• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

DF: The Medium Tech Review: A closer look at Xbox's First Next-Gen Game

sinnergy

Member
Really liked what I played the first hour, great looking game for such a small team, looks great on series X, next gen looking ! Pixels don’t matter right , it’s quality of the pixels that matters.
 

Clear

Gold Member
I'm with you that the "2 games at once" is a hinky way to describe it (and if that's really how Bloober is doing it, I'd be curious why?), but you're talking about the final output; you still need to get through all the game processes before you render those 4k of pixels. To do that, like you said, they need to have the scene built twice, but just because it's half the pixels doesn't mean it's perfectly half the work. You still need pretty much all of the geometry (you could cut it back but then you'd need full-detail on any frame that you cut to full-scene,) you'd need the textures, and you'd need to make all the passes of the elements you'd do at full-screen (again, at some measure of quality if you can get by when squinting at a split-screen doesn't show the fine details, but I would imagine the quality would still need to be high because compromises would probably stand out.) And your second screen has a lot of differences even though the geometry is very similar, so you don't have the advantage that a full-frame does of having one scene to work with; if you're doing the work twice, every surface that informs how another surface behaves has to do that work twice even though it's the same set of surfaces in both scenes.

If the work of doing a split-screen effect were perfectly half, we'd still have split-screen in our racing games and co-op action games, but it's not 1/2, it's something more like an exponential difference (or fraction of exponential) in how much it takes to do a complex scene twice in a frame, even though the final output will be a set resolution.

Its not that simple at all.

For instance in a traditional split-screen arrangement the issue becomes what if every pane needs to display the worst-case scenario in terms of draw load. In say a FPS that would mean not only would the scene geometry have to be rendered (which is going to fluctuate based on whats in the view frustum) but what if each player was facing all the other players making them visible also.
Because there's no way around it without compromising player agency, handling that without the frame-rate buckling becomes a primary concern.

The thing about The Medium is that its all scripted, Bloober have total control over draw load because they set the time, the location, even the camera angle, and they only have one player character to contend with. At the end of the day the actual game-part is pretty trivial, almost no combat, no need for complex AI, and relatively few proper onscreen characters from what I've seen. There's some decently complex looking particle vfx, but they are confined to the spirit world with the real world being pretty static.

I'm not trying to bash them, but as I wrote its not technically that impressive to simply render two datasets per frame. It just requires more data and vram.

Doing it efficiently as well as artfully is worthy of praise, but the Medium doesn't seem to perform that well even on top of the line hardware.
 

DeepEnigma

Gold Member
Really liked what I played the first hour, great looking game for such a small team, looks great on series X, next gen looking ! Pixels don’t matter right , it’s quality of the pixels that matters.
 

Vae_Victis

Banned
Split-screen is one thing (and usually with split-screen you have roughly similar assets/geometry, as usually split-screen is for multiplayer and so you're on the same track or level; this is unusual in that the geometry is mostly the same but they're totally textured and lit differently for each use of the scene.) However, there are parts of the game that literally "blink" in and out of the two worlds, at full frame, in fractions of a second. I'm not a game developer, but I've seen game designers turn on/off graphical elements, and it's not that fast. Turn on/off a grass shader or pop a big tree prop in and out and you'll see the tool take a beat to load the thing, then you'll see it show on screen but maybe with not exactly the right mipmaps needed for that viewpoint and the shadows not fully coming out right on even the baked elements already present, then another beat as the game catches up to the differences in the scene, and then everything's smooth after those first few rough passes tell the system that everything's good to go. Once it's on screen, you can dial up and down the sliders and play with the object/effect and the engine keeps going smoothly, but until the system knows what it's going to be doing with what's in the scene, it has to take a few passes to get everything in sync and in the groove.
That's not what I'm suggesting. You don't have to swap any asset from the environment.

What you do is create TWO separate game maps from the very start, each has their own character and camera. Both variants of the game world exist at all times, in the same 3D space, regardless of what the camera is showing you. Then all you have to do when you change world is to turn off one camera and turn on the other (or move the same camera from one place to the other), which is something you should be able to do pretty much at a sub-frame level.

It's not exactly the same thing I'm describing (there are not two character models, but just because they didn't need them), but you can get the general gist of the idea from here:

 

CamHostage

Member
*Because I don't really want to help keep that argument over DF going, I'm slipping back to an old subtopic...
Yes it’s very similar to the effect they use in rachet

Eh, not really, I wouldn't say?

We don't have a full breakdown of everything Ratchet is or isn't doing (and I haven't seen a breakdown of The Medium's tricks yet either,) but the two games appear to be using different methods (and a variety of different methods) to pull off their realm-switching effects.

(*I am not a game designer, so anybody who challenges my take on this, please knock me down.)

The Medium centers around an idea of parallel worlds that often are literally "parallel". There's two different versions of the world around you, and you see/play them both side-by-side in different split-screen segments, or you glitch between them. (Some of the glitch-teleport effects appear to go to very different places, so I'm not sure the percentage of time the worlds are mirrored versus are totally different, but as I understand it, it's mostly if not always paralleled between the real and spiritual world.)

The common way you would pull this off is to build a stage once, then re-dress it for each realm. You have the same general layout of the game world (same floor, same hallways, same layout) and then add design to the area with texture swaps and lighting changes, according to the mood of each world. One realm is blue and has soft lighting and full of natural materials and fog, one realm is red and high-contrast and the air is filled with billowing particles while all the surfaces are covered in moss or blood or gunk. (You could even add or remove a few props between the two, but to make it easier on the game designer, the "two worlds" probably have all the same geometry in them, and you just turn collision detection off between one or the other, so in one world you see a pile of rubble and can't get past it and in the other world the rubble is invisible and you can walk right through it.)



Swapping stuff within the same general stage layout can be really powerful (your favorite horror games have done it for decades,) be it in either case of The Medium doing split-screen or hot-swapping the full-frame view. Plus, it just makes sense as a game design technique to use the same environment in both realms since you wouldn't want to disorient the player as they move. (You also want to avoid bugs in the game, and if everything is already there in both realms, much of the work's done for you.) You can also do some nice effects like transitions if your transition technique uses a lot of color-shifting and light adjustment and invisibility/visibility effects. You can fade things in or morph some of the geometry, you could even use a a long transition and make it seem like the worlds are collapsing in on one another. Titanfall 2, for example, looks very different when you shift time, but it's certainly the same general environment with different elements, and it might even be that it's the same game map the whole time. Those robots could be there in both worlds, for instance, but in one version their transparency would be 100% and their collision could be off and so they're effectively ghosts until you can see them when you shift. (It's possible that TF2 uses parallel maps instead of altering the map portion you're on, but given that it happens so fast in action sequences at some point of the game and also that it's a lot of open-world spaces without a scripted camera/character model, my guess is that it's swapping look rather than locale.)



The other way of having two places at once is, of course, to have two different places. Build each map or map-chunk totally separate (albeit with similar layouts so the controls match as you move in each), cap each to a reasonable amount of texture/geometry so you can load both as needed, and then have two separate instance running at the same time. Usually one piece is just like a little "room" in your chunks of game map, because you often don't need to be in that area for much longer than a scene or so. (The "tears" in Bioshock Infinite, for example, suddenly show you or bring you somewhere totally different, but you either go back to your regular game world or you come back to an altered version of the world you were just in; I'm pretty sure you never cross maps and play a whole chapter inside a tear.) (It's actually fun in like a Silent Hill to boundary-break the game camera and cross the empty space see the areas that you're not supposed to or to see elements swap while you're not supposed to be looking at them...)



The issue there is, you need both "worlds" to be in memory so that you can switch between them (or you need an incredibly fast data storage system to dump the one you're in and pull the world you want in instantaneously.) Given that The Medium loads worlds in literally a flash (and sometimes blinks between worlds rapidly, in just a couple of frames of time,) it's obviously not "streaming" the two worlds in. Also, The Medium is often split-screen, so while it may mix techniques, it'll generally need both "worlds" in memory because it'll be showing both on the same screen if it splits. Similarly, there are parts where the worlds morph into each other in full-screen view, meaning you have to have both worlds stored so you can intermix them visually. (Aside from needing to stream in two sets of rich data frequently to fill in both worlds as you travel about, I'm actually not sure why we're even talking about SSDs being vital to The Medium's technology?) So you can only build your map chunks to a a certain percentage of detail/scale, because you need to fit both in at the same time in order to use them both. And you can only "stream" in a new chunk that's not in memory if you're either not seeing it or if you're okay with a seam/delay in transition. Cut to black, wooze up the view with dream-type effects, and/or halt play until the new map is ready to use.

You'd also have a real tough time doing some of the cool transitions you could imagine between the two if the scenes were the same, because the two worlds are totally different. If the worlds were the same and just textured/lit/stretched differently, you can blend between them with a slow cross-fade (moving the sliders), you could skew the textures, you could blend transparencies, you could have a prop like a vine "grow" across the surface by cueing an animation once a transition starts, and you can even do some geometry morphing. So even way back on the PS1, a game could have "two worlds" that a shifter could go between at whim, and they were really the same world but by stretching, moving, fading, recoloring, and re-dressing the scene, one world becomes the other right in front of your eyes.

 
Last edited:

CamHostage

Member
...So then, what tricks is Insomniac using for its game?

Ratchet & Clank: Rift Apart, as we know it from the trailer/gameplay demo, appears to have two different types of portals: an Orange Rift and a Purple Rift.

The Orange Rift pulls you through space while inside the same level map. So, that's pretty easy, in that the "two worlds" are the same world. Prey (2006) did this all the time, as did Portal (with some visibility tricks to show you the "other side" of the portal, which I'm surprised the Rift Apart rifts don't?)



We will see how complicated rift-tethering gets (maybe there are a bunch of colors?,) but if the developers wanted to for instance add time instead of just space to the equation, they could use some of the same methods of The Medium or Titanfall, so they'd either do the same color-shifting/light-changing/transparency-substitution to show you the world affected by each side of the Rift, or they'd have a parallel chunk of the world somewhere else in the game map and transport you to that area to give the effect of pulling into a Rift.

Ah, but that's not so crazy; it's the Purple Rifts where things get really crazy. These rifts go from world to world to world, seemingly totally different every time you fall through a different rift. And, from all analysis so far, this game really is loading whole worlds at a time (or at least whole new world chunks, but I'll get to that later.) One second you're chasing Dr. Nefarious through a brightly-lit city after a long battle through the town, the next second you're falling through a purple void, then suddenly you're grinding a rail at night and avoiding trains, then back into the purple space and wham, you're sliding down the side of a building over the edge, then another purple and you're riding a tetradactyl through a desert ravine, then the tetradactyl flies into the purple and you're suddenly swooping over a Fifth Element-type city, and one more purple trip sends you onto a pirate ship for a big battle.



Each area has totally different geometry, lighting, textures, even gameplay, and although a few warped-to sections are barely mini-games as far as play the city level you start at and the pirate level you end up at are full-fledged R&C encounters. (I'm assuming the game not let you at a whim jump off the tetradactyl and play the Fifth Element stage as a complete level, even if a different part of that Fifth Element stage ultimately might be an area you can return to when not in a rift-trip?), I suppose it's possible that the rift-tripping sequences have some tricks to them and there's more of those areas stored in memory than we thing, but for all indications, these are levels streamed from the SSD to the game, one after another.

One of the big tricks, however, is the purple void. You may be in a completely "new world" each time you trip from city to trainyard to canyon to skyway to pirate ship, but that stage needs to dump out while a new stage is coming in. Levels are stitched together with that purple void, which may look like shards of all the levels you could possibly fall into, but are most likely just still/animated graphics of the areas textured onto some floating polygons. Add some purple haze and glow effects and it's a pretty enough space for Ratchet to cartwheel through in the time it takes to dump the old area and load a new area. And given that the gameplay doesn't need an instantaneous or even seamless transition to maintain the play flow, Ratchet can be in that void for as long as the game developers can stand it (it's basically an "in-game loading screen"... this game could conceivably be ported down to PS4 if it were okay for players to be stuck in a 60-second-loadtime void constantly, but that would suck,) so you don't need to perfectly time when one area is in memory or not. So while The Medium can be instantaneous, it can only be the two areas (or however many can fit in memory at appropriate detail) and those things absolutely need to transition smoothly in order to maintain the play flow, whereas Rift Apart can load area after area if it wanted to, as fast as the game can stream them in (or generate them conceivably if it were a procedural game) with a downside being that it takes a second or two or three in say a bland purple void to clear one and load the next one.

I'm not into arguments over whether Rift Apart is "more advanced" than The Medium, so hopefully breaking it down shows why there are different approaches and that there's isn't "one right choice." (If a dummy like me can understand it even slightly, I'm positive almost the techniques used in either game is recognized full well by the other team when they take a look at the other's work, and that they're familiar with if not experienced in testing that method in their office.) I see what's amazing about both... although personally, those streaming Ratchet levels do knock me out each and every time I see it.

That's not what I'm suggesting. You don't have to swap any asset from the environment.

What you do is create TWO separate game maps from the very start, each has their own character and camera...
It's not exactly the same thing I'm describing (there are not two character models, but just because they didn't need them), but you can get the general gist of the idea from here:

Ha, that's funny, you beat me to the Shesez post!
 
Last edited:

gamer82

Member
Oof just watching a demo playthrough and for an exclusive next gen game on the most powerful console its looking really rough

. I get its a small steam but is this best graphics they can get out of the console its looks so dated.
 
Well done. But did you miss the links where this game runs badly on new nvidia 3000 gpus?

So what is it: XsX and all PC configurations are bad or an optimisation problem?

XSX is weak, XSS weaker and PC version has an optimization problem.
 
Last edited:

Radical_3d

Member
...Incidentally, talking about The Medium and realm-switching tricks makes me sad all over again that Soul Reaver: Dead Sun was canceled.

Without Hennig was probably for the best. I’d love if she would retake the franchise, tho. But I’m afraid the IP is long dead and this Z Generation youngsters know nothing about that old blue planeshifter.
 

Durask

Member
Downloaded and tried it.
My gaming monitor now is ultrawide 2560 x 1080 but game does not go ultrawide so it is just plain old 1080p.

RTX2080 Super

Framerates were 90-100 first few minutes then I got bored since this is not my kind of game.
 

Clear

Gold Member
...So then, what tricks is Insomniac using for its game?

Ratchet & Clank: Rift Apart, as we know it from the trailer/gameplay demo, appears to have two different types of portals: an Orange Rift and a Purple Rift.

The Orange Rift pulls you through space while inside the same level map. So, that's pretty easy, in that the "two worlds" are the same world. Prey (2006) did this all the time, as did Portal (with some visibility tricks to show you the "other side" of the portal, which I'm surprised the Rift Apart rifts don't?)



We will see how complicated rift-tethering gets (maybe there are a bunch of colors?,) but if the developers wanted to for instance add time instead of just space to the equation, they could use some of the same methods of The Medium or Titanfall, so they'd either do the same color-shifting/light-changing/transparency-substitution to show you the world affected by each side of the Rift, or they'd have a parallel chunk of the world somewhere else in the game map and transport you to that area to give the effect of pulling into a Rift.

Ah, but that's not so crazy; it's the Purple Rifts where things get really crazy. These rifts go from world to world to world, seemingly totally different every time you fall through a different rift. And, from all analysis so far, this game really is loading whole worlds at a time (or at least whole new world chunks, but I'll get to that later.) One second you're chasing Dr. Nefarious through a brightly-lit city after a long battle through the town, the next second you're falling through a purple void, then suddenly you're grinding a rail at night and avoiding trains, then back into the purple space and wham, you're sliding down the side of a building over the edge, then another purple and you're riding a tetradactyl through a desert ravine, then the tetradactyl flies into the purple and you're suddenly swooping over a Fifth Element-type city, and one more purple trip sends you onto a pirate ship for a big battle.



Each area has totally different geometry, lighting, textures, even gameplay, and although a few warped-to sections are barely mini-games as far as play the city level you start at and the pirate level you end up at are full-fledged R&C encounters. (I'm assuming the game not let you at a whim jump off the tetradactyl and play the Fifth Element stage as a complete level, even if a different part of that Fifth Element stage ultimately might be an area you can return to when not in a rift-trip?), I suppose it's possible that the rift-tripping sequences have some tricks to them and there's more of those areas stored in memory than we thing, but for all indications, these are levels streamed from the SSD to the game, one after another.

One of the big tricks, however, is the purple void. You may be in a completely "new world" each time you trip from city to trainyard to canyon to skyway to pirate ship, but that stage needs to dump out while a new stage is coming in. Levels are stitched together with that purple void, which may look like shards of all the levels you could possibly fall into, but are most likely just still/animated graphics of the areas textured onto some floating polygons. Add some purple haze and glow effects and it's a pretty enough space for Ratchet to cartwheel through in the time it takes to dump the old area and load a new area. And given that the gameplay doesn't need an instantaneous or even seamless transition to maintain the play flow, Ratchet can be in that void for as long as the game developers can stand it (it's basically an "in-game loading screen"... this game could conceivably be ported down to PS4 if it were okay for players to be stuck in a 60-second-loadtime void constantly, but that would suck,) so you don't need to perfectly time when one area is in memory or not. So while The Medium can be instantaneous, it can only be the two areas (or however many can fit in memory at appropriate detail) and those things absolutely need to transition smoothly in order to maintain the play flow, whereas Rift Apart can load area after area if it wanted to, as fast as the game can stream them in (or generate them conceivably if it were a procedural game) with a downside being that it takes a second or two or three in say a bland purple void to clear one and load the next one.

I'm not into arguments over whether Rift Apart is "more advanced" than The Medium, so hopefully breaking it down shows why there are different approaches and that there's isn't "one right choice." (If a dummy like me can understand it even slightly, I'm positive almost the techniques used in either game is recognized full well by the other team when they take a look at the other's work, and that they're familiar with if not experienced in testing that method in their office.) I see what's amazing about both... although personally, those streaming Ratchet levels do knock me out each and every time I see it.


Ha, that's funny, you beat me to the Shesez post!

Rift Apart and The Medium are doing very different things.

The promise of Rift Apart is being able to fully reconfigure the entire rendering and gameplay pipe using data stored anywhere on the SSD in a couple of seconds. The Medium has the ability to run two preset scenes in tandem, and mix/switch between them on the fly.

The difference is that there absolutely have to be two datasets resident in memory for the Medium's needs, meaning that available storage is effectively halved for each view. There's literally no way around this as during split-screen sequences both need to be updated real-time in response to the same user input. The overall frame-time being the sum of the time it takes to rasterize both, hence load fluctuates depending on whether one or both needs to be rendered.

The dynamic loading shown in Rift Apart always has access to the same (full) resource pool both in terms of memory and rasterization budget. Its a pure demonstration of I/O speed that they can basically "fast travel" to any part or level installed on the disk quickly enough not to break the flow of the game. Whats more it can do this without all the data-duplication tricks used by conventional streaming solutions to avoid seek-time delays.

They are really not addressing the same "problem". I/O speed doesn't really have any bearing on what Bloober is doing because all the engine needs to be able to run two displays concurrently, some of the time.
 
Given the fixed camera I just maxed everything and locked to 30. Haven't had any frametime issues that way, but there's enough juice in the GPU tank to try again if it is something to be patched. Not quite a showcase title, but it does look pretty good on occasion
 

RobertsK

Member
Funnily enough, the game is enjoyable.

Potato is delivering.

It is weird, but I have none of these issues. Perhaps, a case of barely noticeable framepacing here and there, but the only time I saw a bad, blurry textures, it was the sky above Niva hotel.
 

Stuart360

Gold Member
To the people saying it doesnt render 2 worlds at once, or because its split screen, it doesnt need twice as much power.
The game does NOT render both scenes consatntly, but power needed in the split screen segments IS doubled.
I'm playing the PC version with Rivatuner OSD and i'm at the first proper segement where both worlds come into play (the abandoned fort in the woods), and full screen single world i was getting 60fps at around 60-65% gpu usage, then when both worlds came into play, and the game goes split screen, i went to 99% gpu usage, and a 38-45fps framerate.
So yeah, the dual world segments basically double the power requirements, and i guess that why XSX halves its resolution in those segments.
 
Last edited:

D.Final

Banned
Given the fixed camera I just maxed everything and locked to 30. Haven't had any frametime issues that way, but there's enough juice in the GPU tank to try again if it is something to be patched. Not quite a showcase title, but it does look pretty good on occasion
Awesome
It will be glorious
 

sinnergy

Member
This game never, ever reaches Quantum Break levels of graphical fidelity. How on earth you can compare both is beyond me.
Because I have both ... that’s why. They clearly are in reach .. and for such a small company it is a great achievement. I never said the surpassed QB, they even do RT.
 
Last edited:

Self

Member
Because I have both ... that’s why. They clearly are in reach .. and for such a small company it is a great achievement. I never said the surpassed QB, they even do RT.

You like the aesthetics, I get it, but don't be silly.



 
Top Bottom