Developing visuals for a game has always been a tradeoff between visual fidelity and performance. Why I believe this to be a fantastic achievement is that visual fidelity is going to be high when visual fidelity matters and performance is going to take precedence when performance matters. People tend to look at visuals when the screen is not busy with a ton of other stuff going on - when exploring, taking in scenery etc. With dynamic resolution, in these situations, visual fidelity can afford to be higher than it could ever be without it.

On the other hand, visuals tend to be a lot less essential when you get into the action - especially for hack and slash games, FPS games etc. Resolution of the picture is the least of your worries when you're shot at from all sides and you need to react quickly - and that's where it drops and gives way to better performance in order to allow you to better cope with unexpected. When things then calm down and you can take your time to actually aim even over a distance, resolution gets upped again as your screen is not as busy in such situations.

It seems to me that you get the best of both worlds - high performance for action and high visual fidelity for when things calm down.

It's not important tho. I bet 90% of people playing videogames don't even have a clue about what 'Resolution' stands for, exactly. And for most of those people, how the game 'feels' (in other words, performance) is going to be a lot more important.

Just going to leave this here.

Right. Because very often only a very small amount of objects or effects actually scale in density with the resolution.

But because there's a "console target", because you can scale without all that much lack of precision downwards, and because people prefer that static wallpaper in HD (whatever that means at the moment) over a grainy shader or something else that was more interesting. Since you typically can't sell "better scaling objects as they approach" with any luck over "everything mysteriously appears to the player at the same distance, always". Then that's what you're going to choose.

And no, "scam" is not a wrong word to describe dynamic scaling at all. It's a "technology" designed to mitigate often pretty bad design-choices. In the sense that you want to make it easier to avoid framedrops, by just applying a magic automated solution. Which is valuable to a lot of companies, because they want to put as much of those expensive static assets at the detail the artists create them, into a very unintelligent rendering engine. So instead of going for a different approach to rendering engines, and coding in each project, you want to just put in something automated. That then certain people are willing to believe can make jewelry out of potato chips. When of course that's not actually possible without losing visual quality.

But if you sell something like "yeah, we guarantee that the rendering time of each frame now never drops below 16.66ms", then you're gong to produce interest. And people will want to believe that this is a great solution. That.. doesn't require specific tweaks in very narrow contexts at all, etc. :p

I'd support a game that designed for PCs first and give the player choices for those with low med and high end systems. I never bought a console but a PS3 was given free to ages ago for buying a laptop. It makes a great blueray player. I'm an old PC gamer. Nothing will change that. If that means I'm stuck with ancient games or believe that one day an odd underdog creative game company comes around? So be it!

*sigh*
[url=Scam]http://www.merriam-webster.com/dictionary/scam[/url] Dynamic resolution scaling is neither. There are bloody articles in gaming magazines about it - it's about as deceptive as people being shocked by a car containing a steering wheel. Anyway,

It's also a technology which allows for higher graphical fidelity on weaker hardware. Of course, lazy implementation will be lazy for any tech. If you have a better tech which solves what you perceive to be underlying issue ready, there's nobody holding you back - create it, feel free to start selling it. Anyway, "Magic automated solution" is a holy grail of any programmer.

i think that with slow games like point and click, be it in pixel style like wadjet eye games ( i have all the games)
should have good nice backgrounds cause you be staring at these for 90% of the time.

So thats why old syberia 1+2 are good examples of old games with good backdrops.
When playing normal (usally 2d , no 3d rotating) adventure games, graphics are a very important part of the game and backgrounds are very important aswell, so the quality needs to be as high as possible.

As for action games: no one care if the background is pixel, ugly, or whatever, like you said: when your live is at stake in the game, who careas or has the time to see how the graphicslook, cause surviving and slayin the enemies has priority one.

And like the holy grail - it may just not exist. At least not easily. I'm don't know any internals about dynamic upscaling to say where exactly it may go wrong but I can imagine a number of possible ways (wrong choice of assets, usage of assets for fixed low resolution, extensive use of effects while neglecting details, using wrong and time consuming effects, taking the target resolution only into account at the latest stage, ...) so that an automated solution without tweaking could result in a bad quality output. Kind of adaptive solution is surely possible but magically guessing the right tweaking parameters to give optimal results, I doubt it. Although during testing maybe nearly optimal parameters for a wide range of systems could be found and then later applied.

Best proposal I heard was a sort of thought-experiment - you'd have automated bots running through the scene to measure the polygon budget, that then would be used in the build step to create new static objects at the right level of detail towards a specific hardware target.

But the weakness of that approach is that when you add more objects, you increase the need for storage by a very, very large and unpredictable amount. And it also limits the amount of actually dynamic objects and effects you can have. So you end up with something like what DICE did with Battlefield 4 for the xbox, where they essentially went through the entire game and built static lighting effects to replace the engine-created effects you find in the pc and.. well.. the ps3 version. This "fixes" the framerate issues, but it replaces the actually dynamic effects.

For example, imagine a kind of shader-intensive effects for light shafts that depend on the scene location and where you stand in it. It could be created to run on fairly modest hardware, over a certain limit. But if you wanted to remove the lower requirement for shader capacity, you would scale the scene down so far that it just becomes a complete blur. So you could keep the effect, but at such a low resolution in the whole scene (which is where that magic code would have to be looking at the scene from) that the result is worse than turning the whole effect off.

Which is why DICE went so far as to go through the entire game and create static effects manually in all the cutscenes specially and also in the rest of the game - to keep the scene detail as high as possible. Meanwhile, using a type of internal dynamic scaling as the detail increases has actually been used extensively on console-ports several times before. Several call of duty titles for example actually have half-resolution vertically applied at the external pipeline, after the rendering and placement of objects happen, to maintain the framerate. So this stuff isn't new by any means.


Point is that adaptive v-sync, just as dynamic scaling, essentially accepts as inevitable that graphics engines will have infinite overdraw. It's completely steeped in the convention where you think the only way to improve framerates is an extensive manual process where you optimize maps by plugging holes, creating location-dependent streaming, and more handcrafted static resources. While you have to design games as corridors that you move through in a more predictable way to spend the polygon budget more optimally, etc.

So if you want artists and designers to have more freedom, if you come from this point of view, then you will want to circumvent the limitations of the design conventions (and reduce the costs of development) by creative uses of dynamic scaling and detail, etc. You think that then it's possible to have higher detail in the scene overall, so you can potentially exploit the hardware to a larger degree. I.e., taking seriously that for the most part in a scene, you can actually run it perfectly fine, but then that peak loads are cushioned enough to get acceptable framerates.

These views typically come from the "visual science" school. And they're not completely without merit. But they don't really understand that the graphics pipeline and the way a graphics context is created already will depend on a lot of scaling and dynamic detail trickery anyway. So a general "magic" approach that simply removes actual resource-handling on the engine level is just an invitation to continue using an approach that has very obvious limits. Basically, if you didn't have game-logic/engine level detail handling, then it doesn't take very long before these general tweaks would dynamically scale the entire scene to a complete blur from trying to draw some distant object you don't perhaps even see. It's just not a solution, and in practice the tweaks also would be on very specific parts of the graphics pipeline anyway.

Bottom line - there's no magic general high-level solution to this that can work externally on the graphics pipeline, simple as that.

Look, I understand that you consider this a patchwork non-solution. And I agree, to an extent at any rate. Hell, we're going this way for ages and instead of striving for universal solution we invent things that effectively add work like distant LOD.

On the other hand, I'm a realist - until we reach the limits of this approach or until somebody successfully introduces a viable alternative, we'll be stuck where we are now. Might as well try to squeeze as much performance as possible while we're at it. (and yes, I do realize you consider postponing the inevitable to be another issue.) As for 'everything becoming a blur', resolution scaling doesn't have to be unlimited, obviously. On the other hand I am worried about it becoming one size fits all solution which would both invite lazy design and negatively influence gameplay. But... We'll see where it goes.

I wonder what time period you're talking about. I only encountered scalable vector based UIs very late, actually, around 2006 or something. The vast majority of old games without an upper limit for the resolution draw the GUI elements in their native resolution, so they appear tiny but in the correct spot due to using coordinates relative to the screen border (i.e. Hitman, Jedi Knight, Half-Life and zillions of other games) which I wouldn't quite call scalable due to risk of the elements becoming too small to be readable. Also many (or even most?) new games which don't have a contemporary or sci-fi setting stick to sprite-based HUDs and use border coordinates as reference points and may have readability issues on 4k screens.

And I know very few old games that scale GUI elements up to always take up the same portion of the screen (such as Incoming, even though its HUD would have been perfectly suitable for vector graphics) and see it more often in newer games (especially multi-platform ones) that stick to sprite-based GUIs as well as HD ports made for consoles. And then the GUI actually stands out as being extremely pixelated or blurry if the original target resolution was too low (currently playing God of War II on PS3 where the HUD is quite pixelated due to this).

I thought the reason Rage looked like shit was because id wanted to squeeze the game onto the 360 and couldn't be fucked to produce better quality versions for the PS3 and PC. They seemed to really hate the PC at that point in time. Maybe things have improved since daydreamer Carmack has left. Doom certainly looks a hell of a lot more promising than it used to, so maybe.

...the really good old days? When the grass was greener and the snow whiter, and so on :p

No, pretty common when you had different resolutions on PC tops. Several of the gamecube/console crossovers were pretty consistent too. Since the first 3d cards ran typically on 640x480, 800x600, 1024x768 maybe, while the gamecube and ps2 target could run in widescreen or letterbox, etc. Either you got that evil "supported resolutions" stuff with handcrafted UI to different resolutions. Or something that stretched to resolutions with artefacts (somewhat better). Or you had, like you see in a lot of smaller titles now that are supposed to run in windowed mode, 4:3 and 16:9, or fullscreen, and anything in between, where the UI is made up of geometry of some sort or vector-based graphics with relative coordinates to the center, and so on. Guess you don't actually "see" it right away in newer titles, but overlays in the game, text in objects, things like that, usually scale like this. Dead Space has a lot of this. DICE's Frostbite engine games deal with this really well. Puzzle mini-games in Warframe work like that as well..

The funny thing is that the less visible things sometimes scale really badly. So if you play with AA off, even when you can put it on, because you don't really like the blur all that much, etc. Then you tend to notice when the textures wrap badly, or when they're switched out, how some shader effects look great from a very specific angle and distance (but nowhere else) and things like that. That's where a lot of graphics engine work can be done. But ..sort of turns out that people don't notice or appreciate things that scale amazingly - if they're not photographic, so to speak, when it covers the entire screen..

Btw, the ps2 emulation for the ps3 that actually decoded the function calls (like Shadow of the Colossus or the God Hand setup.. based around the same virtualisation method, apparently. But not all the titles use that emulation package.. not sure why, but they had a similar one for the ps1 they also scrapped for an emulator similar to what you have on PCs. Very strange, since the emulator that God Hand runs on is an actual emulator that was general, without having to cut out effects, and end up with misaligned and badly scaled objects if the internal resolution was different, etc), these drew the layers on the screen in different passes. 3d context in one resolution, scaled in some way, UI in another. So the UI would scale that way, like it actually did on the ps2. Are a lot of different solutions to this.

Yeah, I'm aware of this method, I didn't mention it because it appears comparably new to me, though (new as in early 2000s while I thought of the earlier 3D games from the mid to late 90's). Yes, PS2 and Gamecube generation games already had a lot of this and funnily their ports usually suffer from smaller scaling issues than even dedicated PC games from that time but may get the artifacts you mentioned if you turn on AA.

bump

i read all posts so far and instead of replying to about 15 of them, i'll just respond in general. as some of the posts are similar. and yes id tech 5 engine is what i meant with modified doom 3 engine. also one could call it current version of quake engine :).

it seems to me like many other posts are missing the point partly or full and/or don't seem to know what the thread is about and/or didn't read the and my [url=http://www.gog.com/forum/general/dynamic_scaling_a_new_scam_of_videogame_industry/post10]2nd response.

so its neccessary to explain, what is meant by general or static game engine scaling. which many users were posting about. its about density, polygon count, resolution and amount in which objects, textures, npcs, enemies, other players, general surfaces and graphical effects are rendered, depending on the position and distance relative to the player. you can partly adjust the scales for this in the settings. but they're fixed during gaming and don't readjust themselfes, therefore static. also this scaling is independent of the rendered game picture resolution. this is an old standart in video game industry.

dynamic scaling isn't that old and its different. aside from messiah, i know only of id tech 5 based games such as rage which are using it. its getting widely used only since current gen consoles release. it means that the above explained scaling and/or rendered game pic resolution readjust itself on the fly by cpu/gpu to maintain a stable framerate. if fps gets lower than x then the engine scales down (less density, polygon count, resolution and amount) and/or the rendered game pic resolution gets lower. if fps gets higher than x then the engine scales up and/or the rendered game pic resolution gets higher.

we're talking here only about resolution related dynamic scaling. therefore, to be more precisely, messiah doesn't count as an example here as its only lod related dynamic scaling.

then there is a scaling of the rendered game pic to fit on the display, which is made by software and cpu (ps3) or hardware (ps4, xb1, xb360) or gfx (pc). also tvs are able to scale a picture from some formats/resolutions to some other formats/resolutions. fenix explained it . although i believe the picture link in that post doesn't show a scaled but just a rough stretched picture. here's an example of [url=http://media.redgamingtech.com/rgt-website/2014/06/tomb-raider-1440p-vs-1080p-close-up.png]the difference between native 1440p and 1080p scaled to 1440p. this is independent of resolution related dynamic scaling and was used before current gen consoles.

here's a list of last gen consoles, you can see that many games were upscaled after rendering, there's no resolution related dynamic scaling.

in fact a game for example for ps4 could use dynamic resolution between 720p and 900p and the hardware scaler of ps4 would scale the rendered pictures from between 720p and 900p to 1080p.

here's a general brush up about framerate, resolution and upscaling a picture.

as mentioned in the op i see 4 ways of how resolution related dynamic scaling works:

1. the full game pictures at the end get rendered at lower resolution. here's a good example, the picture shows the difference between native 1440p and 1080p upscaled to 1440p. this is called dynamic resolution.

2. the textures inside the game getting rendered at lower resolution, kinda dynamic lod for textures. full game pictures are still rendered at full resolution. rage and games on id tech 5 engine are using this and here you can see how shitty it performed at release of rage.

3. game objects with textures getting rendered at lower resolutions. full game pictures are still rendered at full resolution. don't have an example to present :(.

4. a mix of all 3 methods above.

here's a short explanation about dynamic resolution.

this is a good article about dynamic resolution and some context to its appearance on new gen consoles.

then there's also temporal reprojection a complex resolution related dynamic scaling technique, which is used by killzone shadow fall.

„...Here Guerrilla Games has opted for a 960x1080 framebuffer, in pursuit of a 60fps refresh. ... Now, there are some mitigating factors here. Shadow Fall uses a horizontal interlace, with every other column of pixels generated using a temporal upscale - in effect, information from previously rendered frames is used to plug the gaps. ...“ source

devs of shadow fall got sued because they did advertise their title as a full hd game. but this could fill an own thread.

and although dynamic resolution isn't scam by itself, its a scam how the dev/pub and console manufacturer are handling the usage of it in their games. so everytime a ps4 or xb1 game, using dynamic resolution, gets released and the dev/pub with support by or pressure by console manufacturer is officially stating that a game runs at 1080p, its a scam. hell actually the resolution in which the game renders and if its using dynamic resolution, should be mentioned on the box. but there's no such information on xb1 games. and i thought that ps4 games contained at least a resolution info, but its just to which resolutions the game pic gets scaled after rendering.

call of duty black ops 3 on ps4 is one example. there is no official statement, at least at release, that the game is using dynamic resolution. its uncovered by eurogamer after release.

i saw a thread on another forum in summer 2014, can't find it anymore. it listed many games that were already released and were using dynamic resolution on ps4 and/or xb1. i can remember one of them was tomb raider for both consoles.