Xbox Scarlet Hybrid Game-Streaming Version, Kahawai

Don't they have 2 different streaming methods.
1 no dedicated box, can stream to everything including phones
2 dedicated hardware to give a higher quality streaming experience, e.g. lower latency

I think your thinking about number 1?

I think it will be an exceptionally confusing product line and I have to wonder how many third parties are going to invest the engineering resources to make their streamed games work any better on the hybrid device than the streaming only version. MS is going to try and sell people on this more expensive streaming box for "reasons", but will anyone but MS even employ those "reasons"? How many people will find out they paid extra for features barely any games use? And then there are going to be extra, extra payers for a 1080p local version of Scarlet, and an extra, extra, extra version for the premium 4K Scarlet box?
 
There's only one steaming product, number 2. Number 1 is streaming to your phone and stuff. Games won't need to target the steaming box as it'll be handled by MS through the system libraries, I'd expect; they'd know devs wouldn't put in the effort to target it themselves. So devs will build the game for Xbox, maybe with a bunch of settings for different tiers, Xbox covering PC as well, and when this executable is run on the Xbox Streamer, it'll be run in lowest quality while streaming higher quality from the net. The same executable runs on the Xbox 2S at 1080p and on Xbox 2X at 4K.

Or simply, targeting the different devices should be no more hassle than targeting the range of PCs out there. MS have two generations of console library advance, plus their VM experience, plus their extensive BC experience, plus DX 12, so they should definitely know how to put together console-performant libraries that'll scale across their devices.
 
Games won't need to target the steaming box as it'll be handled by MS through the system libraries, I'd expect; they'd know devs wouldn't put in the effort to target it themselves.

I would posit that any "hybrid streaming" solution that is transparent to game developers probably has negligible benefit. It will just be "power of the cloud 2.0" with MS trying to sell people on a fantasy they never manage to demonstrate in practice.
 
I would posit that any "hybrid streaming" solution that is transparent to game developers probably has negligible benefit. It will just be "power of the cloud 2.0" with MS trying to sell people on a fantasy they never manage to demonstrate in practice.
Can I ask why? There should be a little more meat to your post if you're going to just hand wave off everything Shifty just wrote. Nothing he wrote is absurd or of the realms of reality for MS. They spend $1 Billion per month on cloud infrastructure alone, if you're telling me they are incapable of this, and we're being sold snake oil your response has got to be more substantial than just name dropping cloud 2.0 when Mattrick and Balmer were heads of MS.
 
I don't put any faith in anyone promising an automagical solution to a complex problem. They almost always prove to be neither. To be blunt, I don't see how anyone could handwave away the idea of game logic running in two different locations, as much as 200ms round trip apart from eachother as something a game developer will not have to worry about thanks to unspecified "reasons".
 
MS recently demo'd two streaming technologies that were discussed (I was suitably critical of real world viability). However, that's dependent on the nature of the network, and doesn't affect how devs would have to work.

I agree though that I didn't really factor in the networking sync aspect which couldn't be automagical, only thinking about the graphical updates. It ought to be automagical for online enabled games at least, but single player titles might need to be written as online titles, perhaps limiting which games can be streamed in low-latency mode?
 
Last edited:
I posit Crackdown 4 will be the poster child of MS hybrid streaming system, announced with the console, to be released 8 years after. Without streaming.
 
I don't put any faith in anyone promising an automagical solution to a complex problem. They almost always prove to be neither. To be blunt, I don't see how anyone could handwave away the idea of game logic running in two different locations, as much as 200ms round trip apart from eachother as something a game developer will not have to worry about thanks to unspecified "reasons"
I don’t think anyone does. But there is precedent for it which is why it’s not being nearly questioned as much.
I’m not entirely sure which topics you’re trying to address though:
Are you trying to suggest that developer interaction is required ? Because if so Geforce Now does exactly this, your steam account online, a server downloads the game titles of your choice onto that local cloud server and streams it to your client. No developer interaction required and it’s really good ping.

Google has similar performance ratings, I believe DF had the latency down to an added 30ms.

There is a latency issue without a doubt but these issues are solved by bringing the servers closer to last mile. And that comes with time, and they have have time to build out their infrastructure, I don’t recall this service being announced for worldwide use tomorrow. The difference between MS and other companies is that they already are building closer to the last mile because it runs parallel to their azure plans anyway. It’s not like they are investing 12B annually in gaming, they are investing 12B annually in Azure and gaming is taking advantage of the infrastructure being built.

So the precedent exists. It would not be the first time that MS took a complex problem and solved it without developer interaction; see 4K BC 360/OG titles on X1X.

As for cloud powered 2.0. The technology, experience and learning is much more important than the product itself. MS greatest lost here is the acquisition of cloudgine over to Epic. They should have acquired them and continued building out the tech. They did restart and build something, but it’s a regrettable loss of time. Considering where they are going with streaming, having cloud assisted processing to support titles that get streamed to mobile devices anyway seems to work hand in hand.

Crackdown will likely not set the world on fire with or without cloud processing. But the underlying technology and the team that makes it is worth significantly more than the product that they created. If there are whole teams dedicated to graphics engines, having a whole team dedicated to this type of thing is every bit as important looking forward.

We will see more cloud 2.0 in the future especially as the future gets more connected.
 
That's simple streaming. MS's advanced streaming involves running the game in simplified form locally and syncing with the game running on the cloud to supply improved visuals. This requires the local game and cloud game to be synchronised, which requires communication between the two akin to a standard multiplayer client<>server model. If your game isn't developed for that, how can you synchronise game state across the internet latency, especially for when there are significant peaks of latency? 100 ms could result in a significant difference between server and client state if all you're doing is forwarding client actions to the server.

Is there a way MS could automate that process through libraries so it's effortless for devs to incorporate, or is it going to require specific labours by developers to support Advanced Streaming needed in the advanced streaming box, versus simple streaming on any device that simple runs the game locally on the cloud and steams visuals?
 
MS's advanced streaming involves running the game in simplified form locally and syncing with the game running on the cloud to supply improved visuals.
What are we discussing here? Project X Cloud or something else, the else being what's showing up in Crackdown.

If the latter I believe it's work in progress. And it would be a long time for it be minimal developer interaction. That would be one hell of a setup, and they must have discovered some seriously canned solutions for that to happen.
 
Last edited:
Jay described two streaming methods, the first being streaming to any device including mobiles and the second one being the latest examples from MS, Kahawai (first post this thread) which does local processing.

bgroovy sees Kahawai as being a SKU that needs to be specifically targeted.

Does MS's idea of a streaming box with local compute doing things like physics computations require developers to target it, or could it be handled at a system/API level and not require specific work? If the former, will devs target it or will it be a box with a limited library for its local compute capabilities?
 
Jay described two streaming methods, the first being streaming to any device including mobiles and the second one being the latest examples from MS, Kahawai (first post this thread) which does local processing.

bgroovy sees Kahawai as being a SKU that needs to be specifically targeted.

Does MS's idea of a streaming box with local compute doing things like physics computations require developers to target it, or could it be handled at a system/API level and not require specific work? If the former, will devs target it or will it be a box with a limited library for its local compute capabilities?
hmm shoot.. this is where journalism isn't helping us out. I thought Kawahai was Project X Cloud.
Yea I mean, if it's hybrid local processing ala Crackdown, yea I can't see it without being specifically targeted.
 
hmm shoot.. this is where journalism isn't helping us out. I thought Kawahai was Project X Cloud.
Yea I mean, if it's hybrid local processing ala Crackdown, yea I can't see it without being specifically targeted.

Kawahai isn't anything like Crackdown's server augmented physics.

A full version of the game runs locally at low graphics settings. The server runs the low detail and high detail version. The difference between the two frames is sent to the client and blended for the final output. In it's demoed form, it's a way of saving significant bandwidth over having only server side streaming.

I'd wondered if you could us it to offer a 1080p local box that could be augmented to 4k via streaming. If it misses an augmented frame to improve latency it might not even be that noticeable a drop in quality.
 
Kawahai isn't anything like Crackdown's server augmented physics.

A full version of the game runs locally at low graphics settings. The server runs the low detail and high detail version. The difference between the two frames is sent to the client and blended for the final output. In it's demoed form, it's a way of saving significant bandwidth over having only server side streaming.

I'd wondered if you could us it to offer a 1080p local box that could be augmented to 4k via streaming. If it misses an augmented frame to improve latency it might not even be that noticeable a drop in quality.
oh jeez, i'm really fallen off the wagon here. I'm going to look around for source links. I saw this in the past, I think with a Doom demo, but I didn't think they were trying to make it a thing.

edit: LOL. Right you posted this on the first page. hahaha
 
I'm not convinced that Kawahai is part of Project Xcloud. Look at the official announcement on Xcloud...

Developers and researchers at Microsoft Research are creating ways to combat latency through advances in networking topology, and video encoding and decoding. Project xCloud will have the capability to make game streaming possible on 4G networks and will dynamically scale to push against the outer limits of what’s possible on 5G networks as they roll out globally. Currently, the test experience is running at 10 megabits per second. Our goal is to deliver high-quality experiences at the lowest possible bitrate that work across the widest possible networks, taking into consideration the uniqueness of every device and network.
https://blogs.microsoft.com/blog/2018/10/08/project-xcloud-gaming-with-you-at-the-center/

Microsoft insider Mary Jo Foley had this to say...

Update: As "The Walking Cat" on Twitter and a couple of others have noted, Microsoft possibly could be using the Microsoft Research "Project Kahawai" technology to enable high quality gaming using CPU offload techniques. Kahawai is a hybrid, collaborative rendering technology designed to help save bandwidth.

https://www.zdnet.com/article/micro...o-encoding-for-project-xcloud-game-streaming/

She says "could be using". Remember Microsoft Research does a lot of cool projects that never see the light of day. Remember illumiroom? Microsoft made have learned some cool things with Kawahai, but I'm not convinced it's part of Xcloud.

Tommy McClain
 
hmm shoot.. this is where journalism isn't helping us out. I thought Kawahai was Project X Cloud.
Yea I mean, if it's hybrid local processing ala Crackdown, yea I can't see it without being specifically targeted.

Kawahai isn't anything like Crackdown's server augmented physics.

A full version of the game runs locally at low graphics settings. The server runs the low detail and high detail version. The difference between the two frames is sent to the client and blended for the final output. In it's demoed form, it's a way of saving significant bandwidth over having only server side streaming.

I'd wondered if you could us it to offer a 1080p local box that could be augmented to 4k via streaming. If it misses an augmented frame to improve latency it might not even be that noticeable a drop in quality.

I'm not convinced that Kawahai is part of Project Xcloud. Look at the official announcement on Xcloud...


https://blogs.microsoft.com/blog/2018/10/08/project-xcloud-gaming-with-you-at-the-center/

Microsoft insider Mary Jo Foley had this to say...



https://www.zdnet.com/article/micro...o-encoding-for-project-xcloud-game-streaming/

She says "could be using". Remember Microsoft Research does a lot of cool projects that never see the light of day. Remember illumiroom? Microsoft made have learned some cool things with Kawahai, but I'm not convinced it's part of Xcloud.

Tommy McClain

The thing that interests me about Kawahai is that it would fit the "more expensive" streaming box story. Anyone have some good alternatives? (Other than the rumours being bogus!)
 
There's a good chance that project xcloud is the standard streaming to any device based on 1S hardware.

The streaming based on next gen hardware will allow local support. I don't remember MS saying anything about dedicated hardware for xcloud, just that it can stream everywhere.

There was 2 technology demo'd in the past that will probably come into play, one didn't need any code changes, the other minor.
That was a while ago, I'm sure its been improved and added to.
The developer wouldn't need to put much work in I suspect apart from giving it consideration during development to get good local processing and normal offline running.
All the heavy lifting is done in the libraries. But without use or due consideration the studio would be limiting their games performance.
 
Last edited:
How do you account for the synchronisation issues between local client and server?
I don't remember the specifics of the 2 projects, but I assume that it's issues like that which stopped it coming to market sooner.
I believe one project was based around predictive algorithms, so maybe now with ML at scale it's become viable, or using ML means they can cut down on all the different variations that needed to be processed.
It may be using either, both, spin off, totally different tech, but it does seem to be 2 different streaming methods they have planned.
Questions like syncronisation and possible solutions is the interesting tech conversation, which I don't have answers too, maybe I should go back and watch those vids, see if any hints in them.
 
Back
Top