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

[dumbo11]

And I am yet to believe there claims of 10x the performance from 3x the power. It would be nice if they showed a real world scenario, in a game, on a home internet connection.

Some interesting benchmarks that indicate a certain degree of 'PR bull manure' in those quotes...
http://blog.cloudharmony.com/2013/06/value-of-the-cloud-cpu-performance.html
http://blogs.msdn.com/b/windowsazur...rks-show-top-performance-for-big-compute.aspx

I think that's around 50 azure cores = 1TFlop?

IMHO:
- the 3x resource refers to how the cloud is configured. 3 cloud "instances" are created per client - 1 for download, 1 for profile/stats, and 1 for game server.
- the 10x thing is 'out of context' (I suspect it refers to storage capacity rather than performance).

 

[hesido]

Some interesting benchmarks that indicate a certain degree of 'PR bull manure' in those quotes...
http://blog.cloudharmony.com/2013/06/value-of-the-cloud-cpu-performance.html
http://blogs.msdn.com/b/windowsazur...rks-show-top-performance-for-big-compute.aspx

I think that's around 50 azure cores = 1TFlop?

IMHO:
- the 3x resource refers to how the cloud is configured. 3 cloud "instances" are created per client - 1 for download, 1 for profile/stats, and 1 for game server.
- the 10x thing is 'out of context' (I suspect it refers to storage capacity rather than performance).

But they are wording it in a way to make it impossible to know what that context is, purely for deceptive marketing (IMHO).

 

[Shifty Geezer]

They actually did show demo where the amount of asteroids on screen was increased dramatically by the cloud

As previously discussed, that demo means absolutely nothing as there weren't enough details to form an understanding of what work was being done on the cloud ad how the data was being managed.

 

[Silent_Buddha]

As previously discussed, that demo means absolutely nothing as there weren't enough details to form an understanding of what work was being done on the cloud ad how the data was being managed.

IIRC correctly the main point of it was that each asteroid was having it's path, speed, trajectory, location, etc. all calculated in realtime to exactly match the asteroids that NASA tracks.

From here, lightly touching on what computations they are doing...

http://thenextweb.com/microsoft/201...o-asteroid-maps-next-gen-kinect-and-football/

And here, touching on where the data on asteroid movements comes from...

http://www.neowin.net/news/e3-2013-microsoft-shows-us-some-xbox-one-cloud-and-kinect-demos...

I'm pretty sure I also read one article that mentioned the Xbox One being maxed out computationally with the initial 40k (I'm assuming purely CPU and likely just 6 cores? Or less due to overhead for the program itself.), but don't have time to go digging around.

But yes, it'd be interesting to see exactly what calculations are being done. What precision those calculations are using, etc. Are there additional calculations being done to ensure that at any given time, the asteroids match their tracked positions, etc.

Regards,
SB

 

[Shifty Geezer]

The calculations are somewhat immaterial. The question is the data - there's no description of what internet connection is behind this demo, nor the format of the data. It's certainly not 500k 2D float vectors per frame. Without that info, we can't look at how this simulation relates to any other tasks on the cloud. Like the Cell raytracing demo's were no indicator that we'd have qualitatively adjusted supersampled raytraced terrain. We can't even deduce that the cloud could handle 1000 bots in 3D space, because the data requirements of handling one bot is so much greater than modelling a single point in space.

 

[Gipsel]

there's no description of what internet connection is behind this demo

In some article it was stated they had a dedicated (high bandwidth) connection.

I later ask him if this demo is for real. We're seeing it in the middle of the E3 show, after all, where Internet connections are notoriously dreadful. Yes, it's real, he says, "We are on an Ethernet line with a private line out connected to our data centers."

As I said before, there is a significant probability it was just a weird bandwidth test. They claimed up to 500,000 updates per second. Six float values (coordinates + speed, i.e. 24 byte per particle) per update result in ~12 MB/s or just about what you can get over a 100MBit line.

 

[Silent_Buddha]

The calculations are somewhat immaterial. The question is the data - there's no description of what internet connection is behind this demo, nor the format of the data. It's certainly not 500k 2D float vectors per frame. Without that info, we can't look at how this simulation relates to any other tasks on the cloud. Like the Cell raytracing demo's were no indicator that we'd have qualitatively adjusted supersampled raytraced terrain. We can't even deduce that the cloud could handle 1000 bots in 3D space, because the data requirements of handling one bot is so much greater than modelling a single point in space.

I would imagine that at a minimum they would only have to transfer the coordinates of each of the 260,000 extra asteroids (x,y,z). But that leaves you potentially vulnerable to minor hiccups in latency. So perhaps the coordinates for 500 ms worth of calculations initially, and then just each iteration from that point forward? That would cover all but the most erratic of internet connections. I'm not sure that would necessarily be enough for a fluid display, or it could be more information than you actually need.

But yes, additional details would be quite interesting.

Regards,
SB

 

[hesido]

I would imagine that at a minimum they would only have to transfer the coordinates of each of the 260,000 extra asteroids (x,y,z). But that leaves you potentially vulnerable to minor hiccups in latency. So perhaps the coordinates for 500 ms worth of calculations initially, and then just each iteration from that point forward? That would cover all but the most erratic of internet connections. I'm not sure that would necessarily be enough for a fluid display, or it could be more information than you actually need.

But yes, additional details would be quite interesting.

Regards,
SB

Also their example seems to be non-interactive, which also lowers the value of the demo.

 

[Kaotik]

Also their example seems to be non-interactive, which also lowers the value of the demo.

Why? Server based augmentations don't have to be interactive to improve gaming experience

 

[Rurouni]

Why? Server based augmentations don't have to be interactive to improve gaming experience

for me, if it was only for stuff that happening in the background and don't influence the player, then those stuff might as well be faked or use a less accurate technique (use as little resource as possible).
For stuff that at the core needs a server (MMO stuff, competitive stuff), I don't really mind if devs add some new ways in utilizing the cloud. But for true single player game or local game or the single player element of a game, I really don't like the cloud idea. Maybe for stuff like dark soul is okay-ish, but if it alters the graphic and interactivity/gameplay in a significant way that online vs offline experience have a large-ish difference, then no. If the devs choose to disable the cloud, then basically you'll have a difference (usually inferior) experience playing it back then and playing it again in the future.

Edit: I would be somewhat fine if those games that use server augmentations gave a guarantee that the server would be up at least x years after I purchased the game. Of course this would be hard if there are still physical distribution.

 

[Kaotik]

for me, if it was only for stuff that happening in the background and don't influence the player, then those stuff might as well be faked or use a less accurate technique (use as little resource as possible).
For stuff that at the core needs a server (MMO stuff, competitive stuff), I don't really mind if devs add some new ways in utilizing the cloud. But for true single player game or local game or the single player element of a game, I really don't like the cloud idea. Maybe for stuff like dark soul is okay-ish, but if it alters the graphic and interactivity/gameplay in a significant way that online vs offline experience have a large-ish difference, then no. If the devs choose to disable the cloud, then basically you'll have a difference (usually inferior) experience playing it back then and playing it again in the future.

Edit: I would be somewhat fine if those games that use server augmentations gave a guarantee that the server would be up at least x years after I purchased the game. Of course this would be hard if there are still physical distribution.

With cloud they use virtual servers, it shouldn't require much to keep the game servers available almost indefinitely since if no-one plays, there's no need to keep a virtual server up, if one starts playing, pop the server up.

For local / single player games, you could have that "use as little resources as possible" when not connected, but when connected wouldn't you rather choose "a living world" instead of "fake something up"?

 

[Rurouni]

With cloud they use virtual servers, it shouldn't require much to keep the game servers available almost indefinitely since if no-one plays, there's no need to keep a virtual server up, if one starts playing, pop the server up.

For local / single player games, you could have that "use as little resources as possible" when not connected, but when connected wouldn't you rather choose "a living world" instead of "fake something up"?

If the augmentations doesn't affect much, I rather not depend on it since the beginning.
The thing is, even if the cloud scales the load dynamically, I believe the publisher still have to paid for it. I don't know how much excited a publisher will be if suddenly an old game start to ask for some of those cloud power.

 

[-tkf-]

With cloud they use virtual servers, it shouldn't require much to keep the game servers available almost indefinitely since if no-one plays, there's no need to keep a virtual server up, if one starts playing, pop the server up.

For local / single player games, you could have that "use as little resources as possible" when not connected, but when connected wouldn't you rather choose "a living world" instead of "fake something up"?

Raises a great question, will the cloud services require the games to be updated with new code as the cloud gets upgraded with new features, patched for security holes?

With publishers running their own servers and the classic dedicated servers, the code was always in sync.

And asteroids in the cloud sounds like a great tech demo but we still lack a good example that proves the cloud power. And most importantly offers something we havent seen with current games.

 

[Silent_Buddha]

Also their example seems to be non-interactive, which also lowers the value of the demo.

Accurate physics modeling of explosion and debris are pretty much never interactive. Yet they are valuable none-the-less.

Trees with physics modeling are also generally non-interactive yet influence a games feel quite radically.

Hell, a rendered building is generally non-interactive. Furniture in games are generally non-interactive. Yet all of those require real time processing resources from the GPU in order to be rendered on screen.

Those wonderful cars in Grand Turismo? Non-interactive basically.

Regards,
SB

 

[DrJay24]

That's a lot of confounding physics and rendering. I don't see either rendering work for those GT cars nor the underlying driving physic being able to be done on the cloud.Tech demos are made to wow audiences, not illustrate something useful.

I remember Nvidia showing a similar 'all the stars in the Milky way' tech demo to show off the latest GPU last year at GDC. It sure was pretty, but ultimately meant little, they were faking it in part by simplifying the calculation for anything not near the camera, but it was still fun.

 

[Silent_Buddha]

That's a lot of confounding physics and rendering. I don't see either rendering work for those GT cars nor the underlying driving physic being able to be done on the cloud.Tech demos are made to wow audiences, not illustrate something useful.

I remember Nvidia showing a similar 'all the stars in the Milky way' tech demo to show off the latest GPU last year at GDC. It sure was pretty, but ultimately meant little, they were faking it in part by simplifying the calculation for anything not near the camera, but it was still fun.

That wasn't the point of those things.

The point was that something doesn't have to be interactive for it to be impressive and affect the feel or polish of a game. There are plenty of non-interactive things in games that require some amount of processing and would make the game feel barren and incomplete if they weren't there.

Regards,
SB

 

[Le Photographeur]

Question..

How feasible or possible is it to have locally augmentations on games?
Meaning, we move down the cloud to our room

Basically, what I mean is..
Having a second Xbox One at home, both connected through the ethernet connection.
One of these are only used for improving GFX/physics/whatever things.

Having it locally would mean that you dont need to worry about ping/net latency and other things.

Would this be possible?

 

[Arwin]

Question..

How feasible or possible is it to have locally augmentations on games?
Meaning, we move down the cloud to our room

Basically, what I mean is..
Having a second Xbox One at home, both connected through the ethernet connection.
One of these are only used for improving GFX/physics/whatever things.

Having it locally would mean that you dont need to worry about ping/net latency and other things.

Would this be possible?

Yes of course. Sony did some Gran Turismo five demos where four PS4s worked together to create a high resolution (4k) or high framerate (240fps) version of the game.

 

[Silent_Buddha]

Question..

How feasible or possible is it to have locally augmentations on games?
Meaning, we move down the cloud to our room

Basically, what I mean is..
Having a second Xbox One at home, both connected through the ethernet connection.
One of these are only used for improving GFX/physics/whatever things.

Having it locally would mean that you dont need to worry about ping/net latency and other things.

Would this be possible?

It's possible, certainly, but you then run into the same situation that optional things are pretty much never supported outside of a small minority of games on console. And even when supported by games targeted at the optional "thing", that will at some point have support abandoned for it due entirely to it being optional. And when not abandoned, it remains largely a fringe benefit with the majority of games not supporting it (Move for instance), and those that do often times not supporting it well.

In other words, the cloud could have been successful because it was standard. Having a second console do something similar would never take off due to it being optional, and optional in a way that it's highly unlikely that more than a very small minority of owners would have 2x of the same console. Hence, it would receive basically no support outside of the initial proof of concept games. Move could have become an amazing controller supported in the majority of games, IF Sony had made the move to include it in every single PS4 as a standard controller.

So, yes, it could be done. But won't ever be done.

Regards,
SB

 

[Kaotik]

That's a lot of confounding physics and rendering. I don't see either rendering work for those GT cars nor the underlying driving physic being able to be done on the cloud.Tech demos are made to wow audiences, not illustrate something useful.

I remember Nvidia showing a similar 'all the stars in the Milky way' tech demo to show off the latest GPU last year at GDC. It sure was pretty, but ultimately meant little, they were faking it in part by simplifying the calculation for anything not near the camera, but it was still fun.

No, the rendering part wouldn't be done in the cloud, but all the other calculations could be done, including lighting calculations for the rendered objects, it would just push more stress to the GPU, not the CPU which by any standard is PS4/XB1's "weak link" if anything is