamental flaws in assumptions made when FSX was under development about where hardware was going. FSX is still dependant not on co-processing, but raw CPU speed. Therefore while some gains are to be seen when upping the hardware ante, they are disproportionate to the costs. Some older technophiles might remember not just Moore's Law and how it proposed that the number of transistors able to be incorporated into a reasonably-priced integrated circuit doubled roughly every two years (as distinct from how it has come to be commonly misinterpreted as processor speed doubling every eighteen months) but how we used to use the `50/50 rule` to define when best to make that upgrade - i.e. you wait until the hardware you have now is half the price it was when you bought it, then you buy for the same price as you did before and get a doubling in performance. 50 percent doubled meant a 100 percent increase in game performance.

That is simply NOT TRUE for FSX, not even after SP2. Some CPU upgrades (along with appropriate RAM and MOBO) can see in-game performance double, but it is certainly NOT true of GPU performance. And to upgrade BOTH CPU and GPU to achieve a doubbling of consistent fps requires far more than the 50/50 rule dictates.

Or to put it another, simpler way - the Law of Diminishing Returns has hit FSX. Hard.

A new game engine may certainly solve that problem today, but unless it is integrated in such a way as to permit those linear increases to match the hardware then all we will see is FSX-Repeated - a Northbridge Too Far, if you will... :mellow:

I can certainly see and appreciate the developers point to Outerra, but as you say it is perefectly possible to have the SAME level of detail in FSX today, and unless and until the visual and simulation experience doubles, yet the `price` remains consistent in terms of fps cost, or hardware cost to achieve that fps then I don't see that Outerra has anything to offer.

The fundamental schism for me is the current three-pronged approach to texture renders - Mesh+Terrain+Autogen is a very resource-consuming method of displaying assimilated data, but one which the inherited - and inherent - limitations of FSX made necessary. Again, in simple terms if a terrain engine is capable of rendering at the centimetric level, then it MUST, by implication render not just ground imagery but also 3d objects that interract with that ground - hedges, fences, walls, trees, etc. etc. Outerra has yet to display a city scape, nor has it yet enables the weather engine, cloudscape flight dynamics modelling or variation in land, sea and air to meet the needs of a Global simulator operating through all seasons, including extremes.

I am extremely dubious that these aspects can be added to the basic engine without it too, bogging down under the sheer weight of processing.

I don't pretend to know what the answer IS, only that as long as the Earth is simulated in the manner we simmers have come to expect with selectable locations, time, dates and weather, then the exponential load on a fractal engine will probably show an exponential rather than linear reduction in performance as the information load scales, and NOT show a linear performance increase in line with the general performance of the GPU and CPU.

However, if the engine is designed in such a way as to mazimise the inherent strengths of TODAYS hardware, rather than guseeing at tomorrows, then the results might be more predictable. Those Crystal Balls have proved to be somewhat murky of late - SSD's were not even on the horizon when FSX was `born`... yet perhaps a SSD-based simualtor product - sold ON an SSD rather than DVD - might allow for faster, better optimisation...

Link to comment
Share on other sites

You've mentioned the exponential load twice of the fractal engine without giving any proof that this is the case.. could you please use big Oh notation so the rest of us can understand your reasoning?

Link to comment
Share on other sites

I wish I could find the link, but there was an Italian University that was experimenting with fractal engines and high-detail mesh to create a `cityscape` where the buildings were rendered in 3d outline. All was fine while the `pure` rendering was taking place, but their project stalled when they attempted to place texture on the rendered base and then simulate the effects of light and shade. It was concluded, IIRC that the the fractal model worked extremely well on a unified single light source, but that the co-processing required to place a lit, rendered image on that fractal-generated model was disproportionately `expensive` in terms of GPU load. Modern GPU's can serve a very useful second function as a math co-processor, but not do that AND run a polysurface multiple-lit image with object interraction at the same time. That is the reason behind Shafer Model language - it can `get away` with rendering a 3d image at a uniplanar level, so you only see the back of thehouse when your viewpoint says you shold see the back of the house...

I assume from readng the Outerra featuresets that, as they stress the fractal algorithms only for the finer detail that the processing requirement quickly reaches the saturation limit on modern GPU's.

A flight sim engine that only runs on Crossfire or SLI, and then only until you load the rest of the sim up with aircraft, AI and climate ain't gonna change the world, much less render it in exquisite motion ...

But these are early days and who knows what DX11, Shader Models and the massive bandwidth of modern GPU's will come up with. But whatever it is flight sim engines need to go with the flow, not try and dig a new stream.

Link to comment
Share on other sites

  • Aerosoft

You've mentioned the exponential load twice of the fractal engine without giving any proof that this is the case.. could you please use big Oh notation so the rest of us can understand your reasoning?

If I tell you it just won't work, will that help? we tested several engine that rely heavy on fractals and have found out that when push comes to shove old fashioned LODs still work better. It remains a technology that promises more then it can deliver.

See just use the engines out there, they all got demo's. See how your CPU load is. Check out the GREAT weather engines that move clouds looking super realistic around the sky. Check out how some engines can show the whole planet and zoom down to a blade of grass. Check what they do on a quad core system and keep in mind you got about 10% resources for weather and 30% for 'natural world' stuff (an airport is way way way more complex then fractalizing a landscape as it does not follow rules). Any new sims major problem is managing the resources. No matter what you hope for, things do not get a lot better then what FSX does now, in contrast to what people think it is pretty fast if you look how much it spends per polygon. What can be done is shuffling the resources a lot better and even a small step on FSX will make a major difference in many aspects. Snave is putting things (as always) rather harsh, but take my word for it, he knows what he is taking about.

Now DX11 might change that, the jury is still out on that. I am sure hardware will catch up to the idea, just not in the next year.

Link to comment
Share on other sites

  • Aerosoft

If it's going to be based off of any engines it'll probably be Quake I-IV..... :) for good reason. ie the source code is opensource and it's a good starting point.

And totally useless if you want sight lines of 100 miles or more on a curved surface. So for very good reasons it took only minutes for us to check and reject that as an engine. Very nice for a confined shooter, but not for a serious sim. Let's keep things in perspective here ok?

Link to comment
Share on other sites

If I tell you it just won't work, will that help? we tested several engine that rely heavy on fractals and have found out that when push comes to shove old fashioned LODs still work better. It remains a technology that promises more then it can deliver.

Outerra is not like other systems (presumably) since he's doing it all on the GPU and not CPU! So that leave a lot of CPU power for modeling flight dynamics, systems logic etc.

# Fractal algorithms used to generate finer resolution maps are running entirely on GPU.

# Terrain textures are generated on the GPU from the fractal data using per-pixel computed elevation, slope and terrain curvature; advanced material mapper will also utilize climatic data, bedrock maps etc. to more accurately texture the world.

# Fully asynchronous 3D engine, majority of the algorithms running on the GPU

Please keep an open mind and at least give him the opportunity to show you some benchmarks.

Link to comment
Share on other sites

Thats the thing if you look at the current Quake 5 engine its going into finer detail of textures all the time just like fractal grass would, without ANY loss in performance, simply using shaders. And I might add the draw distance is as good as anything for at least the ground. He says he can do the same with Polygons for the next engine, in the million+ count. I mean your talking about future implementation when they already exist using graphics cards that are 5 years old now. But then again he did invent the genre, and by all counts is the best graphics programmer on the planet.

Mathijs: All i said was it's a starting point for ideas.. didn't say you'd use the engine out right, I know that's not even remotely possible :) I would naturally assume you'd take what's useful and discard the rest.... and obviously UpdatePhysics(); and RenderFrame(); would need to be written from scratch... but theres code in there for all sorts of things! When I'm writing code there's nothing better than seeing an implementation done by others, it just reaffirms and reassures what you are doing is right all along. I'm personally using the network code for one of my projects. Sure there's plenty of books on the topic... but to see it working in action is a totally another thing.

Link to comment
Share on other sites

Thank you Mathijs for your contribution to this dicussion. I'm rather new and its nice to see active Aerosoft participation.

I would really like to know what "fractal base engine" you have tested out. I would really like to benchmark these myself. Is there anything current? I have search and found a few, but their core and updates seem to be rather old and undeveloped as of late.

Thank you!

  • Upvote 1
Link to comment
Share on other sites

As I read the fractal model, it should be at its most beneficial in ground-based simulations and games where the centimetric detail could make a substantial difference if you are, say, crawling through the woods pushing the bushes and leaves out of the way as you make your way into the enemies camp, etc. etc. But the planar point of reference and the angular vectors involved render that level of detail unnecessary for a flight simulation. If one is looking that closely the individual leaves on a bush, one is already heading for a reset! :rolleyes:

I suspect the LOD model comes in for a lot of stick not because it itself is poor, outmoded or system-heavy, but because the implementation of it is poor in FSX, for all the reasons previously expounded. Simulation of flight requires not point-detail at the highest resolution, degrading into distance through the field--of-view of that moment, but wide swathes of concurrent LOD, representing the `circles of increasing distance` all around the aircraft.

The problem that needs to be overcome with the LOD model is a technical challenge, but possibly surmountable: The FSX engine treats that `circle` LOD as a `ball` around the aircraft, so it affects not just ground detail, but sky, local autogen and AI. Which is why you can't have ground fog in the distance without some clever addons... When ACES patched FSX they deliberately removed some of that LOD on autogen to ease the display rendering requirement, and in so doing created a problem whereby if the rendering is turned DOWN on lower-spec computers to aid performance, the autogen LOD doesn't match the terrain LOD, and it simply looks horrid!

In essence, the LOD is the wrong way round for a real-world perspective - in the real world items closer to the fast-moving aircraft are blurred, then there is a panoramic field of view that is detailed, followed by a (LOD-correct) loss of that detail into the distance, as well as atmospheric interference. The MS FS engine has never been able to replicate that, yet `motion blur` is a common feature of car racing sims.

Perhaps, form a theoretical viewpoint, we have been looking at this all wrong? What we need in simulation of aircraft is a reversed-LOD model, or induced motion blur to lower the detail load on the GPU at the very highest resolutions, where the rendering impact is greatest?

One thing I constantly notice on zooming in and out in FSX in the VC is that at the `correct` zoom level for scale purposes, the detail INCREASES as one gets closer, whereas in the real world it decreases as one gets closer and the impact of the angular velocities on detail perception is felt. Only when zooming OUT to an artificially-wide zoom level does there appear the first suggestion of motion blur, but at the point where now the scale seems out-of-phase.

It's actually an interesting (and quick!) experiment to try:- set up a default flight with the aircraft on the ILS at 1,500ft AGL, with the wide aspect ratio set to TRUE. As you get lower and closer to the runway zoom in a couple of clicks and feel the greater reality of the visual perspective spoiled by the absence of motion cues. Now repeat the exercise, but this time with the starting aspect ratio set to FALSE so that the starting view position is nose-in on the instruments. Now zoom OUT two or three clicks and again feel how as the field of view increases the perception of speed improves, but only at the cost of true depth perception and distance. The baseline difference between the starting points of TRUE versus FALSE aspect ratio is actually just three clicks of the zoom button - zoom IN three clicks from the wide FOV and you are at the FALSE point of view, and vice versa.

We normally consider this to be a fundamental limitation of trying to display the panoramic view out the window through a letterbox and an inherent limitation of the simulator on the PC unless we move to ultra-widescreen and multi-monitor setups, but if you HAVE a multi-monitor setup and are prepared to change the `obvious` Wide Aspect Ratio from `true` to `false` (and therefore flying in the face of convention) what you actually find is the same holds true. It's not just the visual perception of `width` that is holding things back, it may also be the rendering conventions of scale-v-LOD-v-motion.

A flight-dedicated LOD model that could address all these issues by inducing motion-blur so as to reduce the very highest spot loads on the GPU at a reduced level of detail might actually `feel` more realistic at all fields of view... Or alternatively it might just be a question of changing the scale of the rendered objects. Trees are too big in FSX, buildings too small to provide coverage, but are they perhaps inducing the error. Is there a defined and particular reason for the choice of scale in the sim? If all you want is more detail all you need is increased per-pixel data. And manipulating that at the pixel level is what modern GPU's are for.

  • Upvote 1
Link to comment
Share on other sites

As I read the fractal model, it should be at its most beneficial in ground-based simulations and games where the centimetric detail could make a substantial difference if you are, say, crawling through the woods pushing the bushes and leaves out of the way as you make your way into the enemies camp, etc. etc. But the planar point of reference and the angular vectors involved render that level of detail unnecessary for a flight simulation. If one is looking that closely the individual leaves on a bush, one is already heading for a reset! :rolleyes:

I suspect the LOD model comes in for a lot of stick not because it itself is poor, outmoded or system-heavy, but because the implementation of it is poor in FSX, for all the reasons previously expounded. Simulation of flight requires not point-detail at the highest resolution, degrading into distance through the field--of-view of that moment, but wide swathes of concurrent LOD, representing the `circles of increasing distance` all around the aircraft.

The problem that needs to be overcome with the LOD model is a technical challenge, but possibly surmountable: The FSX engine treats that `circle` LOD as a `ball` around the aircraft, so it affects not just ground detail, but sky, local autogen and AI. Which is why you can't have ground fog in the distance without some clever addons... When ACES patched FSX they deliberately removed some of that LOD on autogen to ease the display rendering requirement, and in so doing created a problem whereby if the rendering is turned DOWN on lower-spec computers to aid performance, the autogen LOD doesn't match the terrain LOD, and it simply looks horrid!

In essence, the LOD is the wrong way round for a real-world perspective - in the real world items closer to the fast-moving aircraft are blurred, then there is a panoramic field of view that is detailed, followed by a (LOD-correct) loss of that detail into the distance, as well as atmospheric interference. The MS FS engine has never been able to replicate that, yet `motion blur` is a common feature of car racing sims.

Perhaps, form a theoretical viewpoint, we have been looking at this all wrong? What we need in simulation of aircraft is a reversed-LOD model, or induced motion blur to lower the detail load on the GPU at the very highest resolutions, where the rendering impact is greatest?

One thing I constantly notice on zooming in and out in FSX in the VC is that at the `correct` zoom level for scale purposes, the detail INCREASES as one gets closer, whereas in the real world it decreases as one gets closer and the impact of the angular velocities on detail perception is felt. Only when zooming OUT to an artificially-wide zoom level does there appear the first suggestion of motion blur, but at the point where now the scale seems out-of-phase.

It's actually an interesting (and quick!) experiment to try:- set up a default flight with the aircraft on the ILS at 1,500ft AGL, with the wide aspect ratio set to TRUE. As you get lower and closer to the runway zoom in a couple of clicks and feel the greater reality of the visual perspective spoiled by the absence of motion cues. Now repeat the exercise, but this time with the starting aspect ratio set to FALSE so that the starting view position is nose-in on the instruments. Now zoom OUT two or three clicks and again feel how as the field of view increases the perception of speed improves, but only at the cost of true depth perception and distance. The baseline difference between the starting points of TRUE versus FALSE aspect ratio is actually just three clicks of the zoom button - zoom IN three clicks from the wide FOV and you are at the FALSE point of view, and vice versa.

We normally consider this to be a fundamental limitation of trying to display the panoramic view out the window through a letterbox and an inherent limitation of the simulator on the PC unless we move to ultra-widescreen and multi-monitor setups, but if you HAVE a multi-monitor setup and are prepared to change the `obvious` Wide Aspect Ratio from `true` to `false` (and therefore flying in the face of convention) what you actually find is the same holds true. It's not just the visual perception of `width` that is holding things back, it may also be the rendering conventions of scale-v-LOD-v-motion.

A flight-dedicated LOD model that could address all these issues by inducing motion-blur so as to reduce the very highest spot loads on the GPU at a reduced level of detail might actually `feel` more realistic at all fields of view... Or alternatively it might just be a question of changing the scale of the rendered objects. Trees are too big in FSX, buildings too small to provide coverage, but are they perhaps inducing the error. Is there a defined and particular reason for the choice of scale in the sim? If all you want is more detail all you need is increased per-pixel data. And manipulating that at the pixel level is what modern GPU's are for.

Very good "perspective" here Simon. In fact, motion blur is very complex for technology to duplicate a real world effect for our immersion experience, and certainly far too problematic for FSX to handle accurately.

Just look at the latest crop of LCD/Plasma TVs; manufactureres are trying to one up each other with ever higher refresh rates but the simple fact is that higher refresh rates are actually diminishing our viewing perception of high speed events. In reality, our eyes don't "see" details above 30 or so FPS for high speed events (i.e. just have a look at a ceiling fan - you can't discern the blades while they are in motion) yet when you watch an F1 race at 240 Hz refresh on a 55inch LCD it's certainly clear (and th problem here is, - it's clear everywhere on the picture, not just on your focul point) to be sure, but unrealistically so.

As Simon suggests, there are two ways you can go depending on what you want out of LOD from FSX...

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. Privacy Policy & Terms of Use