Xbox Scarlet Hybrid Game-Streaming Version, Kahawai

That's a lot of assumption there, I've no clue what this streaming box will have and at what price points to make the service good enough to warrant consumer interest.

Based on the tech and article that's broadly what it'd need to look like. You could scale the CPU back a bit from the full fat machine I suppose, but still needs to run the game, just not render much of it. In theory they could go as far as only calculating geometry locally, but that would up the streaming bandwidth requirement.

Having an xbone (not X) equivelent GPU will be fairly cheap on 7nm.

I think this too and the streamer (and older consoles) hedges on the higher priced machine. The X used a 16 nm Jaguar CPU and they still couldn't make it less than $500. If Scarlet has 2X Navi GPU, 2X RAM (though maybe not even that much), and a modern Zen +/2 CPU, it could be a feat to make it at $400 in 2020.

Cost wise you might end up in this weird place where the PS5 is straddle by the high and low end Xboxs.
 
There's a big risk of messing up the marketing with that approach and confusing the consumer. I think it would be best for them to release the traditional console first and when that's settled bring out the streaming version.
 
This seems to me to sound like how the cloud would be so great (Crackdown demo) all those years ago. Maybe whilst working on Crackdown MS came up with this idea?

Personally I don't think streaming games is for next gen, this will be MS dipping their toe and seeing how it's taken/works in the real world, Sony's solution was pretty good but not popular.
 
Everyone who trusts their ISP not to take advantage of this, post net neutrality, please raise your hand?

Oh, it is latency dependent? Hmmm.. I think we can charge you more to make that possible. Want it in HD? Another charge. Oh, and we are adding a bandwidth cap to these low latency bits.
 
Saw this:

Sources familiar with Microsoft’s plans tell The Verge that Microsoft is currently “all hands” on creating datacenters capable of powering the company’s game streaming service. Referred to as codename “XCloud” internally, Microsoft has been experimenting with combining four lots of custom Xbox consoles into a single server blade for its datacenters. These servers will launch initially with developers in mind to build and develop games in the cloud instead of local debug machines, and then to stream games to consumers.

https://www.theverge.com/2018/7/24/...scarlett-xcloud-game-streaming-service-rumors
 
Based on the tech and article that's broadly what it'd need to look like. You could scale the CPU back a bit from the full fat machine I suppose, but still needs to run the game, just not render much of it. In theory they could go as far as only calculating geometry locally, but that would up the streaming bandwidth requirement.

Having an xbone (not X) equivelent GPU will be fairly cheap on 7nm.



Cost wise you might end up in this weird place where the PS5 is straddle by the high and low end Xboxs.
Who knows if PS5 won't be $500 if they're targeting similar specs? We know even less about it or even if they'll have something by 2020.
 
If it does, why would they need to create a new streaming box?

The same reason there's a Wii. It's not like wiimotes weren't tested on GameCube, and the same hacks work across both systems because it's essentially the same hardware.

Also didn't Spencer say console quality 720p streaming?
 
Microsoft hasn't even managed to ship their original power of the cloud showpiece, so you'd have to have your nose pretty far up their butt to put blind faith in some miracle hybrid solution. It doesn't matter how you move the workload around, there is always a round trip between the client and the server. This sounds like a classic case of an overengineered solution that doesn't actually address the core problems of infrastructure and proximity. Solve that and thin client video streaming like PS Now will work great with cheaper hardware and no complex programmatical solutions.
 
Last edited:
It doesn't matter how you move the workload around, there is always a round trip between the client and the server.
One's gotta agree. They didn't deliver on the Power of the Cloud, 4x the power for every console, etc. And every company makes claims they've solved the network problems including OnLive. Heck, every new tech or idea comes with promises of perfection that rarely pan out in the real world. eg. OLED solves all the problems of LCD - maybe true, but it comes with its own problems.

Until they show something that works in the home, they haven't got anything to get excited about. The laws of physics and network communications still apply. And as I say, the results in that YT video were far from great, although they could easily improve that by allowing higher BW and probably refining the algorithm. I feel an ID buffer would be useful.
 
Microsoft is in a good position with Azure to deliver on streaming. They going all in with the cloud and it's paying off big time.
 
One's gotta agree. They didn't deliver on the Power of the Cloud, 4x the power for every console, etc. And every company makes claims they've solved the network problems including OnLive. Heck, every new tech or idea comes with promises of perfection that rarely pan out in the real world. eg. OLED solves all the problems of LCD - maybe true, but it comes with its own problems.

Until they show something that works in the home, they haven't got anything to get excited about. The laws of physics and network communications still apply. And as I say, the results in that YT video were far from great, although they could easily improve that by allowing higher BW and probably refining the algorithm. I feel an ID buffer would be useful.

That YT video was using Doom 3 to showcase the tech...says it all.

I'm still waiting to play Minecraft with hololens like presented at E3 3 years ago...
 
*ahem* Remember, this is a technical thread, so keep all your snark of business decisions and marketting out of it.
 
That YT video was using Doom 3 to showcase the tech...says it all.

I'm still waiting to play Minecraft with hololens like presented at E3 3 years ago...

It depends on what you were trying to infer from it I guess. Both methods shown are novel in their own way but full quality video streaming has been done and could still be done, most likely with the same issues current systems have.

Doom is a fast paced fps, a stress test I would think for streaming, the same for SFV, where frame perfect execution is required for high level play. I do note this implementation only scores the same as full video streaming by the testers, so whatever Microsoft have planned I hope it's more than just this. A less than perfect experience just using less bandwidth is still less than perfect.

It would be interesting to see how delta processing works with a large folliage filled screen, I suspect the low polly Doom3 did have some advantages there.

The round trip issues, I did wonder if the desire to lower bandwidth was one way to reduce this, for example if it's 7 times less data then they could send upto 7 copies of the possible next frame for the same budget, even more if they can use deltas between these new frames. In effect predicting the players actions which is then selected locally with low to no lag.

They did something similar for cloud be where computer learning helped predict player actions to a fairly high success rate.


Speculative execution on the server a round-trip ahead of the client, in this case the client warps the result if it is wrong, but what if they could send more than one frame for selection?

Who knows but interesting research all the same.
 
Based on the infrastructure they have and continue to develop because of Azure. They even building two data centers here in South Africa.
It's not datacentres alone that can address the problem. Closer datacentres means less jumps for your data, but it still has to pass through local ISPs and local routing where you can't be sure of packet delivery times. There's no way to guarantee low-latency packet delivery short of dedicated connections to the servers that bypass the Internet infrastructure, which isn't plausible.

Ergo, any solutions to streaming have to be highly tolerant of latency and packet loss. Solutions that need upwards communication with the servers have it doubly bad versus streaming video, as IO has to be sent to the server, processed, and the output returned. If MS have a solution for that, it has to come from elsewhere to just local servers.
 
Back
Top