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

Yeah but your figures may not be representative. You deal in soft porn right? Mac owners are more likely to be masturbating over pictures of themselves taken at Starbucks ... [rimshot]

You *almost* got me to spit diet coke out of my nose lol, that was a close one :)
 
You *almost* got me to spit diet coke out of my nose lol, that was a close one :)

Aren't mac owners more likely to be 'using' ios devices?

Mac OS X is a small dot on the personal computer radar. Except when you are around the right people then they think it's everywhere and the norm... :)
 
Is the multiples consoles something like "cloud" processing?
I was going to write it's simple network processing, but it's not even that. Each console is a standalone renderer driving its own screen. You just have synchronisation between them. Network traffic is minuscule.
 
Well, I've just started 70-486, and will get another one related to Azure a little later ... Will let you know if I learn anything interesting
 
For those people who have a couple of pc's networked this opens the door to games running with distributed computing

I wonder if local distributed computing is the motivation for interfacing the ethernet directly to the main SOC instead of through the southbridge.
 
I was going to write it's simple network processing, but it's not even that. Each console is a standalone renderer driving its own screen. You just have synchronisation between them. Network traffic is minuscule.

Which in a sense is making a good case for that distributed processing can work really well :D
 
Long time lurker, lost my old login.

Could the processing ala Cloud also come theoretically from a pc on a local network?
 
Yes, we even discuss the idea from further up on [strike] this very page[/strike] the previous page (Patsu kicks it off post #1033). The idea is perfectly sound the challenge is the wide variance in compute resources (available CPU & RAM), whether such resources exist at all (with more and more homes using ARM devices for general computing) and the support issues with supporting the remote client on such a computer (wide range of legacy O/S, crappy LANs, etc, etc).

What makes MS Cloud plans interesting is that they are taking most of these headaches away and offering fixed set of compute resources and a consistent environment (I presume it's Azure) what they trade for that is the one advantage LAN resources usually enjoy, latency

Edit (dammit hit Tab and posted before finishing my point)
 
Last edited by a moderator:
Yes, we even discuss the idea from further up on [strike] this very page[/strike] the previous page (Patsu kicks it off post #1033). The idea is perfectly sound the challenge is the wide variance in compute resources (available CPU & RAM), whether such resources exist at all (with more and more homes using ARM devices for general computing) and the support issues with supporting the remote client on such a computer (wide range of legacy O/S, crappy LANs, etc, etc).

What makes MS Cloud plans interesting is that they are taking most of these headaches away and offering fixed set of compute resources and a consistent environment (I presume it's Azure) what they trade for that is the one advantage LAN resources usually enjoy, latency

Edit (dammit hit Tab and posted before finishing my point)


I remember Microsoft briefly speaking about it and won't dive further because I cannot find the discussion blogged. They did however say that multiple windows machines over a network could offload heavy processing duties etc etc. It's been years since the whole Cloud discussion, everything just leaves me curious. The architecture, their plans. Before the Xbox One, there were heavy talks about multiple devices, home PC's, STB's, Tabs, Laptops any unified devices sharing workloads.
 
I didn't mean to imply from my post that it wasn't possible or even too difficult it's just seems impracticable for a mass market product like a video game.

For example lets say I design my game for 200 MegaFlops of remote LAN computing resources and I choose to dedicate those to say global lightmap calculations so my whole world map has nice shadows that follow the sun all day (lets not get hung up on what I use the 200MF for). This works great in the lab and represents a real step up in Iq compared to the local shadowing we have now. The kick is that I also have to design my game with a local shadowing engine for those scenarios where that 200 MFlop are not present or risk alienating a portion of my user base. Thus I wind up twice the work my for shadowing engine if not more as now I have to worry about network latency, sudden loss of the resources (your brother fires up BF4 and kills the remote client), viruses on the remote host eating cycles and all sorts of other issues.

What MS have done is eliminated almost all of those issues by saying here is X Flops and Y RAM available on demand over the network. Azure data centres are already positioned globally to provide low latency (by WAN standards) connections. Now I as a develop I can know that those resources are there. The big downside of course and the one that has tripped up most scenarios discussed in this thread is the latency over that WAN link and the nature of the internet and it's fluctuating performance. That's why I think we're all awaiting some games that actually leverage this tech in an interesting way so we can get a feel for what developers think it can do.

Thus far the only implementation we've heard of is the Drivatar tech from Turn 10 which could have been achieved off line as it's basically telemetry analysis. Of course the advantage of the cloud is that you don't have to make the player wait X seconds while the data is crunched locally it can just be sent as a job, while the player plays on.
 
I am doing various MS training courses, and the stuff that I read so far all warned that programming against Azure should be aware of high latency and performance fluctuations. And try to work as stateless as possible. But at the same time highly scalable (two sides to the coin)

State management is a big bottleneck when you don't know for sure that the logic and state is on one server together. So I am guessing that the suggested available resources by MS are such that they can be provided by a single server.
 
Thought you guys might be interested in this :)

Ryse’s Design Director: Xbox One Can Pull Off 10,000 AIs At The Same Time Because of Cloud Computing
/
That's what the title says. The words from Esteves are nothing like. He's talking about a hypothetical 10,000 AIs in the next game. You only need look at the game to see there aren't 10,000 AIs!

Read the original interview and you get a clearer picture. I don't think Ryse is using the cloud at all.
 
His statements came in two different answers.

The first claim was that, should a technical resource like that be available, they would use it, because pushing the tech to deliver new experiences is what they do.

The second was that it was technically possible to do 10,000 AI, for some definition of AI.
The resources he outlined in his hypothetical may in certain aspects exceed the current cloud allocation for Xbox One, however.
He further clarified that is was more of a general projection of something he could see happening during the very long console generation, by somebody, for some future project.
 
Back
Top