At this year's GDC, there's one topic ruling the conversation; streaming services, and the arrival of Google as a player in what's now become a very tangible space. How long ago it now seems from the comically messy launch of OnLive, or from when Microsoft's 'power of the cloud' mantra was a stick used to beat it during the difficult early years of the Xbox One.
Now it's no joke, and Google's imminent keynote where it lays out its plans for its own service feels like a pivotal moment in a streaming future that now seems inevitable. Microsoft has already made its own motions, of course, with the announcement of its own xCloud service last year, headed up by corporate VP of gaming cloud Kareem Choudhry, who we got to catch up with on the eve of Google's big show.
"In the gaming division, our goal is to reach everyone on the planet, to enable them to play the games they want when they want on the devices they want," he tells us in a swish Microsoft office in downtown San Francisco, just a block away from where the Google keynote will be taking place. "Our strategy is shifting from one that's console-centric, where step one is please buy our console and steps two through to 58 are things accessed off the console. We're starting to put the customer at the centre of what we do, and recognise that they have multiple devices, multiple lifestyles, then bring it all together in a cohesive way."
Microsoft's approach, Choudhry says, is focussed on three fronts; content, community and cloud. Community you know from the likes of Mixer and Xbox Live, while content is something that's being taken very seriously indeed - a year ago Microsoft boasted some six first-party studios, but after some canny movements and acquisitions that's more than doubled to 13. The cloud aspect, meanwhile, is something that's really beginning to take shape. We spoke about some of the challenges and some of the solutions.
Kareem Choudhry: Why's Microsoft getting into streaming? In terms of why right now, you get this confluence of technology, consumer expectation, the device agnostic lifestyle that we're starting to lead, the rollout of 5G that's coming - that's an important element, though in terms of our own plans that's not a requirement. I always like to cover some of the hard problems. Music streaming was around the launch of 3G, video was what kicked off with 4G and there's a general belief that game streaming is what will be the killer workload for 5G. Video is non-interactive, and gaming by its nature is interactive. I've spent two decades thinking about the game loop, all the way from input to TV display latency to 30fps/60fps/120fps.
How's it going with your solution to latency here? It's still an issue with many services we've seen already, so what kind of latency will we see on xCloud, and how are you going about minimising that?
Kareem Choudhry: From the data centres we have near Washington we're seeing really good latency - less than 10 milliseconds that's being added by the traversal to the cloud. Frankly we find more latency in the Bluetooth stack, connected to an Android phone. Part of what I get so excited about, having Azure as our first party cloud, is that I'm able to place hardware, regionally, as close to users as possible. When I talk to game developers, latency is something the industry has been dealing with for a long time, but they've always dealt with it from the multiplayer side. When I go and talk to creative game developers, especiallly in first party, is that they're going to take all those latency mitigation techniques that we apply on the multiplayer side and we're going to put them on the enduser input side. I think it's going to work out okay.
The other thing I've noticed is that consistency of latency is much more important than absolute latency, when it's very spiky. The human brain is actually an excellent latency compensator, and if we can use that to our advantage I think that'll be good.
What kind of connections will you be requiring for this?
Kareem Choudhry: We hope to get down to single digit mbps. I think some of the demos we've shown so far have probably gone down to nine, 10mbps. Some of the work that we're doing with Microsoft research, I think we'll be able to get a really good video feed probably around six to five.
At the moment you've got Xbox One S consoles in the data centre - how exactly does that work?
Kareem Choudhry: That's really a mechanical question - version one is go buy a console, take the plastic off and go put it in a data centre. That's what got us up and running, and obviously that happened quite some time ago. If you remember - let me back up, a data centre operating environment is very different to a home environment, and when we design a console for the home environment there's a bunch of things we take into consideration - power consumption, the acoustics, I spent a lot of years of my life recording the decibels of the fan from 10 feet away.
As someone who's got an Xbox One X in their living room, next to another console that sounds like a jet engine taking off, I appreciate the work.
Kareem Choudhry: The data centre is quite different - there are different challenges like supportability, managabilty, blade design. We're now at a place where we've redesigned the Xbox One S motherboard with specific server scenarios in mind. Now we have eight of them packed into a 2U blade, which slides in, and we're leveraging all the hardware design and data centre design that the Azure organisation has built up over the years. It's *largely* the same thing. For a couple of reasons - one, the content doesn't know it's in a data centre. I'm taking the exact same content - much like we did with the backwards compatibility program - the hardware has to be pretty precise so the game operates properly. It's making changes to make it more server suitable, improve the running economics, that kind of stuff. But it's no longer a consumer facing product, so to speak.
The other question would be why not Xbox One X consoles in there? Is that just because this is an early iteration?
Kareem Choudhry: I'd say a couple of things - one, we have a roadmap of what we're doing on the hardware side. But we just decided to start with Xbox One S for a couple of reasons. One was our first target being Android phones - we could send a 4K stream there, but probably 1080p, no-one's going to tell the difference. That was an easier place to get started, and frankly we wanted all our Xbox One X chips to go to consumers like yourself.
Mobile's a target for this. A basic usability question arises around the UIs. I remember when there was the division between SDTVs and HDTVs, with interfaces being almost illegible, which is the same now when trying to play these games designed for bigger TVs on a mobile screen. What ways do you have around that?
Kareem Choudhry: There's a couple of different things - TV safe zones, HD versus SD, we've been playing in that space for a long, long time. The other thing we're looking at is just crispness of text from an encode and a decode perspective. We've got Microsoft research actually looking at that - one of the advantages that we have is we know to some degree what's happening in the stream. We're not using generic encoders, we can use an encoding strategy that can even be custom tailored using machine learning to the actual specific game. We can do things like say, we know the HUD is right here, so do something different over here to get actual crispness.
In terms of a game having a specific text size - we're not going to be able to change that without cracking the game open. And then it gets a little bit more to our forward looking strategy. The first step is lifting and shifting content that works on our consoles today into the cloud scenarios. The next stage of that evolution is something we call cloud aware, where we'll actually inform game developers at a runtime perspective, saying hey you're running in a cloud scenario and they can choose to do different things to make things better for their users if they want.
Is the idea to have multiplatform stuff on there eventually too?
Kareem Choudhry: Obviously we're very close with all the different publishers and developers - our approach is that we're building a technology - and it's like the backward compatibility programme - I built a platform and it's in a way that all the content could come and participate. It's up to the IP owners, they'll be the ones who decide when and how it gets exposed.
There's been a lot of excitement about the potential on Switch. What would it take to get xCloud working on Switch?
Kareem Choudhry: Well I've never developed on the Switch, for obvious reasons! So I'm not a good person to answer that question, but from an aspiration perspective, really I want to enable content on any device that has a screen and an input. We're starting on Android phones, but you could extrapolate where else we could go next - mobile phones is an obvious starting point. We spend a lot of time thinking about things like smart TVs, and other inputs.
How do services like GamePass sit alongside xCloud?
Kareem Choudhry: Obviously GamePass is our subscription service, it's really where we want to deliver great experiences and value to our customers. And right now it's a console product - we have aspirations to bring it to more users and more places. Project xCloud and GamePass are going to co-exist in some reasonable way.
A broader philosophical question - some people see cloud gaming as the death of console, or the start of it. Is it that extreme?
Kareem Choudhry: It's not that extreme at all! I've been building consoles for 15 years - we're not getting out of the console business, we spoke about it in our xCloud videos. We love our consoles, we love that business, and we're super proud to have the most powerful console out on the market place today and that leadership position we hope to retain going foward. And I also believe your best premium experience is going to be dedicated hardware running under your TV in your living room. It's an 'and' conversation, not an 'or' conversation. Everyone loves to jump to the death of consoles, and I think it makes a great headline, but we don't think that way at all.