What you need to know
- Microsoft officially discussed its Xbox Game Studios Publishing cloud gaming division at GDC 2022.
- This division of the Xbox publishing group is working on cloud-native games.
- Kim Swift, head of the cloud gaming division, discussed the kinds of challenges the team is looking to tackle.
Microsoft officially talked about its Xbox Game Studios Publishing Cloud gaming division at GDC 2022. This division is working to create "cloud-native" games for the Xbox ecosystem.
In a presentation (as shared by VGC), senior director of Cloud Gaming Kim Swift talked about the process for developers working with the publishing organization. Unlike streaming a game through Xbox Game Pass Ultimate, a cloud-native game would theoretically be processed almost entirely server-side, before then being streamed to a player's device.
Swift is a veteran game designer who has worked project lead on Portal and designer on Left 4 Dead at Valve before also working as a game design director at the now-defunct Stadia Games and Entertainment. Swift was hired by Xbox Game Studios Publishing back in June 2021.
"We believe that cloud gaming has a huge potential upside for reach and accessibility," Swift explained, while adding that the team is looking to publish cloud-native games.
Swift also notes that cloud gaming is still relatively early on as a technology, and further areas of research not yet figured out include better graphics, more complex or numerous NPCs or even more players that possible on native hardware.
According to reports from last year, Xbox is working on deal to publish one of the next games from Hideo Kojima and Kojima Productions, though we'll have to wait and see if anything is officially announced.
Best of all worlds
Xbox Game Pass and Xbox Live Gold in one
Xbox Game Pass gives you access to over 200 games for one monthly fee. Xbox Game Pass Ultimate also adds Xbox Live Gold to the package so you can play online with your friends. You also get access to EA Play.
On the graphics front, one concept that has intrigued me with cloud is an entirely different way for calculating the scene that is then displayed to a player in 3D games. On local gaming, there are many tricks and shortcuts taken to improve performance around only rendering what the player sees (to avoid wasting processing power to render something outside the field of view or behind a barrier). But if you have a large multiplayer game, different players see different views, so those hidden areas may need to be rendered anyway. If you had a powerful enough super system, you could render everything, then just give each player a view of the relevant scene. With enough players, that may be more efficient than rendering thousands of different views independently. This is a graphic option that only becomes available with a very, very powerful central system and effectively dumb terminals just displaying the graphic feed. In this scenario, using true ray tracing actually REDUCES burden on the system, because that becomes the only way to properly light a scene for multiple users from different player/viewer angles. Fascinating thinking about the programming opportunities and challenges this creates. Will this happen in the near term? I suspect not in the first generation of games, but maybe it's coming in the next few years. Or, maybe the processing needs for that central supercomputer are still so high that for a long time it will still be easier to render a separate scene for each player using conventional tricks and approaches, but these are the kinds of things that true multiplayer cloud gaming enables.
That's not how rendering works, at all. Constructing a 3D space with multiple people in it works today on any device capable of 3D math, be it a PC, a phone or a toaster oven. Rendering is when you need that representation to appear on a display device, usually in 2D based on that 3D data. When 3D game engines don't render obscured content or content outside the camera's viewport (the latter is called frustum culling), that doesn't mean the content outside the viewport "disappears". Cloud compute has nothing to do with what you described. Also, when you talk about a supercomputer, that goes directly opposite to where cloud computing is today, which is increased scalability often made possible by making systems more distributed.
Well, server farms are each effectively a supercomputer. 😇
Gabe, I don't disagree with your factual points, but you're not looking at the big picture and the way those very algorithms change over time. Your points are why I acknowledge that we probably won't see anything like this in the first iteration of this -- it's just too big a change not just in development and tools and processes, but in the underlying algorithms used and how they are used. I'll give just two examples to illustrate: 1. Colored lighting and lighting levels are typically grafted into a scene by the game artists. They are not usually done by truly calculating light and color from light sources in a dynamic way (games with true ray tracing being an exception). This is because the computational power needed to do this in real time is substantial and the fake models game developers use work really, really well to create an illusion of actual lighting. Gamers with top-of-the-line graphics cards could turn on ray tracing and get this effect dynamically, but the lion's share of users with their older, less powerful cards would get nothing, so the fake approaches are needed. However, if you have a supercomputer (many processors running in parallel is still a supercomputer), it can calculate all of this in real time with ease even for the user with the least powerful hardware, because the streaming video is almost purely a function of the resolution, with scene complexity being irrelevant for streaming (I say "almost," because compression algorithms are impacted by scene complexity). Furthermore, the supercomputer running this only has to calculate the lighting once, not once per user, as in the case of traditional game rendering engine. This doesn't help with object reflections, which must still be calculated per user (reflections are angle-based, so every user's point of view creates a new reflection calculation need), but ambient lighting can be completely covered with a calculate once-and-done model for the entire world, regardless of the number of players. 2. Number of triangles composing the world, which affects things like how many blades of grass, how many independent tree branches, how much detail in crowds of NPC's, etc. Same as #1, calculating this per user is intensive and exactly multiplied by the number of players -- 100 players takes 100 times as much computational power as 1 player -- but that's entirely different when the scene itself is being defined once by the central computer. Traditional MMO's just use a few data points to show where other characters are (x, y, z, angle facing, and some data for position and actions -- the idea is to keep this data to a minimum specifically because that keeps everything responsive for remote players) and then leave all of the scene construction and then rendering to the local PC for each player. That whole approach is turned on its head when a central system is determining where every triangle is once. Lastly, both of those are processes that are, even with current rendering systems, entirely parallelizable. Perfect for cloud computing with something like thousands of of RTX 3090s all running a single model. This is how current graphics cards already handle these things.
While this sounds good in concept, lets not forget that processing games online was basically just slap in a xbox series x on the server and call it day. At what point do we seriously NEED cloud only games? processing the game almost entirely server side? what does that even mean? We already have it processed and then stream to a device.
Yeah, if that's all they do -- provide every user with an Xbox Series X in the cloud -- then this is much less interesting. I hope by the fact that MS is hiring for this that they're looking at a completely new and different computational model where the central system running the game is instead thousands of Xbox Series X's running as a single supercomputer. That would benefit from a radical re-thinking of world-building and graphical design for games and enable some incredible transformations.
Bring xcloud to every country and we'll start talking.
It is pointless until I see a concrete plan to show me exactly how much "free" processing power is allocated to each xCloud members. Because currently, it is not free, thus, games like Crackdown 3 don't want to pay for it and make the cloud physics as ****** as possible.
I find the timing of this announcement amusing, coming just before the FTC starts evaluating the Microsoft purchase of Activision. Most of their concerns (aside from big is bad) is that MS might make games XBOX and PC exclusive. Well, cloud native games are user-hardware agnostic. Kinda puts that one to bed.
Interesting. Clever insight. I wonder if that was part of the motivation. Hadn't occurred to me.
GHz and Gpu.
Thank you for signing up to Windows Central. You will receive a verification email shortly.
There was a problem. Please refresh the page and try again.