Digital Foundry Article Technical Discussion [2019]

Status
Not open for further replies.
But as it is, you'll still have to pay for most of the games. Currently out of the 30-ish (was it 30? or was it 20?) games available on Stadia...only 2 of them can be played with just the subscription fee alone.

12.
22.

Originally it was just 12, but then they added more to hit 22. You'll have to wait until next year for 30ish.
 
For people where price is the main driving force, how many of them can afford the type of internet connection that would be required for a good Stadia experience? Currently it requires the best internet connection you can get just to get a "passable" if compromised experience.

You're trading off a one time console hardware fee for local play versus a continuous monthly subscription fee for a compromised streamed play experience.
I am not disagreeing with you in regards to the price model as it is today, but if they want to compete, they have to change the model in someway. But for the sake of argument, let's say the price is cheaper with Stadia vs Console vs PC.

Then what I tried to convey, is that technically they can succeed, it's very much doable. It basically works well enough for the people that are price sensitive and prefer the convenience and don't need 4K/8K super duper leet stuff. I would love to see the market research for streaming game services and how the profile of the target customer looks like.

In regards to internet connections having to be top grade, well that is not really true, or rather its true now, but that is because everybody has been blinded by SPEEEEEED. Today any broadband connection challenge is solved with SPEED, but I am of the opinion that some nice QoS and a better CPE is the proper solution. Without everybody needing 1Gbps links, still the weak link in the chain is wifi, speed wise it might give you throughput numbers that are higher than your internet connection. But the QoS/QoE on wifi is best effort and really really really hard to guarantee in anyway form or shape and it shows in the goodput numbers.

You can have 1Gbps connection, but when somebody in the house starts a download then that download will kill any other traffic flow unless there are some QoS/shaping in place, not only on the CPE but on the upstream COE boxes also.

If I was an ISP today, I would look into doing that shaping based on traffic type. Any latency sensitiv stuff gets highest priority and a guaranteed bandwidth and the rest gets best effort and whats left of the bandwidth. And you sell that as your service. Streaming TV will "always" work and same with streaming games and voice and what not. Bulk/Burst stuff like downloads and surfing and youtube etc will work, but since they have less latency sensitive characteristics you can run them with best effort.

Networking rant over :D
 
I am not disagreeing with you in regards to the price model as it is today, but if they want to compete, they have to change the model in someway. But for the sake of argument, let's say the price is cheaper with Stadia vs Console vs PC.

Then what I tried to convey, is that technically they can succeed, it's very much doable. It basically works well enough for the people that are price sensitive and prefer the convenience and don't need 4K/8K super duper leet stuff. I would love to see the market research for streaming game services and how the profile of the target customer looks like.

In regards to internet connections having to be top grade, well that is not really true, or rather its true now, but that is because everybody has been blinded by SPEEEEEED. Today any broadband connection challenge is solved with SPEED, but I am of the opinion that some nice QoS and a better CPE is the proper solution. Without everybody needing 1Gbps links, still the weak link in the chain is wifi, speed wise it might give you throughput numbers that are higher than your internet connection. But the QoS/QoE on wifi is best effort and really really really hard to guarantee in anyway form or shape and it shows in the goodput numbers.

You can have 1Gbps connection, but when somebody in the house starts a download then that download will kill any other traffic flow unless there are some QoS/shaping in place, not only on the CPE but on the upstream COE boxes also.

If I was an ISP today, I would look into doing that shaping based on traffic type. Any latency sensitiv stuff gets highest priority and a guaranteed bandwidth and the rest gets best effort and whats left of the bandwidth. And you sell that as your service. Streaming TV will "always" work and same with streaming games and voice and what not. Bulk/Burst stuff like downloads and surfing and youtube etc will work, but since they have less latency sensitive characteristics you can run them with best effort.

Networking rant over :D

Wifi makes things worse currently for Stadia or any gaming WRT to latency and consistency. However, wired ethernet doesn't solve problems with requiring top quality internet.

For example, a person with high speed cable internet could have a good experience. Someone living in the same area but on another block with the exact same internet may have a horrendous experience. The difference being that they are on different trunks of that ISPs service. That's because in order to offer both high bandwidth and low cost (to attract customers) internet trunks are oversubscribed based on what the ISP uses as an "average" daily use case for the "average" user.

You can see this in play all the time in MMORPGs, where depending on how many people on a network trunk are using their internet (and for what purpose) a player could have either a good experience or see players rubber banding all over the place. And this is with people spending 100+ USD for gigabit internet.

QOS/QOE isn't going to solve this problem with ISPs and profitability (IE - low enough prices to get people to sign up for internet services).

Unless you can afford to pay for guaranteed bandwidth and line service (usually reserved for businesses and is quite costly) then there is no way to guarantee that someone will have a good experience on any streaming service.

That's going to be a problem no matter how far into the future we go as long as there are things that can use a lot of bandwidth and/or generate a lot of internet traffic (connections).

If there was a solution then DOS attacks would never succeed.

Regards,
SB
 

Completely unrelated to Ray Tracing, but man that brought back memories of competitive gaming in Q2 with the Threewave CTF mod. I was the god of grapple stealing the opposing team's flag and staying alive via grapple dodging.

The grapple hook implementation and mechanics in that mod are what I still consider the best implementation in any game, ever.

Regards,
SB
 
It is amazing how much difference lighting and materials can do to low such a low poly game.

Does it expand the color palette from the ugly black, brown, gray, and red?
 
Wifi makes things worse currently for Stadia or any gaming WRT to latency and consistency. However, wired ethernet doesn't solve problems with requiring top quality internet.

For example, a person with high speed cable internet could have a good experience. Someone living in the same area but on another block with the exact same internet may have a horrendous experience. The difference being that they are on different trunks of that ISPs service. That's because in order to offer both high bandwidth and low cost (to attract customers) internet trunks are oversubscribed based on what the ISP uses as an "average" daily use case for the "average" user.

You can see this in play all the time in MMORPGs, where depending on how many people on a network trunk are using their internet (and for what purpose) a player could have either a good experience or see players rubber banding all over the place. And this is with people spending 100+ USD for gigabit internet.

QOS/QOE isn't going to solve this problem with ISPs and profitability (IE - low enough prices to get people to sign up for internet services).

Unless you can afford to pay for guaranteed bandwidth and line service (usually reserved for businesses and is quite costly) then there is no way to guarantee that someone will have a good experience on any streaming service.

That's going to be a problem no matter how far into the future we go as long as there are things that can use a lot of bandwidth and/or generate a lot of internet traffic (connections).

If there was a solution then DOS attacks would never succeed.

Regards,
SB

Oversubscription is business decision, internet has moved on from bulk downloads to streaming, if you want to be in business you need to make sure that streaming is something that can be done on your network. If not, you are going out of business. Now Youtube, Netflix, Spotify type of services are streaming, but they can buffer up data, so its less real-time sensitive than gaming and voice/video calls.
Current mainstream gaming is not bandwidth intensive, but Stadia, PS Now and Xbox xCloud are. Again you want to stay in business, you need to adapt your services/products.
And if you lag when playing at night because your neighbour is torrenting then I would say it's a sucky ISP.

Now what I argue is that it does not take that much change to actually do it. Just an ISP with half a brain and a half way modern network. You can QoS / shape the traffic in your network. You set that a customers link has a certain "profile", this is just like the frequency stuff cable operators are used to do already. But we do it on the IP layer in addition to what ever they do on the cable layer.

So if you say you have a 50Mbps DL service/product. You portion in out ala, 5 Mbps for voice/video calls/conferences, 15Mbps is reserved for game streaming, 15 mbps for video streaming, 15 for the rest. Which means your ftp/torrent downloads can max get 15Mbps as long as there is voice,gaming and video going on. But if there is no video or gaming, then your download can get those chunks also. But the second you start gaming or video, the download looses that 15Mbps. This is quite easy. And if the customer wants to have multiple video / gaming streams, then they need to upgrade.

This you can do for any tier of speed service you sell as an ISP and its easy and doable on most hardware today. We did this for customers back in 2006, so I know it works and with todays much more powerful DPI stuff it's a cake walk. :D

Question is what are the minimum speeds you can get away with?


As for DOS attacks, they are often bandwidth based, but they do not have to be and the best ones are not, they attack bugs in TCP/IP stacks, like the good old Ping of Death.

As for speeds, according to speedtest.net the average speed is 70Mbps.

https://www.speedtest.net/global-index
 
Last edited:
As for DOS attacks, they are often bandwidth based, but they do not have to be and the best ones are not, they attack bugs in TCP/IP stacks, like the good old Ping of Death.

Not just bandwidth as I noted it has a lot to do with the amount of connections attempted as well, similar to the old ping of death.

It's enough to saturate a user (household), branch, or node with so many connection requests that it cannot all be processed in time. You can see this effect at the local level just by setting the number of connections used to a slightly unreasonable level in a torrent application. And that happens relatively frequently. If too many people do that on a neighborhood network trunk, it will then affect everyone on that network trunk. And some torrent clients attempt to mask their connection as other data types in order to get around ISP QOS shaping.

However, it need not be as all encompassing as a DOS attack (whether bandwidth based, connection based, or bug based) in order to affect gaming or streaming or any other service. Orders of magnitude less, in fact. And it doesn't have to be something that is maliciously or consciously done. This isn't something that needs to completely deny service, but just to introduce enough of a load that the infrastructure at certain junctions starts to take a few hundred more milliseconds in order to process it.

Twitch's service is entirely built on relatively quick live streaming (unlike Netflix) and here you can somewhat regularly see its streaming service have bubbles where stream quality is degraded and/or interrupted. This situation gets much worse if you enable low latency mode.

And that happens to people regardless of the quality of their ISP albeit people with good ISPs have this happen less often. But even still it can get to the point where it affects even people with guaranteed service contracts (expensive business internet lines).

Regular streaming has everything processed and delivered at an average latency of around 10-15 seconds, IIRC. Low latency reduces that but the latency is still in seconds.

And that's with a live streaming service that is much larger than Stadia (number of users) is currently but with demands much smaller than Stadia (significantly less bandwidth per user). Netflix isn't really applicable here as it isn't doing live streaming.

In order to achieve what you envision, the entire Internet would need to be reworked with infrastructure redone on every node of every backbone provider and every node of every ISP. Network trunks that service neighborhoods would also need to be entirely reworked.

I don't think Google has enough money or power to make something like that a reality.

Overprovisioning is a business decision that is never going away. Let's say your average user of a 30-70 Mbps connection is expected to average roughly 5-10 Mbps per day on average (it's quite a bit less than that for most neighborhoods) and maybe 15-20 Mbps during busy hours (like when people just get off from work, and likely less for most neighborhoods).

Now what happens if users with Stadia are suddenly averaging quite significantly more than that due to Stadia's bandwidth and latency demands?

What happens is that the average bandwidth per user calculations need to be redone and ISPs would need to adjust pricing accordingly (more bandwidth allocated per user) or implement bandwidth caps (same bandwidth per user, but now enforced with a cap).

So the reality for people wanting a good Game Streaming experience (at least WRT Stadia bandwidth demands) could either see their monthly ISP costs triple or quadruple (possibly more) and/or ones without a bandwidth cap will suddenly find they have a bandwidth cap which means their monthly ISP costs could balloon even further if they are charged for going over that cap (like Comcast, ARGH).

Although in Comcasts defense, the quality of service (average latency in games and connection service requests fulfilled time) on my internet has improved now that people on my network trunk are disincentivized from using a lot of bandwidth per month. Of course, that also means I can no longer stream nearly as much 4k video as I would like. Heck I can't even stream 1080p/60 video as much as I would like. So it's a mixed blessing.

And this is with me ONLY using Comcast for video streaming. I have another low cost (10 USD per month) and low bandwidth internet service for everything else which has unlimited bandwidth but is limited to only 15-25 Mbps depending on the time of day. And at certain really busy times it can drop as low as 6 Mbps. Basically can't do 4k video on this connection reliably and even 1080p/60 can't be reliably done. At the time of this posting, I'm only averaging 10 Mbps on this connection, yay.

I can't afford to do all the other things (game downloads, OS downloads, etc.) in addition to video streaming on my 1 TB per month Comcast internet.

Basically on Comcast I would have to pay a minimum of an extra 50 USD per month to remove the bandwidth cap if I wanted to use Stadia. If Stadia actually succeeds and a lot of people start using it? I expect that 50 USD per month cost to balloon to at least 150-200 USD for unlimited bandwidth per month.

So yeah, Stadia most certainly does not represent good value for money in my area. Many areas where there is no bandwidth cap, people often suffer from highly variable ping times (latency) for games depending on the time of day. This was extremely noticeable when I still did hardcore raiding in MMORPGs where variable ping times meant complete raid wipes and wasted time. Especially in FFXIV where timing for the Savage raids was very precise and timing windows were very short.

Regards,
SB
 
Last edited:

Interesting results from their xcloud testing.

Also MS should either team up with a manufacturer or make an Android phone, for xcloud not for high volumes, but geared to low latency and attachable controls.
Although considering it can beat some tv's with game mode off, is it necessary?
 

Interesting results from their xcloud testing.

Also MS should either team up with a manufacturer or make an Android phone, for xcloud not for high volumes, but geared to low latency and attachable controls.
Although considering it can beat some tv's with game mode off, is it necessary?
You do things one at a time; especially when you're testing the market for feasibility. MS will confirm the market is there and likely expand to partners before taking on a handheld device.
 

Interesting results from their xcloud testing.

Also MS should either team up with a manufacturer or make an Android phone, for xcloud not for high volumes, but geared to low latency and attachable controls.
Although considering it can beat some tv's with game mode off, is it necessary?

Still faces many of the same problems as Stadia, but at least MS are approaching things a little more realistically.
  • The beta is free unlike Stadia. Yes, Stadia supposedly launched, but lets face it, it's a glorified beta where you have to pay for most of the games.
    • So while you are still going to run into problems and features might be missing, at least you aren't paying for it.
  • They limit it to 720p/60 in order to use as little bandwidth as possible while still offering a reasonably good gaming experience.
    • Internet connection requirements should be significantly reduced hopefully allowing for use of cheaper internet plans.
    • Doesn't put as much stress on existing internet infrastructure.
      • Caveat here is that since it's aimed a mobile phones, congestion on cellular wireless can be significantly worse than wired internet leading to more varied game experiences.
      • On wired internet, however, the experience should be more universally consistent (when using the same devices) than Stadia due to low resource demands.
  • Target the smaller screens of smartphones to try to hide the lower quality and potential compression artifacts.
  • 1 "machine"/blade = 1 person gaming. No load balancing multiple games on a "system."
    • Likely more expensive to roll out, but should guarantee a certain level of user experience.
I still don't have much faith that Xcloud will succeed in the way that stream gaming proponents think it will, but at least it's targeting a more realistic situation.

And it's definitely not something for me. 50 ms (or more depending on internet conditions) of additional latency (more importantly, potentially variable latency) is basically a no go for me. But I understand that most gamers are going to be more tolerant of additional latency, so it'll probably be good enough for some people.

Good luck to them.

Regards,
SB
 
You do things one at a time; especially when you're testing the market for feasibility. MS will confirm the market is there and likely expand to partners before taking on a handheld device.
I doubt they have any interest in producing the hardware. Nothing to do with testing waters or a slow roll out.
But you never know, they've been more flexible and open to change nowadays.
 
If there was a phone, it'd almost certainly be a partnership. Though that said, given Surface Line, a Surface Phone doesn't seem implausible either. I doubt there's much market for such a device though making it not worth it unless either as a flagship product or high-margin niche.
 
Surface Pro X is a curious one. Surface Neo is supposed to be using Intel Lakefield, I think?

Seems odd to go back to ARM for just one line of pro tablets for developers, developers, dev-.
thinking-face-apple-icon.png
ninja.gif

¯\_(ツ)_/¯
 
Last edited:
Status
Not open for further replies.
Back
Top