Support NeoGAF

[jetjevons]

How about some of those sexy, gratuitous physics (physx) particle swirls or destruction orgies you see on high end NVIDIA cards? Could any of those details be calculated remotely?

 

[Zaptruder]

So if GTA5 had this cloud computing feature, you'd be able to have more pedestrians in the city around you all of whom have individualized AIs?

This might be a possibility.

But your console doesn't crunch the AI - it gets sent the pathing information, kinda like animation data.

So on your console, the AIs are following predetermined paths and behaviours, and only when you bump into and interact with them would the local console break them out of those behaviourial loops and interject its own in there.

Of course, the scene would have to be carefully constructed - you couldn't have too many people too close together at once (which kinda misses the point of cloud compute AI?) - in case the player does something that affects all of them at once (i.e. pulls out a gun in the middle of night club and shoots - you'd expect everyone in the scene to start panicking and taking cover).

Maybe you can have more complex nodes of behaviour for cloud computed AI, but when they're locally controlled for such a large group, they become 'dumb' - i.e. panicking and all running away from the player, before cloud AI kicks in and individuals start to seek for cover and pull out their cell phones, or etc.

 

[oVerde]

How about some of those sexy, gratuitous physics (physx) particle swirls or destruction orgies you see on high end NVIDIA cards? Could any of those details be calculated remotely?

Sure, as long the "impact" can be predicted before, so the calculation comes in time.
Theoretically works, but you won't see player directly generating those physx particles animations if are timed sensitive and can't lag a bit.

 

[Demoncarnotaur]

Lets just pretend they are being honest and can get this to work-

Would only online only games work with this? Or could they perhaps scale the visuals (better visuals when online and speaking to the cloud)?

 

[Zaptruder]

It is just not possible(at this time) or at the very least practical. What the fuck are some of you smoking?

Care to elucidate more on the technical limitations in game engines and cloud computing?

It seems to me that it would require a new way of thinking about structuring your games, and even a doubling of work in a lot of areas (i.e. have a cloud compute version of this function/asset, and a local version/function). But it doesn't seem like an intractable issue to me.

Just less of a selling point up front than MS hopes. Eventually, I think this will be the obvious direction of computing in general.

Lets just pretend they are being honest and can get this to work-

Would only online only games work with this? Or could they perhaps scale the visuals (better visuals when online and speaking to the cloud)?

The latter is a possibility.

 

[fin]

If MS can produce a demo with and without cloud computing, that should work. Also considering Epic had to scale back the UE4 demo on PS4 (I think it was the global illumination?), I'd be interested in what they think of the cloud...probably find out at E3.

 

[bes.gen]

welcome to digital foundary heaven

xbox+cloud vs xbox vs ps4 vs pc

 

[IHaveCandy]

Dynamic lighting does need fast and immediate responses. This doesn't make sense.

 

[Zaptruder]

Dynamic lighting does need fast and immediate responses. This doesn't make sense.

Not all lighting is created equal.

Head lights and torchlights and explosions != day/night/weather cycles.

 

[sun-drop]

so did this work back when the cell did it? .... right ..


no one is going to waste the programming resources to code for this, hell lowest common denominator porting as it is is rife in the industry ..and that's coding to hardware platforms where specs are set and don't move.

rule of thumb, if yr console is launching in a few mths ..and nothing specific is detailed about things like this ...it's FUD.

great counter material to when ever any one asks why ps3 hw is more powerful though .... but something tells me MS sound have dedicated more of this time brainstorming effective responces to press questions about the DRM situation ... amiright

 

[wildfire]

Alright MS. Show this at E3 and tell us the price and then we'll know if you are offering something of value.

 

[kitch9]

They haven't demonstrated a game yet period...

Let MSFT prove it at E3 and beyond. If they don't then you call them out and gloat. If they do, you eat crow (with about half the other GAF members). The over reaction on the forum with the current (non) information is honestly crazy.

I've used Skydrive....

How they think they can get this to work when they can't get that to work right is beyond me.

 

[SeaOfMadness]

I realize this is new for many people, but it's not new in the computing field and it's not even new for games (as I've said: FPS games work exactly like this on a much smaller scale already)

It's fine to be skeptical, but I think people should be open minded about it.

The key to all of this is a commitment to providing the infrastructure for every game to have access to this service which is unprecedented. Many games already use their own servers for all kinds of stuff, but they're expensive to run and maintain and don't scale very well. This seems like it will address those issues and it's really just up to the creative people to figure out how best to take advantage of that.

I don't think this should be dismissed just because it's not well understood.

And it's being picked up by mainstream business press as well

 

[zomgbbqftw]

So this really is just turning into Microsoft's "power of the Cell" then. Everything is left unexplained and vague and executives just keep saying "power of the cloud" or "cloud enhancements".

I think the difference is that Sony just about managed to get away with it because their team of first party developers are perhaps the most gifted in the whole industry so they managed to extract every last drop of power from the Cell and PS3 first party games were appreciably better looking than other games on either console. I'm not sure I can picture first party Xbone games looking better than first party PS4 games because of some undefined Cloud power.

 

[jetjevons]

Sure, as long the "impact" can be predicted before, so the calculation comes in time.
Theoretically works, but you won't see player directly generating those physx particles animations if are timed sensitive and can't lag a bit.

In most cases I think they're just eye candy.

 

[Xplatformer]

I think its a really interesting idea with potential down the road, but I also see it effectively killing offline gaming. Not cool.

The first quote in the OP speaks to having some way to intelligently handle situations where you lose your connection, but doesn't offer any details. The second quote is the kicker though, since he speaks directly towards moving more and more processing to the cloud over time to actually in-place upgrade the graphical capabilities of the console. That to me means they want and expect offline gaming to deteriorate in quality and effects over the lifetime of the console.

An always online console is fine, provided you can guarantee uptime and quality to the extent that say, electricity is dependable, but you can't. Always online is also a measure of control, an ad delivery mechanism, a possible invasion of privacy via kinect, DRM, and one way to justify requiring a subscription account.

So for right now I see it far more about control than about possibility.

Office 365...for games.

Microsoft wants to control ALL content from the cloud. And we thought they were greedy when it came to the desktop OS.

Again, why can't this be done on the PS4 as well, with the majority of the graphics for multiplatform titles still looking slightly better than they do on XBox One'

 

[SeaOfMadness]

So this really is just turning into Microsoft's "power of the Cell" then. Everything is left unexplained and vague and executives just keep saying "power of the cloud" or "cloud enhancements".

I think the difference is that Sony just about managed to get away with it because their team of first party developers are perhaps the most gifted in the whole industry so they managed to extract every last drop of power from the Cell and PS3 first party games were appreciably better looking than other games on either console. I'm not sure I can picture first party Xbone games looking better than first party PS4 games because of some undefined Cloud power.

Personally, I think the "looking better" aspect of this isn't the interesting part. It's more about enhancing AI, turn based games, massive scale MP games, stuff that can be done in the background and used later, not stuff that needs to be sent down immediately (and certainly not every frame... that's just silly).

Maybe we won't see too many games use this in the beginning. But I think it could catch on once the devs have had time to figure out how it helps them.

 

[Guevara]

This is 100% bullshit. Quote me a year from now if it turns out otherwise.

 

[Erasus]

Personally, I think the "looking better" aspect of this isn't the interesting part. It's more about enhancing AI, turn based games, massive scale MP games, stuff that can be done in the background and used later, not stuff that needs to be sent down immediately (and certainly not every frame... that's just silly).

Maybe we won't see too many games use this in the beginning. But I think it could catch on once the devs have had time to figure out how it helps them.

Thats down to being dedicated servers instead of P2P.

Also how will they handle multiple types of broadband speeds?

It all sounds very cool if it works! AI, SSAO, lightning could be probably done but its alsways gonna lag behind a bti. But seems like a lot of extra work for devs...
But having a good server infrastructure ready is always good.

 

[Goodlife]

Dull question time (could already have been covered) this whole cloud based processing could be utilised by any of the boxes (even the current gen stuff) couldn't it? There is nothing specific you'd have to include in your hardware designs, is there?

 

[la flama blanca]

Not all lighting is created equal.

Head lights and torchlights and explosions != day/night/weather cycles.

And how unique or practical would it be to offload these calculations to a cloud based server? What about if someone's connection is spotty? When would Microsoft stop supporting an older XBOX ONE game with this supposed "cloud processing"? Why wouldn't anyone of us most likely have known of this kind of advanced technology up until MS' CONSOLE RELEASE? I'll bet my fucking account that this turns into another gimmick.

 

[P44]

This might be a possibility.

But your console doesn't crunch the AI - it gets sent the pathing information, kinda like animation data.

So on your console, the AIs are following predetermined paths and behaviours, and only when you bump into and interact with them would the local console break them out of those behaviourial loops and interject its own in there.

Of course, the scene would have to be carefully constructed - you couldn't have too many people too close together at once (which kinda misses the point of cloud compute AI?) - in case the player does something that affects all of them at once (i.e. pulls out a gun in the middle of night club and shoots - you'd expect everyone in the scene to start panicking and taking cover).

Maybe you can have more complex nodes of behaviour for cloud computed AI, but when they're locally controlled for such a large group, they become 'dumb' - i.e. panicking and all running away from the player, before cloud AI kicks in and individuals start to seek for cover and pull out their cell phones, or etc.

We're sort of bumping into another problem here.

I've always felt the limiting factor in better AI was the developers generally not prioritising it after a point. I mean, I reckon very nice AI has been possible on the PC for years, but its basically stalled in many ways.

so did this work back when the cell did it? .... right ..


no one is going to waste the programming resources to code for this, hell lowest common denominator porting as it is is rife in the industry ..and that's coding to hardware platforms where specs are set and don't move.

rule of thumb, if yr console is launching in a few mths ..and nothing specific is detailed about things like this ...it's FUD.

great counter material to when ever any one asks why ps3 hw is more powerful though .... but something tells me MS sound have dedicated more of this time brainstorming effective responces to press questions about the DRM situation ... amiright

Also I can't help but think this might be true. The X1 will need extra work to get this going, work that, for multiplats, probably won't be worth it, especially since the architectures this gen are so similar that all a dev has to do is turn off a few shinies like a few less particular effects and volumetric effects etc; minor stuff thats a lot easier and efficient to do (whilst I don't know how difficult it will be to implement this solution, I can guarantee it won't be easier than toning down or off a few effects here and there). It basically from there falls on the first party to do something with the technology by and large, which isn't the greatest way to proceed.

By the sounds of it, any projected increase of power would be limited by the fact that they can only fiddle with latency insensitive areas really; the actual effects could be less as a result.

 

[zomgbbqftw]

Personally, I think the "looking better" aspect of this isn't the interesting part. It's more about enhancing AI, turn based games, massive scale MP games, stuff that can be done in the background and used later, not stuff that needs to be sent down immediately (and certainly not every frame... that's just silly).

Maybe we won't see too many games use this in the beginning. But I think it could catch on once the devs have had time to figure out how it helps them.

None of the stuff you have mentioned is particularly intensive for a modern GPU and it would just introduce latency for enemies and break the game. My connection to the local MS server will have to be less than 3-4ms so that it can calculate everything and send it back in time for the next frame in 33ms time. I don't see it personally.

 

[west]

As someone who has been working with Azure for years, I had a pretty strong reaction to the PR statements. Just to clarify, I do not work for MS nor in gaming; my work is business app related - a field where MS has been pushing it's cloud offering for a while now.

First of all: These servers have been around for a while, I doubt these 300k claims are based on something Xbox specific. If someone knows better please correct me, but I believe 300k is just the total number of all MS cloud servers currently. So for example the server running our company email is already listed.

Calculating power: The current servers are no graphical beasts, they have 8 core processors and a lot of memory, but quite basic GPU:s that are shared with up to 8 other customers ( Everything is virtualized ). Currently the servers mostly host databases, websites, sharepoint deployments etc. While they certanly might be 4x Xb one in CPU calculations, I doubt they will be doing any heavy graphics calculations anytime soon.

Price: A singe 8 core machine runs at about 500 dollars a month at the current pricing model. I just don't see how it can be feasable in the long run in games that have no monthly fees. If they actually want to calculate anything worthwhile graphics related, we are looking at a several cores per user. It simply makes zero sense to build complicated cloud support for just one extra core.

What the Cloud is good at: The cloud is good for handling large amount of data ( Disk and memory is cheap ) and doing userbase wide calculations. It might calculate statistics or auto balance the game based on a lot of user data, or it might store persistent world states for large worlds ( Online destructible environments ). But actually improving graphics, that will most likely not happen; they would lose their game profit way to fast. Again, 8 cores cost 72 dollars for 100 hours of usage at current Azure prices. How much are those shadow calculations worth?

What we will see: EA will propably make websites for all their sport games based on in game statistics and data. Lineups will be handled trough the cloud and you can compete with your dream teams or whatever. I believe most of the "exclusive" EA sport features are related to this. The big advantage in hte cloud is accessibilty and data. I see more companion iPad apps like Destiny has in the future. You can most likely see Cod statistics live from any device.

Why I am sceptical: The cloud is nothing new and Azure has been around for years. How many games use cloud calculations currently? I only see this as a way for MS to bring current web based features under one roof / API. Thus giving MS a steady revenue, and simplifying development of these kind of "off-console" experiences.

 

[derFeef]

They should have showed an example right up front. Bold claims and PR words don't cut it. Show me a game running and poulate the world, change achievements, spawn new collectibles (ORBS!)...

 

[le.phat]

Maybe they can use the cloud to give us that free AA they promised us since last gen.

 

[Zaptruder]

And how unique or practical would it be to offload these calculations to a cloud based server? What about if someone's connection is spotty? When would Microsoft stop supporting an older XBOX ONE game with this supposed "cloud processing"? Why wouldn't anyone of us most likely have known of this kind of advanced technology up until MS' CONSOLE RELEASE? I'll bet my fucking account that this turns into another gimmick.

If you can't connect to the cloud processor, obviously your Xbone will explode like a petulant child, killing any players seating in front of it, and setting the house ablaze.

Obviously, if the connection drops off, it'll stop looking as fancy/behaving as well. I don't know if this means it shouldn't have access to cloud computing resources - I think it just means it'll stop looking as fancy. Which is ok, if at bare minimum it looks better than console games from this gen by some degree.

 

[la flama blanca]

If you can't connect to the cloud processor, obviously your Xbone will explode like a petulant child, killing any players seating in front of it, and setting the house ablaze.

Obviously, if the connection drops off, it'll stop looking as fancy/behaving as well. I don't know if this means it shouldn't have access to cloud computing resources - I think it just means it'll stop looking as fancy. Which is ok, if at bare minimum it looks better than console games from this gen by some degree.

Can you provide any existing example of a cloud based solution offering this kind of processing?

 

[slider]

This is all very intriguing. I wonder how much cynicism there'll be if MS show off stellar looking games at E3 "enhanced by the cloud".

And I wish there were some parameters set for cloud usage - it all feels a little too abstract for me at the moment.

 

[zomgbbqftw]

As someone who has been working with Azure for years, I had a pretty strong reaction to the PR statements. Just to clarify, I do not work for MS nor in gaming; my work is business app related - a field where MS has been pushing it's cloud offering for a while now.

First of all: These servers have been around for a while, I doubt these 300k claims are based on something Xbox specific. If someone knows better please correct me, but I believe 300k is just the total number of all MS cloud servers currently. So for example the server running our company email is already listed.

Calculating power: The current servers are no graphical beasts, they have 8 core processors and a lot of memory, but quite basic GPU:s that are shared with up to 8 other customers ( Everything is virtualized ). Currently the servers mostly host databases, websites, sharepoint deployments etc. While they certanly might be 4x Xb one in CPU calculations, I doubt they will be doing any heavy graphics calculations anytime soon.

Price: A singe 8 core machine runs at about 500 dollars a month at the current pricing model. I just don't see how it can be feasable in the long run in games that have no monthly fees. If they actually want to calculate anything worthwhile graphics related, we are looking at a several cores per user. It simply makes zero sense to build complicated cloud support for just one extra core.

What the Cloud is good at: The cloud is good for handling large amount of data ( Disk and memory is cheap ) and doing userbase wide calculations. It might calculate statistics or auto balance the game based on a lot of user data, or it might store persistent world states for large worlds ( Online destructible environments ). But actually improving graphics, that will most likely not happen; they would lose their game profit way to fast. Again, 8 cores cost 72 dollars for 100 hours of usage at current Azure prices. How much are those shadow calculations worth?

What we will see: EA will propably make websites for all their sport games based on in game statistics and data. Lineups will be handled trough the cloud and you can compete with your dream teams or whatever. I believe most of the "exclusive" EA sport features are related to this. The big advantage in hte cloud is accessibilty and data. I see more companion iPad apps like Destiny has in the future. You can most likely see Cod statistics live from any device.

Why I am sceptical: The cloud is nothing new and Azure has been around for years. How many games use cloud calculations currently? I only see this as a way for MS to bring current web based features under one roof / API. Thus giving MS a steady revenue, and simplifying development of these kind of "off-console" experiences.

This post should be added to the OP. Real world experience with Azure.

 

[SeaOfMadness]

None of the stuff you have mentioned is particularly intensive for a modern GPU and it would just introduce latency for enemies and break the game. My connection to the local MS server will have to be less than 3-4ms so that it can calculate everything and send it back in time for the next frame in 33ms time. I don't see it personally.

I don't either and that's not what I was suggesting. Sending data down every 16 or 33ms is NOT what this is designed for. It's for things that can be done in idle time, or done while a level is loading. I already gave examples of terrain generation, high level AI decisions, things of that nature.

I wish I had better examples. But not having to do those things locally, means more of the CPU and GPU can be used for the real time stuff that can't be sent to/from a server.

 

[P44]

As someone who has been working with Azure for years, I had a pretty strong reaction to the PR statements. Just to clarify, I do not work for MS nor in gaming; my work is business app related - a field where MS has been pushing it's cloud offering for a while now.

First of all: These servers have been around for a while, I doubt these 300k claims are based on something Xbox specific. If someone knows better please correct me, but I believe 300k is just the total number of all MS cloud servers currently. So for example the server running our company email is already listed.

Calculating power: The current servers are no graphical beasts, they have 8 core processors and a lot of memory, but quite basic GPU:s that are shared with up to 8 other customers ( Everything is virtualized ). Currently the servers mostly host databases, websites, sharepoint deployments etc. While they certanly might be 4x Xb one in CPU calculations, I doubt they will be doing any heavy graphics calculations anytime soon.

Price: A singe 8 core machine runs at about 500 dollars a month at the current pricing model. I just don't see how it can be feasable in the long run in games that have no monthly fees. If they actually want to calculate anything worthwhile graphics related, we are looking at a several cores per user. It simply makes zero sense to build complicated cloud support for just one extra core.

What the Cloud is good at: The cloud is good for handling large amount of data ( Disk and memory is cheap ) and doing userbase wide calculations. It might calculate statistics or auto balance the game based on a lot of user data, or it might store persistent world states for large worlds ( Online destructible environments ). But actually improving graphics, that will most likely not happen; they would lose their game profit way to fast. Again, 8 cores cost 72 dollars for 100 hours of usage at current Azure prices. How much are those shadow calculations worth?

What we will see: EA will propably make websites for all their sport games based on in game statistics and data. Lineups will be handled trough the cloud and you can compete with your dream teams or whatever. I believe most of the "exclusive" EA sport features are related to this. The big advantage in hte cloud is accessibilty and data. I see more companion iPad apps like Destiny has in the future. You can most likely see Cod statistics live from any device.

Why I am sceptical: The cloud is nothing new and Azure has been around for years. How many games use cloud calculations currently? I only see this as a way for MS to bring current web based features under one roof / API. Thus giving MS a steady revenue, and simplifying development of these kind of "off-console" experiences.

Mate, I really appreciate this post. I'm very out of my element trying to work out how this will all work, so, some real life experience with the servers in question is a real help in working out where this is going.

 

[Plaguefox]

.

 

[Krilekk]

Three servers for each console? Am I reading this right?

More like three virtual machines for each console. With how weak those consoles are it should easily be possible to have multiple VMs per server. And don't forget, they've had seven years of collecting data on concurrent users, regional loads and shifting of loads. They should know exactly how many servers they need to offer the service on every day except Christmas eve. ;-) And since server tech actually advances unlike console tech you wouldn't need more servers, just faster servers in the future.

The reason they are doing this is fairly simple: They have their cloud (Azure) already and it doesn't have a lot to do. IIRC in 2011 they had average numbers of 880.000 calculations per second. That basically means Azure idles all the time but it needs to be ready for corporations to use it. So they have their cloud service layered around the developed world, just waiting for tasks. Might as well just give them to it.

 

[equippedwithtowel]

If you can't connect to the cloud processor, obviously your Xbone will explode like a petulant child, killing any players seating in front of it, and setting the house ablaze.

Obviously, if the connection drops off, it'll stop looking as fancy/behaving as well. I don't know if this means it shouldn't have access to cloud computing resources - I think it just means it'll stop looking as fancy. Which is ok, if at bare minimum it looks better than console games from this gen by some degree.

So the cloud is handicapped to a small number of effects on the periphery. Not that 4x performance boost MS is going on about.

 

[Demon Ice]

Looking forward to it being like UE3 on the fly texture streaming, but with...lighting and physics. Lol.

In all seriousness, there are very few aspects of a game that I can think of that are latency insensitive. Even lighting and physics need to be dynamically calculated if they're going to react to the player's actions.

Edit: I don't know why all these developers keep saying you can do AI on the cloud. AI needs to react instantaneously, absolutely no way it can be offloaded to the cloud. Unless they're talking about keeping basic AI on the console and doing advanced stuff on the cloud, but then what, you have an internet outage and you suddenly get dipshit AI?

 

[P44]

So the cloud is handicapped to a small number of effects on the periphery. Not that 4x performance boost MS is going on about.

Yeah that's more or less what I assumed - the raw performance icnrease might be 4x or whatever, but in real terms, I don't think the games will look a lot nicer, purely due to what they can and can't cloud compute.

 

[Klocker]

RE: Servers

allegedly these are 300k new servers just coming online this coming year and will be for Xbox Live.

that is one reason it is not being shown yet, the infrastructure is just being completed. they mentioned in the Engineer panel discussion that it would be very soon now that they could start leveraging it.

 

[Krilekk]

This post should be added to the OP. Real world experience with Azure.

Xbox Live already runs on Azure. As does Outlook. Everybody has real world experiences with Azure.

 

[Erasus]

text

Yes, I was a bit sceptical. Servers usually dont have GPUs so lightning and stuff would be hard. AI is CPU driven but having only 8 cores per server it dosent seem worth it.

As you said, companion apps and databases for massive worlds can be on the "cloud", like how an MMO works.

Xbox Live already runs on Azure. As does Outlook. Everybody has real world experiences with Azure.

Well using it does not mean knowing how it works or what is in it. Driving a car does not make me an expert on engines.

 

[SeaOfMadness]

Looking forward to it being like UE3 on the fly texture streaming, but with...lighting and physics. Lol.

In all seriousness, there are very few aspects of a game that I can think of that are latency insensitive. Even lighting and physics need to be dynamically calculated if they're going to react to the player's actions.

That's kind of the issue here. No one has really come up with any compelling scenarios that justify this, but that doesn't mean there aren't any.

15 years ago, I would guess people had the same conversation about GPUs and what they could be used for. I think we know the answer to that very well now

 

[LabouredSubterfuge]

It's nice in theory but wouldn't it have been better to put all those resources into simply having better hardware in the box to deal with these things more quickly and offline?

 

[SeaOfMadness]

Yeah that's more or less what I assumed - the raw performance icnrease might be 4x or whatever, but in real terms, I don't think the games will look a lot nicer, purely due to what they can and can't cloud compute.

I don't know why everyone is so hung up on this somehow making games look nicer. Maybe someone will figure out some way to leverage the cloud for that, but in the short term, I'm not sure you're going to see it used that way.

We're talking about general purpose computing power which can be used for lots of non-graphical things.

 

[SenjutsuSage]

If Sony was talking about this same thing on this same level, people wouldn't be so "skeptical." Microsoft wouldn't be lying so outright about what's possible on the new xbox with the cloud infrastructure they've setup, contrary to what some think.

 

[zomgbbqftw]

Xbox Live already runs on Azure. As does Outlook. Everybody has real world experiences with Azure.

Yes because I use outlook for my work email it makes me an expert since I am the sysadmin at work and get the server time from MS. Clearly all of us here have had contact with Microsoft's Azure team to get server space for their Outlook.com account and XBL. I mean just last week I called up the Azure team to renew my XBL subscription.

 

[P44]

I don't know why everyone is so hung up on this somehow making games look nicer. Maybe someone will figure out some way to leverage the cloud for that, but in the short term, I'm not sure you're going to see it used that way.

We're talking about general purpose computing power which can be used for lots of non-graphical things.

I think I'm just cynical about developers actually going to lengths to program better AI, or physics simulations etc to take advantage of these things. It creates fragmentation in a gen where the architectures are similar, for most multiplat devs, it will be more of a headache than its worth I think.

 

[Gotchaye]

As someone who has been working with Azure for years, I had a pretty strong reaction to the PR statements. Just to clarify, I do not work for MS nor in gaming; my work is business app related - a field where MS has been pushing it's cloud offering for a while now.

First of all: These servers have been around for a while, I doubt these 300k claims are based on something Xbox specific. If someone knows better please correct me, but I believe 300k is just the total number of all MS cloud servers currently. So for example the server running our company email is already listed.

Calculating power: The current servers are no graphical beasts, they have 8 core processors and a lot of memory, but quite basic GPU:s that are shared with up to 8 other customers ( Everything is virtualized ). Currently the servers mostly host databases, websites, sharepoint deployments etc. While they certanly might be 4x Xb one in CPU calculations, I doubt they will be doing any heavy graphics calculations anytime soon.

Price: A singe 8 core machine runs at about 500 dollars a month at the current pricing model. I just don't see how it can be feasable in the long run in games that have no monthly fees. If they actually want to calculate anything worthwhile graphics related, we are looking at a several cores per user. It simply makes zero sense to build complicated cloud support for just one extra core.

What the Cloud is good at: The cloud is good for handling large amount of data ( Disk and memory is cheap ) and doing userbase wide calculations. It might calculate statistics or auto balance the game based on a lot of user data, or it might store persistent world states for large worlds ( Online destructible environments ). But actually improving graphics, that will most likely not happen; they would lose their game profit way to fast. Again, 8 cores cost 72 dollars for 100 hours of usage at current Azure prices. How much are those shadow calculations worth?

What we will see: EA will propably make websites for all their sport games based on in game statistics and data. Lineups will be handled trough the cloud and you can compete with your dream teams or whatever. I believe most of the "exclusive" EA sport features are related to this. The big advantage in hte cloud is accessibilty and data. I see more companion iPad apps like Destiny has in the future. You can most likely see Cod statistics live from any device.

Why I am sceptical: The cloud is nothing new and Azure has been around for years. How many games use cloud calculations currently? I only see this as a way for MS to bring current web based features under one roof / API. Thus giving MS a steady revenue, and simplifying development of these kind of "off-console" experiences.

Good post. The obvious application to me is shipping pre-computed data to consoles.

Consider the lighting examples. A lot of the sorts of calculations being done are deterministic and have nothing to do with player input. The reason current consoles and PCs do all this computing when a level loads is just that it's not worth putting the pre-computed light map on the player's hard drive or on the disc. But on the cloud, you can compute that light map once and then store it there, sending it out to clients upon request.

This way real graphical improvements can be provided without having to dedicate significant computational resources to each client. The only downside to losing the internet connection would be longer load times, in the best case.

 

[Plaguefox]

.

 

[Gotchaye]

Forgive my ignorance; why is it not worth putting pre-computed light maps on the player's hard drive or the game disc?

I'm mostly talking out of my ass. There's presumably a reason they don't do it, and it's hard to see what possible reason there could be other than "takes up too much space".

If the lighting is changing in some predictable way (time of day, say), you might end up with a very large number of discrete lighting conditions, even if an instantaneous light map is small. You could store one-second resolution light maps on a server and ship out 60 of them at a time to clients pretty readily, I figure. The storage requirements are minimal because your stored copy is being used by a large number of clients.

 

[NemesisPrime]

I am actually working on a game that uses Azure for procedural generation and the Cloud version is up to 100 times faster in loading levels. (since Azure can generate the procedurally generated data much faster than a mobile phone).

Minecraft for example could generate a world on the cloud in a few seconds instead of 20 or so on 360.

The problem is finding a good fallback scenario for when you have no internet connection. This is the biggest problem atm.