The Tomorrow Children: Closed Beta Technical Analysis
Q-games are not a studio that many may have heard, being based in Japan and with the founder Dylan Cuthbert having roots back to much older generations of games. But most will know of his work including the Snes classic Starfox from Nintendo and maybe even more famously the Bath Duck tech demo for the PS2.
But he has always pushed technical boundaries in whatever projects he has been involved in.With his own company long since established and they have been busy working on a PS4 exclusive that you may have heard me talk about many times before, now even at first glance it may look, to some, to not be pushing the delicate balance of software technology as far as previous works but nothing could be further from the truth, but I race ahead and first I want to talk about the game and just what this soon to be a classic game is delivering and changing.
The story centres around humanities need to create, meddle and ultimately destroy, following a failed experiment in the 60’s that has reduced human consciousness into a place ominously called the Void but hope is not lost as you are the hero this world needs, not to mention young couples everywhere to rebuild it all. As the world has been created from the depths of the combined human psyche expect it to be it a bit of a nightmare.
But how do you go about this Lego related harvesting saviour & construction you may ask? Well I have the answers or at least some after the closed Beta launched this month which some of the footage you can see in the video. With the game running within a persistent online world you have to mine the land, build, gather and create for all your worth. You are humans last hope as they have created these ‘projection clones’ or Avatar to mine the land and this can be chosen from classes such as Radio, Engineers, miners but your work is not just simple digging. As your town and settlement will come under attack from Godzilla wannabe’s in the Izverg monsters (think Kaiju) you must connect and work as a whole to save and defend your town’s folk. As you play you will see other players fade In and out that are doing the same as you and they help you as much as you do them, this combined ecology is a vital part of the game. As you are hunting for human DNA that is scattered across the land in the Matryoshka dolls.
Now this persistent world, online play, scavenging mixes both synchronous and asynchronous structures into its gameplay, but more impressively it is also one of the growing number of games that also uses both within its GPU resource control. To explain in brief this means that with the standard Direct X path of rendering work passing pixel creation work over to the GPU that it will then allocate to the appropriate AL Units as is most efficient. Both consoles and AMD GCN cards being endowed with a more coherent design that better utilises the constant wasted cycles or bubbles within a GPU, helping it achieve more of its maximum potential whilst not reducing its current scheduling. The key here is the ACE in the PS4 allows up to 64 jobs to sit in the queue and be reallocated per cycle, enhanced further with its UMA memory system allowing the GPU to be used for Non-rendering tasks that would normally be CPU bound. The team whilst not being part of Sony’s worldwide studios has closed links with them and Nintendo from Dylan’s long established programming history. And they have strived forward as is typical of them with embracing new technology and using it to empower greater games, being able to use the GPU resource for non-Rendering tasks was always a key part of Mark Cerny’s foundation from the start and will help keep the PS4 (and to a lesser degree the XboxOne) in the game for longer. But like all new methods and tech this will take time for teams to get their head, engine and skills around but be sure it has been used in released games and this will only increase and improve as we move forward. Q-games have been very open with this even sharing a clear example of the improvements it has made to their game in its presentations which are linked here and this article on our site. From the shared information at the time they managed to save up to 30% of rendering time in heavier geometry scenes and could not promote enough to all other developers to look at using the Async compute function of all AMD GCN cards and the results can be clearly see in the on screen results.
All this means nothing if we cannot see or explain just where this resource is being used but I can help enlighten you on that. The games visuals are both striking in design and production, with the carved look to each character and world mixed with a game that looks very close to its Pre-rendered target that it could very well be a Tim Burton movie. One of the biggest reasons for this is the games lighting system which as it stands is one of the most accurate and technically advanced one in a production game, to be clear all solutions are far from a Ray tracing one but work in a similar method. Global illumination is mentioned much now in games and has been used since the last generation but most of the solutions use Pre-baked methods that are rendered off-line or a single or maybe 2 bounce system that has a dynamic element or a combination of both. But TTC uses its own GI panacea created using Cascaded Voxel Cone tracing which is a form of Ray tracing in that it will evaluate space or volume within each scene, samples are taken at intersection points and as it uses Voxels throughout the scene it would use a large portion of memory, reduced by storing the information of fixed geometry within the scene and when these are found extending the cone/cascade until it is all sampled delivering a real-time solution. Without digressing too much here into the technology you can read the presentation from the team yourselves as it is both interesting and very intelligent. Helped by its interleaved updated cascade function being Z based so that the further away from the camera the geometry the update could be every second, 4th and so on frames. I may cover in more detail within another article if needed but I do not want to dwell long here. The long and short is that data is stored from samples into a 3D texture a method allowing the team to deliver a system that calculates the irradiance and radiosity of up to 3 bounces of light. This gives the mesmerising results you see on screen that combined with the design and Post effects could fool many into thinking this was a stop motion movie. The solution allows multiple light bounces that give in-direct shadows and light in conjunction with colour bleed within the bounce that carries over the colour from the reflecting material back onto the recipient surface and this is a truly dynamic solution and a damn impressive one at that. It's is limited to fixed scene geometry only but the team have aims further along with shadow improvements a method of screen space sampling could be used in a later title to allow dynamic objects to also become a reflective source.
Not wasting the techniques they have created for the rendering process they also abandoned SSR methods for reflections instead with the Voxel data they used Ray Marching for reflections of all static and dynamic objects. This solves all the occlusion and fading out issues of a screen space solution. Allowing the reflections to show even when not within the camera view and of an equally high resolution. This method also allows light to refract and reflect along with the world in an incredibly accurate and convincing manor that delivers and leaves a lasting impression.
Injecting even more flair into this is the Non-Photorealistic style of the game allowing its conviction to be more authoritative in action. Colour grading is used to enhance the feel and hue of scenes right down to individual pixels which is used very well for depth shots or cavern scenes. A great motion blur is also fully up to speed and combined with its solid performance helps enhance the look, feel and smoothness of the overall presentation. Geometry is also created in a new way or at least handled with the landscape being excavated by yours and others own fair hands means that it needs to be destructible which is achieved by using Layered Depth Cubes which again is volume based within the Voxel lighting system and the team can use simple flags (Boolean) to remove or add sections with them being drawn as polygons when the player gets up close and personal. As you dig through the scenery a simple Normal map decal is added and expanded to show the impending destruction. This entwined process results in the lighting still being calculated around, over and through the objects again dynamically within play as you extend your reach. It cannot be stressed just how accurate and convincing it all is as you really get to see light react and bounce in multiple direction across and under surfaces that you forget all the hard mathematics going underneath to deliver such wonderfully lit beauty.
Some of the Post effects and lighting could be toned down slightly with the Bloom being one area along with the Vignette being relatively intrusive during the darker moments which does add to the disturbed and oppressed feeling of the game, but an option to turn these down or off I feel would be a good addition.
Even though the Beta was very short and I got no-where near enough time to enjoy it or analyse it fully I am more than happy to put by wallet forward when the final games ships planned sometime later this year. With my short stay in the void enough to wet my appetite to play this refreshingly different and hauntingly beautiful game.