• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Avalanche: Microsoft's cloud power advertising "is misleading at best".

That's a bizarre comparison. The availability, speed and capabilities of those multi-core processors were physical constants. No such certainties will ever apply to the cloud. It's like having a component in your console that might or might not even exist at a particular moment in time.
Exactly. Even more than that, the transition to parallel computing was forced upon developers by physical realities (it was no longer possible to greatly increase sequential performance).
Comparatively, a voluntary transition to a more expensive, harder to implement, harder to test and harder to port asynchronous cloud development model which imposes additional restrictions on your customers seems incredibly unlikely to happen.

And that's true even if all the things MS PR promises were possible, which they are not.
 
Were there any actual sane people who thought the magic cloud was going to make xbone games look better or run faster? Like the man said, the main benefits will likely be for multiplayer games.
 
I haven't really been keeping up with these "cloud-things".

Will games that use the cloud work at all if your connection goes down or will single-player games stop working if you lose connection for a few minutes?
 
This is the number one reason I cannot see it being used. Although I suppose if you bought Forza for $60 and agreed to pay $30/month for access to ForzaCloudPowar it could work in terms of economics.

This is the most obvious aspect.

Why would a developer do that? Instead of offloading "trivial" stuff like physics and AI to the cloud, they will just implement a weaker local version that can be actually handled by the console itself.

It's less complicated, it's way cheaper, and "cloud-based physics" is nothing gamers would buy a game for, let alone pay extra money (directly or indirectly).

People pay for cloud services if they have real benefits, like online gameplay, media streaming, sharing, or whatever. But not game physics on a massive multi-billion cloud infrastructure. Come on...
 
That's a bizarre comparison. The availability, speed and capabilities of those multi-core processors were physical constants. No such certainties will ever apply to the cloud. It's like having a component in your console that might or might not even exist at a particular moment in time.

(sorry, missed your post)
It's only a comparison about "needing more work to master a new technology". Of course coding for cloud computing is not the same thing as coding for multi-threading. That's why it is additional work for a developer, even if he has already mastered the latter.
It was an answer to "why would developers bother to learn that new tech". Developers need to do that all the time.
 
Were there any actual sane people who thought the magic cloud was going to make xbone games look better or run faster? Like the man said, the main benefits will likely be for multiplayer games.
I'm going to imagine that most here have no idea and have to rely upon the word of different personalities in the game's press and on forums to spell out whats realistic and whats not.

I'm a web programmer and I have no idea how many "latency insensitive operations" will be in use in the next gen of games, especially games that may be purpose built to leverage this server architecture, much less how many local resources could be freed up for a dev to repurpose towards latency-sensitive operations.

But, I expect to see that explained or debunked come gameplay at e3 where we should see gameplay footage for titles enhanced by the 'infinite power of the the cloud'. I'll be watching.
 
Did MS say 300,000 physical servers? I'm almost sure that number is for virtualized servers.

That's a whole other story than Jonathan Blow had a go at - it could just be 20 physical servers running 300 K tiny virtual servers for all we know.

It's too early to say anything much about the infrastructure without more specifics.

I will say I thought the interview was good and he seemed clear and open and honest in his replies. Not sure why there has been some bashing of the company around quotes - that seems perfectly fair and even handed.
 
About that 300k servers figure... I got the impression that it was for both XBL and Azure...

http://news.xbox.com/2013/06/connected
Future proofed with power from the cloud: Microsoft has created a global network of more than 300,000 Xbox Live and Windows Azure servers, to help creators realize their visions of what is possible with a connected system.

But between all the mixed messages, it's hard to tell.
 
That's a whole other story than Jonathan Blow had a go at - it could just be 20 physical servers running 300 K tiny virtual servers for all we know.
I'm not entirely up on my clouds but I am familiar with virtualization, and a very real risk in virtualization is that one physical host/machine going down can take out several (even dozens) of these virtual servers.

Does the cloud nomenclature mean that this isn't the case? ie. that physical hardware failures instantly migrate the virtual instances to other hardware? Thats how I've been taking the cloud terminology in context- to mean almost zero hardware dependencies through total virtualization.

Am I correct?
 
this "power of the Cloud" this is just a way for consumers not to know what the 24Hr check in is really for on the Xbox One, DRM..
 
Does the cloud nomenclature mean that this isn't the case? ie. that physical hardware failures instantly migrate the virtual instances to other hardware? Thats how I've been taking the cloud terminology in context- to me mean almost zero hardware dependencies through total virtualization.

Am I correct?

If you want to have an instant migration of compute states in cases of failure, you basically have no choice but to run a redundant copy of the server software in parallel.

In traditional cloud-based services, the state is constantly read and written transactionally from a database (or many hybrid database systems). If a service frontend fails, another one can take over and retrieve the state from the persistency infrastructure. This is not possible for games because of the high performance penalty involved. So you have to keep the state in-memory. But in this case you cannot implement a sensible "instant" method of fail-saftey without significant overhead.

I am actually doing a PHD in a research project that engineers technology for cloud-based, time critical controls of "smart" power grids. We basically face exactly this problem there. And the effort necessary to deal with that particular problem is not just the learning of a new technology, its paradigm-shifting.

Of course, you can just live with failures. In this case, the game would just have to handle the complete loss of all cloud-based game states and "reboot" somehow.
 
Finally some sense from Avalanche. Cloud claims from Microsoft are bollocks, but expect to see dedicated server benefits and some low level stuff for show.
 
Were there any actual sane people who thought the magic cloud was going to make xbone games look better or run faster? Like the man said, the main benefits will likely be for multiplayer games.
Well even if the main benefits are for MP it's still a pretty big advantage. If they are actual quality servers I can see MP games getting a preference on X1 despite looking or running worse. I wouldn't mind paying a sub on the PS4 if it meant proper dedi's for all MP games.
 
Of course, you can just live with failures. In this case, the game would just have to handle the complete loss of all cloud-based game states and "reboot" somehow.
Thanks for the response. It makes sense that how you manage state would have a dramatic influence on how you handle failures.

This stuff honestly sounds fascinating.
 

Yes, but Avalanche are currently working on *three* AAA games, with three different publishers.

So just because they're developping for playstation 4 (which has been stated officially on their twitter/facebook page and all) doesn't mean they don't have an Xbone exclusive in the making too.

After all, they haven't shown anything since Just Cause 2, and that was three years ago (well there was Renegade Ops, but that was just a small team).
 
Exactly. Even more than that, the transition to parallel computing was forced upon developers by physical realities (it was no longer possible to greatly increase sequential performance).
Comparatively, a voluntary transition to a more expensive, harder to implement, harder to test and harder to port asynchronous cloud development model which imposes additional restrictions on your customers seems incredibly unlikely to happen.

And that's true even if all the things MS PR promises were possible, which they are not.
The very best part from the Ars Interview:

“Without getting too into the weeds, think about a lighting technique like ambient occlusion that gives you all the cracks and crevices and shadows that happen not just from direct light. There are a number of calculations that have to be done up front, and as the camera moves the effect will change. So when you walk into a room, it might be that for the first second or two the fidelity of the lighting is done by the console, but then, as the cloud catches up with that, the data comes back down to the console and you have incredibly realistic lighting.

Wtf am I even reading here??? Screenspace ambient occlusion is real-time, because it reacts to dynamic lighting!!! Otherwise you could just bake an occlusion map onto the texture and call it a day, no computation required at all!!!!!!!!!!

Unless you like shining a flashlight into a corner and always wait 2 seconds until the cloud has calculated the occlusion for this viewpoint, you better not move that flashlight. Actually, thinking about it, to render occlusion in the cloud, the server would have to run an instance of the game in fullblown wire-mesh mode or he can't even calculate the correct lighting to send back. This is not what any normal server does unless it's in a renderfarm (happy paying for that).
edit: Or you upload the whole wireframe-and-lighting-scene to the server so he can calculate it then which will take how long??

Someone correct me if I got any of the steps wrong.
 
I think this is the first somewhat negative thing he's actually ever said in regards to the platform.

I was going to quote more, but I think every other quote he has given has had a thread from another site.

He seems pretty up on it overall.


Source: http://gamingbolt.com/interview-wit...loud-functionality-pushed-as-a-marketing-tool

One thing that should be summed up is.. there is NO special advantage Microsoft has in cloud computing.

Anything they can do, their competitors like Sony, Nintendo, Apple and Google can for their games.


Also, anything that can be done on XBONE can be done on PS4, Wii U, PS3, 360, PC, etc.
 
Interesting. The MS press conference is becoming more and more interesting to see how they handle all dem rumours.

you already know how they're going to handle it. they're handling it right now by flooding message boards and article comment threads with bullshit talking points like "Steam is just as bad!" and "people complained about needing electricity in 1900, too!"
 
Someone correct me if I got any of the steps wrong.

No, it's that ridicolous. For such things to work, the console and the cloud have to maintain a conversational state. That basically means that there has to be a (quite large) session state permanently managed in the cloud, coupeling your game to a very specific server instance, thereby casually destroying the cloud's main benefit of high efficiency through shared resources.

The main point here is, that no developer would do that. They will just implement a lesser form of visual candy that can be handled by the console itself.
 
One thing that should be summed up is.. there is NO special advantage Microsoft has in cloud computing.

Anything they can do, their competitors like Sony, Nintendo, Apple and Google can for their games.

They can, but will they ? The main advantage Microsoft has right now, is that they already invested in that strategy (which isn't a small investment, like they already said, and other posters here too).
It also helps that they have their own professional cloud service. If anything, they could even let their competitors use their Azure services, if the additional income is worth it.
 
If anything, they could even let their competitors use their Azure services, if the additional income is worth it.

They are already doing that. That is the whole point of Azure. Just go to http://www.windowsazure.com/, give them your credit card, and you're ready to go with as many servers as you like.
 
I know that. I meant for specific game-related applications, like those of the Xbox One.
If they only "rent" them Azure servers, it's no different than what other "cloud providers" would offer.
But there's no technical reason PS4 games couldn't use the X1 implementation of remote processing tasks, so developers would only have to implement it once, using a single SDK.
 
I know that. I meant for specific game-related applications, like those of the Xbox One.
If they only "rent" them Azure servers, it's no different than what other "cloud providers" would offer.

And what is the difference and from which source do you know that? Or put in other words...

But there's no technical reason PS4 games couldn't use the X1 implementation of remote processing tasks, so developers would only have to implement it once, using a single SDK.

...what exactly is a "remote processing task", how does it work, what makes it different from any other call-request RPC between server and client, and where do you got the information from that it is part of some SDK?
 
And what is the difference and from which source do you know that? Or put in other words...



...what exactly is a "remote processing task", how does it work, what makes it different from any other call-request RPC between server and client, and where do you got the information from that it is part of some SDK?

I don't have any information on that, but it seems logical to think that if MS want developers to use their cloud features in their games, they will include it in their SDK, and have standard API to include common "cloud features" more easily. I would expect it to support at least X1 and PC, so why not other consoles.
It would be much required, since it is a difficult task that developers have to learn and any help in doing so would be welcome. And that's what MS does anyway : most of their activity is developing software environments, and tools to use them.
 
For the consumer? Nothing.

For companies, it makes servers cheaper to host. That's the long and short of it.

Ironically, a transition to elastic, commoditized cloud platforms from specialized, dedicated game servers was not only inevitable, but already happening. MS is just the first to spin it as a marketing advantage.
 
Man I can already see the BS comparisons coming in their E3 conf if they "demonstrate" that stuff.

"On the left, our game running on XB1 without the cloud. On the right, the same game running on a PC with 32GB RAM and 4 tit- I mean XB1 with the power of the cloud"
 
I know that. I meant for specific game-related applications, like those of the Xbox One.
If they only "rent" them Azure servers, it's no different than what other "cloud providers" would offer.
But there's no technical reason PS4 games couldn't use the X1 implementation of remote processing tasks, so developers would only have to implement it once, using a single SDK.
Yes, besides it being completely smoke and mirrors and having a massive overheadin the GPU compared to the X1 which is a GPGPU so it can easily handle any hypothetical task that the X1 offloads into the cloud. AI, Physics?? That's exactly the forte of a GPU besides graphics.

Just accept it, the "infinite power of the cloud" is Microsoft's attempt to hide that their system is simply de-facto weaker by quite a margin.

Normally a company will simply play it's strengths but Microsoft seems really unsure what those are... we'll see what games E3 brings.
 
I don't have any information on that, but it seems logical to think that if MS want developers to use their cloud features in their games, they will include it in their SDK, and have standard API to include common "cloud features" more easily.

So you are basically just wishing that all this stuff will work because Microsoft told us so, and you can't really explain how Microsoft wants to overcome all the economical and technological absurdities of this concept?

They are lying to you. They are sugarcoating their performance deficit. They are literally providing nothing that anyone couldn't have done years ago. Cloud-IaaS has been arround for years. Architecture and communication patterns for distributed systems have been arround for decades. If there would be any sense in this, somebody would have already done it. But nobody has done something even remotly similar to partially offloading game calculations. I am not even aware of any research paper outlining something like this. Because its BS. It does not work. It has a ridicolous ratio between benefit and complexity/costs. And Microsoft won't pull any magic out of their buttocks to suddenly make it a rational thing.

Cloud gaming might be a thing in a few years, but only for games completly running in the cloud, streaming audio/video to heterogeneous, light-weight clients. But a hybrid approach that uses an expensive client (= a console with expensive dedicated hardware), a distributed cloud infrastructure for pure computation and goes into all the software engineering hell that comes with it, is ridicolous. It combines the worst of both worlds.

The cloud will be used for online-multiplayer, sharing, media streaming, digital content distribution, data synchronization/backup. That's what it can do well and efficiently. Nothing else.
 
Man I can already see the BS comparisons coming in their E3 conf if they "demonstrate" that stuff.

Yes. I expect tons of bullshit "simulations" of what a game will be like with the "Infinite Power of the Clowd". And no one will be able to gainsay them until this thing is out and it runs like dirt in the wild.
 
A nice balanced interview there and good read.

I think this marketing BS is really bad, I mean quotes like this...

It's also been stated that the Xbox One is ten times more powerful than the Xbox 360, so we're effectively 40 times greater than the Xbox 360 in terms of processing capabilities [using the cloud]. If you look to the cloud as something that is no doubt going to evolve and grow over time, it really spells out that there's no limit to where the processing power of Xbox One can go.

Source

Are so bad, the benefits are going to be negligible at best IMO. 40x the power? No limits? Please shut up.

Also, is anyone expecting this to be available to silver account holders (if silver still exists)? I'd be surprised if they are going to be doing all this for free, so yay lets pay to play single player games! Granted they haven't mentioned this at all yet, so I might be being disingenuous on this point.
 
So you are basically just wishing that all this stuff will work because Microsoft told us so, and you can't really explain how Microsoft wants to overcome all the economical and technological absurdities of this concept?

Er, no, that's not at all what I'm saying.

First, I don't think the concept itself is a "technological absurdity". There are things that can obviously be done (high latency, high processing of small data). Others that obviously cannot (low latency of heavy data). And many other things that we don't know. Based on that, I have no reason to doubt that MS is really trying to do some remote computing when they say so. And if they do, it is only expected that they develop specific tools for that.

Second, the economical part is obviously something important that we haven't discussed yet. Someone has obviously to pay for those features, but I would also think that MS has thought that through, since hosting and selling cloud servers is a big part of their activity, and something on which they intend to expand.
There are too many unknowns there to speculate on the price of that (we don't even know what kind of tasks will be running on the servers, and how much data/resources they require). Maybe they'll have the publishers pay. Maybe they made a deal with them in exchange of other benefits. Maybe cloud features will be in a XBL Platinum package. Who knows ?

They are lying to you. They are sugarcoating their performance deficit. They are literally providing nothing that anyone couldn't have done years ago. Cloud-IaaS has been arround for years. Architecture and communication patterns for distributed systems have been arround for decades. If there would be any sense in this, somebody would have already done it. But nobody has done something even remotly similar to partially offloading game calculations. I am not even aware of any research paper outlining something like this. Because its BS. It does not work. It has a ridicolous ratio between benefit and complexity/costs. And Microsoft won't pull any magic out of their buttocks to suddenly make it a rational thing.

Your speech reminds me of one I had not too long ago. After 6 years experience working in computer vision at the time, I wouldn't believe the rumor that MS would build and sell a low cost depth camera, and do real time full body tracking for gaming applications. Decent depth cameras were as expensive as cars at the time, and there was no published tracking algorithm that could do what was required, especially on the specs of a X360. I did call that bullshit, and economically/technologically absurd. And I was wrong.
 
First, I don't think the concept itself is a "technological absurdity".

Ok, just a simple question then: why has Microsoft invested in a massive cloud infrastructure, that supposedly is (to the most part, "3 of these things in the cloud") dedicated to a lot(!) of very specific game calculations, that could also be run locally on a more powerful console, and why have they commited themselves to massive ongoing operation costs, instead of just building a more powerful console, that could do all these calculations (as well as latency/bandwidth critical ones) in a much, much more simpler way and without ongoing infrastructure costs?
 
Top Bottom