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

Well, MS made those hand-wavy comments about how every XB1 would have access to the equivalent of three more XB1's worth of CPU and memory up in the cloud. I took that to mean that there might be a few hundred GFLOPS of CPU on tap up there, but definitely not several TFLOPS of GPU resources. They certainly would have made a big deal about that, if it had been the case. And of course, Azure wasn't built primarily for gaming/graphics. It's essentially a bunch of Web/DB/Application servers.

It would seem that Sony/Gaikai's idea of the "cloud" is much closer to that of nVidia's: Lots of GPU in the cloud, able to do actual (complete) game rendering, or failing that, at least do render-assist stuff like these "bonus" irradiance calculations that the nVidia paper talks about.

Sadly, I take the grand plans from Sony-Gaikai (great cyberpunk corp name!) that we've heard about with a grain of salt. Building that up from scratch, and then monetizing it, sounds like a tall order given Sony's resources.


Azure is pretty capable of delivering gaming services, may not be very public yet BUT i'm guessing we'll see it proved within this coming year ...
http://blogs.msdn.com/b/david_gristwood/archive/2013/07/25/orleans-windows-azure-and-halo-4.aspx
 
Azure is pretty capable of delivering gaming services, may not be very public yet BUT i'm guessing we'll see it proved within this coming year ...
http://blogs.msdn.com/b/david_gristwood/archive/2013/07/25/orleans-windows-azure-and-halo-4.aspx

He is talking about rendering games on the cloud, the stuff that 434 has done with Orelans and Azure is more based around getting statistics and information out to other people, not rendering 3d graphics.

http://msrvideo.vo.msecnd.net/rmcvideos/198324/dl/198324.mp4
 
Sadly, I take the grand plans from Sony-Gaikai (great cyberpunk corp name!) that we've heard about with a grain of salt. Building that up from scratch, and then monetizing it, sounds like a tall order given Sony's resources.

There are 2 different models for the industry:
- the console is in your living room, and you "augment it" via "the cloud". This is a very interesting idea, it has very nice latency for *offline games*, but has huge problems for developers, infrastructure and basic maths.
- the console is in a datacenter, a dumb terminal is in your living room. This has nice latency for *online games*, and meh latency for offline - but is mostly trivial to implement for developers.

In truth the second model is more immediately valuable - being able to sell PSN+ subscriptions on smartTVs/STBs/tablets/toasters may be a very profitable route for Sony. (the thing Sony needs to avoid is too much 'vita-focus', they don't need to control the platform of the device if they control the distribution channel).

The bigger question is whether augmentation is really a 'thing' or just some buzzwords stringed together. Currently, everything we've seen indicates 'buzzwords'.
 
There are 2 different models for the industry:
- the console is in your living room, and you "augment it" via "the cloud". This is a very interesting idea, it has very nice latency for *offline games*, but has huge problems for developers, infrastructure and basic maths.
- the console is in a datacenter, a dumb terminal is in your living room. This has nice latency for *online games*, and meh latency for offline - but is mostly trivial to implement for developers.

In truth the second model is more immediately valuable - being able to sell PSN+ subscriptions on smartTVs/STBs/tablets/toasters may be a very profitable route for Sony. (the thing Sony needs to avoid is too much 'vita-focus', they don't need to control the platform of the device if they control the distribution channel).

The bigger question is whether augmentation is really a 'thing' or just some buzzwords stringed together. Currently, everything we've seen indicates 'buzzwords'.

That second method is also significantly more expensive per user as you require graphics rendering with X amount of GPU rendering power, Y amount of CPU compute, and additional resources for video recording and compression per user whereas with the current compute only cloud it's a variable amount of compute per user and relatively easily scaled.

That GPU rendering power and CPU compute power aren't scalable per user for method 2. It must have X and Y amount of each per user as it requires that in order to render and run the game before it is captured and compressed prior to sending the compressed video stream to the end users device. Making it profitable thus isn't easy and it remains to be seen if Sony will be able to succeed where others have failed. With OnLive, even when they were basically giving it away for free, they were unable to increase userbase in any significant way.

Regards,
SB
 
Tom Hiring, an engineer working with Cloud technology wrote the most comprehensive article on the cloud to date, imho. :smile: He is working on Titanfall.

He also explains the differences between player-hosted servers, dedicated servers and the Cloud.

http://www.respawn.com/news/lets-talk-about-the-xbox-live-cloud/

Dedicated Servers
Dedicated servers are when a computer sitting out on the internet handles all of the host duties, leaving every client free to just be a client.

  • You can get even more
    • CPU on your dedicated servers to do new things like dozens of AI and giant autopilot titans!
    • Suddenly you have no more host advantage!
    • Bandwidth for the servers is guaranteed from the hosting provider!
    • You can use all of the available CPU and memory on the player machines for awesome visuals and audio!
    • Hacked-host cheating isn’t an issue!
    • Matchmaking can be lightning fast since it’s guaranteed that everyone can connect to your servers.
    • And since the servers aren’t going to go disconnect to watch Netflix, you don’t need to migrate hosts anymore!
    The player experience is so much better. This sounds awesome!

    But it costs a LOT of money.

    This is something I have worked on for years now, since coming to Respawn. A developer like Respawn doesn’t have the kind of weight to get a huge price cut from places like Amazon or Rackspace.

    And we don’t have the manpower to manage literally hundreds-of-thousands of servers ourselves. We want to focus on making awesome games, not on becoming giant worldwide server hosting providers. The more time I can spend on making our actual game better, the more our players benefit.


    I personally talked to both Microsoft and Sony and explained that we need to find a way to have potentially hundreds-of-thousands of dedicated servers at a price point that you can’t get right now.

    Microsoft realized that player-hosted servers are actually holding back online gaming and that this is something that they could help solve, and ran full-speed with this idea.


    The Xbox group came back to us with a way for us to run all of these Titanfall dedicated servers and that lets us push games with more server CPU and higher bandwidth, which lets us have a bigger world, more physics, lots of AI, and potentially a lot more than that!

    What is the Cloud?


    Amazon has a cloud that powers websites. Sony has a cloud that streams game video so you can play a game that you don’t have on your machine. Now Xbox Live has a cloud that somehow powers games.

    Cloud doesn’t seem to actually mean anything anymore, or it has so many meanings that it’s useless as a marketing word.


    Let me explain this simply: when companies talk about their cloud, all they are saying is that they have a huge amount of servers ready to run whatever you need them to run. That’s all.

    So what is this Xbox Live Cloud stuff then?


    Microsoft has a cloud service called Azure (it’s a real thing – you can go on their website right now and pay for servers and use them to run whatever you want). Microsoft realized that they could use that technology to solve our problem.


    So they built this powerful system to let us create all sorts of tasks that they will run for us, and it can scale up and down automatically as players come and go. We can upload new programs for them to run and they handle the deployment for us.

    And they’ll host our game servers for other platforms, too! Titanfall uses the Xbox Live Cloud to run dedicated servers for PC, Xbox One, and Xbox 360.


    But it’s not just for dedicated servers – Microsoft thought about our problem in a bigger way. Developers aren’t going to just want dedicated servers – they’ll have all kinds of features that need a server to do some kind of work to make games better.

    Look at Forza 5, which studies your driving style in order to create custom AI that behaves like you do. That’s totally different from what Titanfall uses it for, and it’s really cool! So it’s not accurate to say that the Xbox Live Cloud is simply a system for running dedicated servers – it can do a lot more than that.

    How is this different from other dedicated servers?


    With the Xbox Live Cloud, we don’t have to worry about estimating how many servers we’ll need on launch day. We don’t have to find ISPs all over the globe and rent servers from each one. We don’t have to maintain the servers or copy new builds to every server.

    That lets us focus on things that make our game more fun. And best yet, Microsoft has datacenters all over the world, so everyone playing our game should have a consistent, low latency connection to their local datacenter.


    Most importantly to us, Microsoft priced it so that it’s far more affordable than other hosting options – their goal here is to get more awesome games, not to nickel-and-dime developers.

    So because of this, dedicated servers are much more of a realistic option for developers who don’t want to make compromises on their player experience, and it opens up a lot more things that we can do in an online game.

    Wrapping up…


    This is a really big deal, and it can make online games better. This is something that we are really excited about. The Xbox Live Cloud lets us to do things in Titanfall that no player-hosted multiplayer game can do.
  • That has allowed us to push the boundaries in online multiplayer and that’s awesome. We want to try new ideas and let the player do things they’ve never been able to do before! Over time, I expect that we’ll be using these servers to do a lot more than just dedicated servers. This is something that’s going to let us drive all sorts of new ideas in online games for years to come.


    I know this got pretty technical and long-winded, so I thank you for reading this far. Hopefully I’ve cleared some things up, and you can see why I’m so excited about what Microsoft has done here and how it is letting us do awesome new things for our game. I’ll see you online in the spring to play some Titanfall on our dedicated servers!
 
That second method is also significantly more expensive per user as you require graphics rendering with X amount of GPU rendering power, Y amount of CPU compute, and additional resources for video recording and compression per user whereas with the current compute only cloud it's a variable amount of compute per user and relatively easily scaled.

It depends how 'hard' that compute reservation is. If you're doing something real-time, then there may not be a huge difference between the 2 systems.

But I'd certainly expect gaikai to be more expensive than a compute cloud, but I also see profit being far more likely than trying to compete against amazon/google for cloud compute.

With OnLive, even when they were basically giving it away for free, they were unable to increase userbase in any significant way.

Onlive were buying hardware from A, an operating system from B and games from C/D (and then updating the hardware every ~12 months). They were then selling PC games to PC gamers... There was no obvious profit margin, and no obvious demand.

Sony would be buying the majority of their resources from themselves, and attempting to sell "console games to non-console gamers".

Beyond the initial structural investment, the costs for this would appear to be very low.
 
It depends how 'hard' that compute reservation is. If you're doing something real-time, then there may not be a huge difference between the 2 systems.

But I'd certainly expect gaikai to be more expensive than a compute cloud, but I also see profit being far more likely than trying to compete against amazon/google for cloud compute.



Onlive were buying hardware from A, an operating system from B and games from C/D (and then updating the hardware every ~12 months). They were then selling PC games to PC gamers... There was no obvious profit margin, and no obvious demand.

Sony would be buying the majority of their resources from themselves, and attempting to sell "console games to non-console gamers".

Beyond the initial structural investment, the costs for this would appear to be very low.

It'll be relatively cheap for PS3 games, a whole PS3 on a board shouldn't be "that" much. For PS4? You'll need the SOC and 5 or more GB of GDDR5 at a minimum for each user that is using the service at any given time. Compared to OnLive's service for PC gaming, they reduced both resolution and in game quality settings to fairly low levels in order to use cheaper hardware. At the settings they used, you could get away with 1 GB of system RAM or less and at the very most 1 GB of video ram (probably closer to 256 MB of video RAM at the quality settings they usually chose). Per user and per game, it should be significantly more expensive for Sony when offering PS4 games over Gaikai, For PS3 games, probably not. As well, unless you want to have users faced with "no Gaikai server available" when they try to play a game, you have to overprovision how many PS3 based boards are in any given area based on what you expect to be your max user load at any given moment. Since you can't run PS3 software on PS4 and vice versa that also means you have to have a PS4 for each location in the same or more numbers per maximum expected users. Unlike Onlive you can't run games from 10 years ago on the same machine as you run future games.

Now when getting to general compute versus running full games. Does a game using cloud compute require +10% of the local machine's (PS4 or Xbox One) compute capabilities? +20%? +100%? Even at +100% you still don't need to include the cost for GDDR5 (expensive memory) or even need a GPU. And since the processing required scales per game, it's unlikely you'd ever approach a 1:1 relationship between games using the cloud per user and CPU/memory resources per game/user.

Meaning that at any given time, serving 1 million users will always be far cheaper for purely cloud compute versus 1 million users for console game streaming. Even at +100% compute of the local machine but in the cloud (so doubling the CPU compute available to a given game), that's still going to be far cheaper as you don't need to replicate the resources for graphics rendering. Something like an MMO will require multiple times the compute and memory resources of the consoles, but then you are also serving potentially thousands of users per MMO world, once again reducing the per user compute and memory requirements to less than each individual users machine.

And that doesn't even get into bandwidth costs associated with streaming reasonable quality 720p video (most PS3 gen games) or up to 1080 (PS4 gen games). That's going to be significantly higher than anything that will be used for compute. 720p video will use around 4-6 Mbps depending on quality. If you sacrifice enough quality you can probably get a bit lower, but then who would want to play at greatly reduced graphics IQ? Microsoft recommended 1.5 Mbps for Xbox One for best quality. And for the absolute simplest situation of dedicated servers or MMO's at most we're talking around 100 Kbps if a LOT of stuff in happening on screen (a Raid in an MMO for example). I'll be interested to see just how much network traffic Titanfall is going to generate. I'm going to guess nowhere near the 1.5 Mbps recommended by Microsoft for online.

Regards,
SB
 
Having read all of this post its nice to see the change from disbelief ...I.E..just Microsoft using cloud to divert focus from there weak machine to a belief in the possibility of cloud based improvements in some games as well as the hope of new game ideas coming out of this tech.

When I first heard about the possibility of cloud enhanced gaming I didn't think it was a marketing ploy .....maybe I'm naive but my first thought was cool can't wait to see what developers come up with .

Being a life long gamer I am always looking for the next big leap in the hobby I enjoy .......maybe this is it after all games looking prettier is nice and all but what I really want to see is a leap in game play through AI in world like skyrim .....fallout ....maybe this tech will give us the next big leap forward much like the dedicated GP did all the years ago .
 
More from that article:

Turn 10's approach, Greenawalt explained, is to use the cloud servers for its new Drivatar feature; virtual racing profiles that learn real gamers' abilities, skills and techniques and are then fed to other players' games to give computer opponents an all-new dimension. Without the cloud, these profiles would not be possible. They take a lot of processing to devise and update, but not a lot of bandwidth to move back and forth between consoles.

"Here's what the cloud's great at; how about I throw data at it once in a while - let's call it at the end of every race, maybe even every second - and throw a lot of data up there. I just trickle it up there. And these servers are just crunching away at supercomputer rates, all of this data. It's exactly what Bing does, it's a lot of data," he said.

"I can't do that in a box, you have got to do that in the servers. But it doesn't have to be done every split second. It just sits up there, and it crunches and crunches and then trickles back down to where the box is. That's what we do with Drivatars."
 
I'm going to guess they mean either...

1) Only for games which are also available on Xbox One.
or
2) For games that aren't on Xbox One, it'll share resources with the rest of the Azure platform.

Regards,
SB
 
Hmmm, the Titanfall guys said it could be used for both Xbox One and PC.

TitanFall uses "Xbox Live Cloud" for PC, XB360 and XB1, but that doesn't mean it uses the 300k servers dedicated for XB1 for all of them

edit: Clarification: It uses MS Azure Cloud but only XB1 uses the 300k servers dedicated for it
 
Last edited by a moderator:
I believe that Xbox Live Cloud is an umbrella term.
The XB1 has access to some parts (API's etc) that x360 and PC's does not.
I think that's where the confusion is coming from.

It's about finding out the details to what are XB1 only.
 
I have HUGE curiosity related to CPU power and... the cloud.
I really hope the times now are a bit more mature for talk about the "cloud" topic.

A part of all the talking about extra AI and extra Physic etc... achivable via cloud (that seems something that they are really implementing in Titanfall, by the way, but we will see), at the moment, I would like to talk only about the purely "multiplayer" aspect.

Which would be the impact on the CPU of the "multiplayer managment system" in a game like Titanfall if the multiplayer could be managed in the old fashion way (without dedicated servers)?

Some of my "tech wiser" folks seem to believe that the impact on CPU could be quite big. And this without mentioning the much weaker performances in term of gameplay (and the absence of extra AI and extra physic etc...).

In specific, my 2 questions are:

1) Titanfall could be doable without dedicated servers?

2) Which would be the impact on the CPU, without dedicated servers?
 
1) Titanfall could be doable without dedicated servers?

2) Which would be the impact on the CPU, without dedicated servers?

Why not? Assuming it's client-server, it'll cost the host more CPU to "host" the game. Also the network bandwidth would be headache, that's pretty much why most console FPS are capped at a relatively low number of players in a game (24?).

"Cloud" is more of a marketing terms, the way that I see it is that companies will be able to dynamically scale up/down server capacities on demand, as oppose to plan and host a dedicated pool of servers. Hopefully former will be cheaper in either the time or the money, maybe both.
 
Why not? Assuming it's client-server, it'll cost the host more CPU to "host" the game. Also the network bandwidth would be headache, that's pretty much why most console FPS are capped at a relatively low number of players in a game (24?).

"Cloud" is more of a marketing terms, the way that I see it is that companies will be able to dynamically scale up/down server capacities on demand, as oppose to plan and host a dedicated pool of servers. Hopefully former will be cheaper in either the time or the money, maybe both.

But, if it cost the host more cpu to host the game, this cpu cost/reservation had to be present for everyone in order to work client-server.

So, which would be the impact on CPU of a Titanfall managed via client-server?
1% of CPU ? 10%? 1 core? I do not know.
 
Back
Top