Server based game augmentations. The transition to cloud. Really possible?

[ThatGamer]

Diablo 3 needs the internet for DRM purposes. Its is not an example of server based augmentation. The game is run and processed on your hardware locally.

I will take your word.

You didnt understand the subject of the discussion entirely hence why you bring up online gaming and unrelated examples like Diablo 3.
Cloud based game augmentation is something different. A portion of your game is processed on your local hardware and another portion is processed from an external network source

That is exactly what happens when you play an MMO. That is why numerous people brought up MMOs/persistent worlds as the obvious example:

http://forum.beyond3d.com/showpost.php?p=1738426&postcount=21
http://forum.beyond3d.com/showpost.php?p=1739634&postcount=106
http://forum.beyond3d.com/showpost.php?p=1739105&postcount=55
http://forum.beyond3d.com/showpost.php?p=1738449&postcount=24
http://forum.beyond3d.com/showpost.php?p=1740574&postcount=220
http://forum.beyond3d.com/showpost.php?p=1740245&postcount=181
http://forum.beyond3d.com/showpost.php?p=1740117&postcount=158

These are just some of the posts. You can read through the first couple pages if you would like. If anything that is something most posters have agreed is an obvious example. In fact even MS have mentioned it as a possible use. Most people are debating how effective it can be at improving IQ. That isn't the only practical application though.

My point is simply that as most have agreed that MMO's/persistent worlds are the obvious use case. We can also look to real implementations and see how they have dealt with internet problems. Latency spikes and internet drops aren't new problems. Many games have been large successes despite them.

My conclusions are related to the current situation only and not the future situation which, like you, I also expect to change and improve and hence support cloud based game augmentations it is meant to be

There is data available to help project the future. If we limit ourselves to the current average internet bandwidth, which pre-dates the release, then these conclusion could be pretty much useless for years: 3,4,5,6,7,etc. of the console's life cycle. We will be having this discussion in another couple years.

 

[Shifty Geezer]

I think your misunderstanding what was said.

Sorry, I probably should have said, 'do not run deterministically.' Newtonian physics ought to be predictable, but set up the same stack of colliding cubes in any physics engine and run it from the same starting state, you get different results. A simple modern example of this is LBP. Create a physics setup in that and it can have different results. I was a long time ago when I was programming with physics engines though, so I don't know how things have improved, if they have. I would expect that chucking a grenade at a wall in BF will result in different arrangements of rubble though when calculated on different machines, and syncing that rubble over the internet will be hard.

 

[Nesh]

I will take your word.

That is exactly what happens when you play an MMO. That is why numerous people brought up MMOs/persistent worlds as the obvious example:

http://forum.beyond3d.com/showpost.php?p=1738426&postcount=21
http://forum.beyond3d.com/showpost.php?p=1739634&postcount=106
http://forum.beyond3d.com/showpost.php?p=1739105&postcount=55
http://forum.beyond3d.com/showpost.php?p=1738449&postcount=24
http://forum.beyond3d.com/showpost.php?p=1740574&postcount=220
http://forum.beyond3d.com/showpost.php?p=1740245&postcount=181
http://forum.beyond3d.com/showpost.php?p=1740117&postcount=158

These are just some of the posts. You can read through the first couple pages if you would like. If anything that is something most posters have agreed is an obvious example. In fact even MS have mentioned it as a possible use. Most people are debating how effective it can be at improving IQ. That isn't the only practical application though.

My point is simply that as most have agreed that MMO's/persistent worlds are the obvious use case. We can also look to real implementations and see how they have dealt with internet problems. Latency spikes and internet drops aren't new problems. Many games have been large successes despite them.

The other posts isolate the discussion to cloud based augmentations to specific game scenarios and types of games such as MMO. And they are specific for the same reasons I pointed you earlier. They are food for thought about where and how.

On the other hand you werent that specific and used MMO's as a broader evidence that people are ready to accept the inefficiencies in their general gaming experience right now.

In addition you were talking about CURRENT MMO's which do not use cloud based game augmentations whereas these people were talking about how FUTURE MMO's can apply them.

If you were more specific we wouldnt be having this discussion

There is data available to help project the future. If we limit ourselves to the current average internet bandwidth, which pre-dates the release, then these conclusion could be pretty much useless for years: 3,4,5,6,7,etc. of the console's life cycle. We will be having this discussion in another couple years.

As I said, I also expect it to be more relevant in the future.

 

[ManuVlad3.0]

Shots fired:

There was a report last week where Microsoft claimed that the Xbox One will become more powerful with the use of cloud services. Blow thinks this is bullshit and in a series of tweets he criticizes Microsoft’s business strategy for the Xbox One.

In another tweet he questions the business strategy for Xbox One and he does not like when companies lie contemptuously.

He further claims that each of the servers are actually virtual and he himself can can make 10,000 servers per host. Apparently, he thinks this is all a lie. (For your information: In Virtual Servers, multiple customers are hosted together on a single server and each of them share the total resources. This is unlike the dedicated server where each customer is allocated a set number of resources which is normally more compared to a Virtual server).

However, there is no confirmation on whether Microsoft will be using virtual servers, so we have to go by what Blow says above.

He does make sense in many ways though. What if your internet connection goes off and you are disconnected to the cloud? Will the Xbox One be able to render advance physics or other complex simulations in such a scenario? Many such confusing questions remain. Let’s hope that Microsoft clears the dust at E3.

Stolen from NeoGaf

 

[BRiT]

Shots fired:

Stolen from NeoGaf

Not on topic at all, since his main beef is doubting the number of servers that MS will make available on the service for Xbox One. I'll repost my reply in all the other threads that it's been dumped into:

Damn, he seems so worked up, as if MS personally killed his entire family. Personally, I learned to ignore anyone who is that worked up over anything as they are not level headed. Their own personal agenda gets in the way of them seeing the situation clearly and with an open mind.

 

[warb]

Shots fired:



Stolen from NeoGaf

But they also claimed around 4x Xbox One for each system in the cloud. They can't rely on sucky virtual servers for that. And I believe the emphasis was on 300k worldwide to minimize latency.

 

[fehu]

Probably this has been picked in the previous pages, but will developer have to pay to use the servers?
If so, the majority probably will opt to port as best as they can from the ps4, and all this enhanced server processing will be forgotten in some year as all the others pr stunts... :/

 

[Jedi2016]

Are we fairly certain that these "augmentations" are required, or optional? Will the games not work at all if the servers aren't there, or will they simply not work as well? I hope like hell they make it optional, for the people that don't have the ability to constantly run everything online.

 

[Scott_Arm]

Let's throw out something.

There is a broad spectrum between what Gaikai do and running everything locally.
For example I could theoretically run my entire game on the server minus the graphics, and instead of streaming video, I stream object state.
That seems no less feasible to me than the Gaikai model. Unless you think streaming video is lot less data that streaming game state.
Since all the physics/AI etc is in the cloud you can use all the local resources for making it look prettier.

I have perhaps 40K a frame at 30fps on a 10Mbit link, that's a lot of state and the only state I need is the stuff immediately around me. Most online games manage with less that 1/10th of that.

And there are is again a lot of spectrum between that and running locally.

Data caps obviously become a huge issue as they do with Gaikai.

It's certainly interesting, and to be honest I haven to applaud MS for putting the infrastructure in place. Whether it gets significant use in the next few years remains to be seen.

Where is the line drawn between physics and graphics? I'm assuming the result of rigid body or fluid simulation is a mesh, as Shifty suggested. Are there types of physics that are too closely coupled to graphics to make them good targets for cloud computation? Are there any types of graphic calculations that might be an exception and be possible cloud targets? The arstechnica interview mentioned fog specifically, some precomputed lighting data, and someone in this thread mentioned fire.

Thank you for participating in this thread and bringing in some technical perspective.

 

[Shifty Geezer]

Are we fairly certain that these "augmentations" are required, or optional? Will the games not work at all if the servers aren't there, or will they simply not work as well? I hope like hell they make it optional, for the people that don't have the ability to constantly run everything online.

For the purposes of this thread it doesn't matter whether server-side game improvements materialise or not - it's just a discussion on whether they can happen and how.

 

[ERP]

Where is the line drawn between physics and graphics? I'm assuming the result of rigid body or fluid simulation is a mesh, as Shifty suggested. Are there types of physics that are too closely coupled to graphics to make them good targets for cloud computation? Are there any types of graphic calculations that might be an exception and be possible cloud targets? The arstechnica interview mentioned fog specifically, some precomputed lighting data, and someone in this thread mentioned fire.

Thank you for participating in this thread and bringing in some technical perspective.

The bulk of the result of rigid body physics is going to be a collection of positions, orientations and various derivatives there of.
Fluid dynamics doesn't seem like a great fit to me today, most of the demos I've seen are height fields or particle systems, and that seems like a lot of data to push over a network connection. Though it may be possible to do some sort of low resolution computation on the cloud that can refined locally, or the data may compress extremely well, hard to say.
Things that remain largely static once computed like the fog example, don't seem like too much of a stretch.

It will be interesting to see how if at all the cloud compute is used in the short term. I think there is a good chance it will be completely ignored.

And of course there is still the question of how it's paid for.

 

[Cjail]

Probably this has been picked in the previous pages, but will developer have to pay to use the servers?
If so, the majority probably will opt to port as best as they can from the ps4, and all this enhanced server processing will be forgotten in some year as all the others pr stunts... :/

They didn't say it but one way or another costs will go up for developers and publishers.
MS might decide to cover the part of the costs for the servers but I doubt they will cover the development costs even in part, first party exclusives aside.

 

[astrograd]

Where is the line drawn between physics and graphics? I'm assuming the result of rigid body or fluid simulation is a mesh, as Shifty suggested. Are there types of physics that are too closely coupled to graphics to make them good targets for cloud computation? Are there any types of graphic calculations that might be an exception and be possible cloud targets? The arstechnica interview mentioned fog specifically, some precomputed lighting data, and someone in this thread mentioned fire.

Many next gen water simulations seem to be tessellation-based. On the other hand, that type of thing would also already have baked into it a gradual sort of LOD system to handle the tessellation so ocean waves far away from the player could still see hugely complex physics computations governing them I'd presume.

The bulk of the result of rigid body physics is going to be a collection of positions, orientations and various derivatives there of.
Fluid dynamics doesn't seem like a great fit to me today, most of the demos I've seen are height fields or particle systems, and that seems like a lot of data to push over a network connection. Though it may be possible to do some sort of low resolution computation on the cloud that can refined locally, or the data may compress extremely well, hard to say.
Things that remain largely static once computed like the fog example, don't seem like too much of a stretch.

It presumably shouldn't need to be literally static. Just non-interactive in the sense it can't be updated within a frame or two. If they are streaming in scripted animations (based on pre-computed physics interactions) the assets can still be animated obviously.

On the physics side specifically I am personally more concerned with how assets are created to utilize things like destructible environments or deformable terrain etc. For example, say there is an avalanche in a game and at the base of the mountain trees are uprooted or knocked down. How does the cloud deal with the roots/debris/general destruction? Either some artist has to model it and include that imagery on the BR disc or somehow it must be generated procedurally.

Seems to be there could also be room for relatively pseudo-dynamic stuff too. For instance, in the example I just gave say the player were to trigger the avalanche. That's a large scale physics effect that the player wouldn't be able to influence considerably most likely. So when the player gets within some range the computation can be done in the cloud and sent to the console which waits for the trigger from the player.

 

[ninzel]

Even though I'm skeptical of cloud to greatly enhance the power of the Xbox1 I find it interesting to think of how devs might work through the process of implementing this stuff.
Forgive me if this stuff has been discussed.
So if you are dev how do you work through the process.
I guess at first thing you would have to look at best and worst case scenarios?
Best case would be a gamer that has a fast low latency always always on connection.
Worst case scenario would be a gamer that meets just the most basic of connection requirements.
Worst case scenario I guess would be what you call the Standard game that ships on the disk.Best case scenario would be an enhanced version.
So where do you go from there.
Do you build multiple version or states of the game,something similar to lowest and highest system requirements on a PC game? And if you can ship that on disk,why do you need the cloud.
So lets say we look at something specific like an animation system. Do you ship on disk a scalable animation system that activates or engages depending on what the system check tells it? Low connection equals low animation,high connection requires better animation? And again if you ship the scalable animation system on disk why do you need cloud. How does the cloud "turbocharge" what on the disk?
Sorry guys I'm just trying as a luddite to wrap my head around what the cloud could provide over the net given ideal conditions that makes more sense than just shipping on disk.
Going back to the animation system,instead of shipping one scalable animation system,do you ship one basic animation system,then a separate cloud animation system let can potentially run in parallel under ideal conditions.
So I guess in order to run a game at times calculations have to be done by hardware. Is this what the cloud provides the game engine,the ability to feed finished results back to the game engine that could not have simply been done by the game or hardware in the first place?
At and at some point the two separate game states have to be seamlessly blended.
Crazy complicated stuff as if developing a regular game wasn't hard enough.

 

[Cyan]

Legendary RPG developer Obsidian says that the company has cloud based ideas for the Xbox One.

I wonder if what bkilian said about creating more believable, richer worlds and AI would become a reality in this case.

Since those technical aspects of a game aren't latency sensitive, using Skyrim as an example of a game which would greatly benefit from the use of the cloud, leads me to believe that the whole RPG genre could benefit from it too.

http://news.softpedia.com/news/Obsidian-Has-Cloud-Based-Ideas-for-the-Xbox-One-356259.shtml

I am not totally sold on this cloud thing though -in fact, I couldn't care less-, but it's interesting news to RPG fans in any case.

 

[Lalaland]

AI and richer modes of interaction in RPGs are about the only thing that I can see benefiting from the cloud. I don't mean tactical AI in combat but rather large scale faction tracking or allowing large numbers of characters to have individual like/dislike tracking (ie if you piss off Faction A some members whom you have helped would still be onside). Still the examples he cites are quite MMOish rather than being SP focused

 

[Scott_Arm]

There is also the possibility of player analysis on a group level to alter the game. Maybe a game that can algorithmically tune its own gameplay or difficulty. Maybe have an economy that changes the values of goods based on how all players spend their money even though the game is single player. There is also the ability to record player data that can be crunched in the cloud to provide devs with better ways to make better game patches, rather than relying on the small sample that provides feedback on forums. Some of that is not directly augmentation, but I believe player analysis can be of indirect or direct benefit to gamers. Group learning AI would be cool.

As a concept, imagine a game that constantly collects and analyzes data to dynamically create new difficulty levels. Replayability could be taken a long way. You could also make better recommendations to players as to what difficulty they should play on by putting them through a tutorial and comparing to the whole population.

This post was inspired by the player analysis that's being done in the back end for Halo 4 for a number of things including player banning.

Hell, if they want to do skill matching they could start doing WOWY (with or without you) and quality of competition comparisons like the guys who have started doing advanced stats analysis for the NHL.

Edit: Here's a Peter Molyneaux like concept. A 2 hour game with no saves or check points that is meant to be played in one sitting like a thrilling movie. By collecting player data and biometric feedback (heart rate) from the entire player population the game can adjust and provide you with an exciting and challenging playthrough that scales each time you play and improve. Dieing destroys the narrative of cinematic games. This concept could get rid of that.

 

[Scott_Arm]

The bulk of the result of rigid body physics is going to be a collection of positions, orientations and various derivatives there of.
Fluid dynamics doesn't seem like a great fit to me today, most of the demos I've seen are height fields or particle systems, and that seems like a lot of data to push over a network connection. Though it may be possible to do some sort of low resolution computation on the cloud that can refined locally, or the data may compress extremely well, hard to say.
Things that remain largely static once computed like the fog example, don't seem like too much of a stretch.

It will be interesting to see how if at all the cloud compute is used in the short term. I think there is a good chance it will be completely ignored.

And of course there is still the question of how it's paid for.

Thanks for for the reply. I guess we'll have to wait and see if anything shows up at E3. My guess is Fable or Forza being good candidates for cloud talk.

 

[Strange]

Diablo 3 needs the internet for DRM purposes. Its is not an example of server based augmentation. The game is run and processed on your hardware locally.

Not exactly true. Many players (especially wizards) know that damage output is heavily ping dependent on certain builds. The better your connection to the server is, the better the build is. This wouldn't be the case if the game is entirely run and processed on our hardware locally.

 

[KKRT]

Not exactly true. Many players (especially wizards) know that damage output is heavily ping dependent on certain builds. The better your connection to the server is, the better the build is. This wouldn't be the case if the game is entirely run and processed on our hardware locally.

Its because of sync issues, not because of needed calculations in cloud.