Google Stadia in desperate need of promised "negative latency" mode

[atxcyclist]

How would you have "negative latency" without clairvoyance?

[vanished]

2 hours ago, atxcyclist said:

How would you have "negative latency" without clairvoyance?

The original plan was to use AI to predict your movements so it could make them before you even do so, thus allowing the game to act, the frames to be rendered, and sent on their way toward your unit so that by the time you actually press the button, the frames are there and it's as if there was no latency.  I honestly can't believe they actually thought that would work, it sounds completely insane and impossible.  My title for this post is just a reference to that.

[MadyTehWolfie]

I don’t know how my friend defends stadia and says it’s great. Couldn’t possibly be because he has fios and pays for 500mb upload and download. Completely unrealistic and stupid when he only uses it to stream games to his living room which a steam link can do far better.

[Eniqmatic]

14 hours ago, MadyTehWolfie said:

I don’t know how my friend defends stadia and says it’s great. Couldn’t possibly be because he has fios and pays for 500mb upload and download. Completely unrealistic and stupid when he only uses it to stream games to his living room which a steam link can do far better.

To be fair, the general consensus is Stadia works great, the underlying technology is there and the games are very playable. What most of the complaints are about now are the lack of basic features, which is understandable but hopefully they will deliver on those. This is good news because the actual hardest challenge (getting the games to a playable latency) seems to be great.

 

Another point is, it isn't Google's fault that people have data caps/terrible speeds. Data caps in Europe are non-existent so it is perfectly useable for us over here. Of course, its a consideration you should have when buying.

 

Steam link is great, if you already have the gaming PC to power it, for casuals who don't have $1000+ PC's, I think its a good option that will hopefully only get better if Google keep at it.

 

Just my opinion!

[Thaldor]

1 hour ago, Eniqmatic said:

To be fair, the general consensus is Stadia works great, the underlying technology is there and the games are very playable. What most of the complaints are about now are the lack of basic features, which is understandable but hopefully they will deliver on those. This is good news because the actual hardest challenge (getting the games to a playable latency) seems to be great.

 

Another point is, it isn't Google's fault that people have data caps/terrible speeds. Data caps in Europe are non-existent so it is perfectly useable for us over here. Of course, its a consideration you should have when buying.

 

Steam link is great, if you already have the gaming PC to power it, for casuals who don't have $1000+ PC's, I think its a good option that will hopefully only get better if Google keep at it.

 

Just my opinion!

What I have gotten is that Stadia is basicly just shit. It's not about basic features, it's about the whole concept of having paid subscription and on that you even need to pay for the games (apart from even smaller library of games included in the subscription) and the worst thing is that once you stop paying monthly, you loose those games you paid the full price for. Also that the whole service is currently on it's knees because a lot of registering, activating and device problems and a bit more on those.

 

The worst part for Stadia is that it already has competition and that competition is lightyears ahead of it in every way. mainly the competition is now already few years in beta been GeForce NOW and it whacks the shit out of Stadia. Not only does it support more devices than Stadia (almost every Mac and Windows PC +Nvidia Shield vs. Chromecast Ultra which may overheat and die, some limited tablets and Pixel phones), but at least currently it's free (as long as you get into the program) and even if it had paid subscription it would be million times better because you don't need any separate game library but you have the familiar Steam, Origin, Epic, Battle.net, Uplay and others where you can keep the games even if you stopped using GF NOW. Also just that GF NOW actually works while Stadia is unusable (I have 100Mbps VSDL2 and 200Mbps 4G (both without any caps or restrictions) in use and with both my friends Stadia (which I loaned to try out because it didn't even connect with my friends connection) was stuttering and lagging like granny in snow while GF NOW works flawlessly). And the fun thing is GF NOW is big project for Nvidia but it's not their primary thing that they would advertise and try to sell to people, while Google tries to sell Stadia and could have probably poured a lot more money in it than Nvidia even could dream to pour on GF NOW and they still suck at it.

 

Then all the promises about 4K gaming at 60FPS which have been more or less lies so far. Like yeah, good luck getting up from that hole you dug for yourself, Google.

[Eniqmatic]

1 minute ago, Thaldor said:

What I have gotten is that Stadia is basicly just shit. It's not about basic features, it's about the whole concept of having paid subscription and on that you even need to pay for the games (apart from even smaller library of games included in the subscription) and the worst thing is that once you stop paying monthly, you loose those games you paid the full price for. Also that the whole service is currently on it's knees because a lot of registering, activating and device problems and a bit more on those.

 

The worst part for Stadia is that it already has competition and that competition is lightyears ahead of it in every way. mainly the competition is now already few years in beta been GeForce NOW and it whacks the shit out of Stadia. Not only does it support more devices than Stadia (almost every Mac and Windows PC +Nvidia Shield vs. Chromecast Ultra which may overheat and die, some limited tablets and Pixel phones), but at least currently it's free (as long as you get into the program) and even if it had paid subscription it would be million times better because you don't need any separate game library but you have the familiar Steam, Origin, Epic, Battle.net, Uplay and others where you can keep the games even if you stopped using GF NOW. Also just that GF NOW actually works while Stadia is unusable (I have 100Mbps VSDL2 and 200Mbps 4G (both without any caps or restrictions) in use and with both my friends Stadia (which I loaned to try out because it didn't even connect with my friends connection) was stuttering and lagging like granny in snow while GF NOW works flawlessly). And the fun thing is GF NOW is big project for Nvidia but it's not their primary thing that they would advertise and try to sell to people, while Google tries to sell Stadia and could have probably poured a lot more money in it than Nvidia even could dream to pour on GF NOW and they still suck at it.

 

Then all the promises about 4K gaming at 60FPS which have been more or less lies so far. Like yeah, good luck getting up from that hole you dug for yourself, Google.

If you think its shit then you've been paying attention to the media spin, not the actual users. Media has hated on it from the start, not sure why. If you read community posts you will see 95% of users are having a fantastic experience. 

 

Your statement about losing games if you cancel your monthly subscription is 100% false, this is why its getting a bad rap in the media is because people are spreading misinformation: https://www.pcgamer.com/uk/cancelling-stadia-pro-wont-lock-you-out-of-your-discounted-games/

 

The service is 100% not on its knees, far from it. There was issues for 5 days after launch which they should definitely have done better with, no doubt about it. But since then, zero issues.

 

The competition is not light years ahead, infact Stadia does not have any direct competition. xCloud isn't a competitior because its Android only and 720p only. GeForce Now is pretty similar but the technology isn't the same at all. Stadia supports every device, not sure what you mean here? That's literally the whole point of Stadia is it's available everywhere?

 

Those that I have seen with problems is down to their network setup being poor, who then come back and say they fixed it after doing X, Y, Z. I've even seen people playing it literally on a moving train on Wifi. So not sure why you are having issues.

 

Agreed, they need to sort the 4k60FPS out with the developers.

[Kisai]

On 12/13/2019 at 3:47 AM, Eniqmatic said:

If you think its shit then you've been paying attention to the media spin, not the actual users. Media has hated on it from the start, not sure why. If you read community posts you will see 95% of users are having a fantastic experience. 

 

 

 

Where?

 

I've played with it for a while and it feels like a laggy mess. If I compare it to playing any MMORPG, I've last seen crappy MMO functionality from latency back in 2009 or so.

 

If someone can't notice the latency, then they likely live close enough to where the servers are to not feel it as badly or never played the games featured before and have no idea how it's supposed to feel. But anyone who has played MMO's for years can certainly feel how laggy it is.

[Derangel]

On 12/13/2019 at 5:47 AM, Eniqmatic said:

If you think its shit then you've been paying attention to the media spin, not the actual users. Media has hated on it from the start, not sure why. If you read community posts you will see 95% of users are having a fantastic experience. 

 

Your statement about losing games if you cancel your monthly subscription is 100% false, this is why its getting a bad rap in the media is because people are spreading misinformation: https://www.pcgamer.com/uk/cancelling-stadia-pro-wont-lock-you-out-of-your-discounted-games/

 

The service is 100% not on its knees, far from it. There was issues for 5 days after launch which they should definitely have done better with, no doubt about it. But since then, zero issues.

 

The competition is not light years ahead, infact Stadia does not have any direct competition. xCloud isn't a competitior because its Android only and 720p only. GeForce Now is pretty similar but the technology isn't the same at all. Stadia supports every device, not sure what you mean here? That's literally the whole point of Stadia is it's available everywhere?

 

Those that I have seen with problems is down to their network setup being poor, who then come back and say they fixed it after doing X, Y, Z. I've even seen people playing it literally on a moving train on Wifi. So not sure why you are having issues.

 

Agreed, they need to sort the 4k60FPS out with the developers.

People have "spread misinformation" because Google has done a piss poor job of actually explaining anything. Nothing was clear before launch. They didn't even talk about all the promised features that wouldn't be at launch until directly asked about it in an AMA. If not for that AMA they never would have revealed their lies.

 

Stadia supports every device? BULLSHIT! Now, you're the one spreading misinformation. The app works on a lot of devices, but you can't stream games on them. It works on Chrome, but only at 1080p with no surround sound, no HDR, nothing but plain 1080p. Outside of that you can play on ONLY Google Pixel phones or the singular Chromecast Ultra that ships with the bundle. Nothing else, until some nebulous point next year. Oh, and, you can't even play it over mobile data at the moment, unlike xCloud. The xCloud beta might currently be limited to Android devices, but at least it works on more than three of them. I have it installed on my Shield and Fire HD. Stadia is a beta product being called a full release, I really wouldn't put it that much a head of xCloud or Geforce Now. Also, you seem to have conveniently ignored Playstation Now, a fully released service that tends to do things better than Stadia.

 

Most of the US has a poor connection. That may not be Google's fault, but it is their problem. They make all those grand promises that simply cannot be delivered on current connections, at least in the largest consumer market for this stuff. That's without getting into Google's blatant lies about connection speed requirements.

[TechyBen]

On 12/12/2019 at 12:16 AM, Ryan_Vickers said:

The original plan was to use AI to predict your movements so it could make them before you even do so, thus allowing the game to act, the frames to be rendered, and sent on their way toward your unit so that by the time you actually press the button, the frames are there and it's as if there was no latency.  I honestly can't believe they actually thought that would work, it sounds completely insane and impossible.  My title for this post is just a reference to that.

We keep needing to add, that it does work for retro games, but you need 8+ or so renders of each frame/state (when a modern computer can run over 100 copies of Window 95 smoothly, this is no trouble). Which lets you get the frame and only accounts for delivery latency, it completely removes render latency from the delays. But this is also because there are only 8 buttons on the controller/etc, so also a LOT less possible states than modern gaming.

[vanished]

22 minutes ago, TechyBen said:

We keep needing to add, that it does work for retro games, but you need 8+ or so renders of each frame/state (when a modern computer can run over 100 copies of Window 95 smoothly, this is no trouble). Which lets you get the frame and only accounts for delivery latency, it completely removes render latency from the delays. But this is also because there are only 8 buttons on the controller/etc, so also a LOT less possible states than modern gaming.

It's true that retro games with far fewer potential actions per frame are more likely to work, but I'm curious now so let's do the math on that.

 

Lets say we are seeking to eliminate 200 ms of lag on a 60 fps game.  That's 12 frames it has to predict into the future.  I don't believe reliable prediction will ever be possible so lets forget that and just consider what it would take to pre-render every possibility.  As you say, there's maybe 8 buttons.  Each game will bring with it more or less possibilities but that seems like a reasonable middle ground.  That means from any given point in time, there's 8¹² (aka almost 69 trillion) different states the game could be in 200 ms from now.  Even for a retro game that seems quite impossible to pre-render and cache.  Even if we drop it to something vastly smaller and more doable like 4⁸, that's still over 65 thousand possibilities.

 

However we now have to reconsider that prediction statement I made earlier since the entire system relies on that.  What (theoretically) makes this system work isn't pre-rendering all the possibilities, it's choosing one and then sending that to the user before they've even requested it.  This is where the prediction comes in - you can't send one unless you know which one is going to be requested.  You also obviously can't send all of them due to the huge figures mentioned above, so there's not really a way around this.  In the likely event that a mistaken prediction was made, it's going to be a minimum of whatever your latency is (say 200 ms) before you can correct it, and in the mean time the user has witnessed something very jarring and been totally thrown off.

 

In the end, I think the practical issues are trivial comparatively speaking.  The crux of this is how good your prediction can be, and if it's not perfect, it just won't work, and I think it's safe to assume for now that it will never be perfect.  Maybe some day enough will have changed to solve one of these issues or another but for now it seems impossible to me.

[MadyTehWolfie]

On 12/13/2019 at 4:25 AM, Eniqmatic said:

To be fair, the general consensus is Stadia works great, the underlying technology is there and the games are very playable. What most of the complaints are about now are the lack of basic features, which is understandable but hopefully they will deliver on those. This is good news because the actual hardest challenge (getting the games to a playable latency) seems to be great.

 

Another point is, it isn't Google's fault that people have data caps/terrible speeds. Data caps in Europe are non-existent so it is perfectly useable for us over here. Of course, its a consideration you should have when buying.

 

Steam link is great, if you already have the gaming PC to power it, for casuals who don't have $1000+ PC's, I think its a good option that will hopefully only get better if Google keep at it.

 

Just my opinion!

Never heard of it being playable. All the videos on youtube showed latency was trash even when they did their presentation of it. My frien has multiple PC's. Hes using stadia to do what steam link can do. Even defends having to buy games and pay the subscription service.

[TechyBen]

2 hours ago, Ryan_Vickers said:

It's true that retro games with far fewer potential actions per frame are more likely to work, but I'm curious now so let's do the math on that.

 

Lets say we are seeking to eliminate 200 ms of lag on a 60 fps game.  That's 12 frames it has to predict into the future.  I don't believe reliable prediction will ever be possible so lets forget that and just consider what it would take to pre-render every possibility.  As you say, there's maybe 8 buttons.  Each game will bring with it more or less possibilities but that seems like a reasonable middle ground.  That means from any given point in time, there's 8¹² (aka almost 69 trillion) different states the game could be in 200 ms from now.  Even for a retro game that seems quite impossible to pre-render and cache.  Even if we drop it to something vastly smaller and more doable like 4⁸, that's still over 65 thousand possibilities.

 

However we now have to reconsider that prediction statement I made earlier since the entire system relies on that.  What (theoretically) makes this system work isn't pre-rendering all the possibilities, it's choosing one and then sending that to the user before they've even requested it.  This is where the prediction comes in - you can't send one unless you know which one is going to be requested.  You also obviously can't send all of them due to the huge figures mentioned above, so there's not really a way around this.  In the likely event that a mistaken prediction was made, it's going to be a minimum of whatever your latency is (say 200 ms) before you can correct it, and in the mean time the user has witnessed something very jarring and been totally thrown off.

 

In the end, I think the practical issues are trivial comparatively speaking.  The crux of this is how good your prediction can be, and if it's not perfect, it just won't work, and I think it's safe to assume for now that it will never be perfect.  Maybe some day enough will have changed to solve one of these issues or another but for now it seems impossible to me.

No. You render the 8 possible states. If it takes 100ms of *original* code, due to timings/programming design (so not removable if emulating original system) then "future branch prediction" on the CPU would, as you say, be 8^12 or more.

 

However, as you know the user can only make 8 inputs, you can run 8 versions of the game simultaneously with no branch prediction. Then kill the 7 threads that are irrelivant, and keep the 1, then branch that one off, each frame. If the game is running 60fps, you have 16ms per frame time allowance. If its' 5ms to render the frame (retro game, super "light" on GPU/CPU) and 10ms to spin up/cull the threads/branches, that's giving you true 60fps 1 frame accuracy, to a game that previously had 100ms latency because of the way the code worked (it's still a game running 100ms in the "future", but you can remove that limit to it now). I realise doing 3 or 4 frames ahead then needs 64 or 4096 threads/branches, but this is again, much more manageable and selective in code than the 8^12 you mention.

 

Think small, think little steps, but also think the big picture. You just went off the deep end, and missed some obvious optimisations.

 

This of cause is impossible on anything past Street fighter 2, or Super Mario. We are talking needing to be able to emulate at least 8x for 1 frame, or 100s of times for multiple frames (some code could shorten this, if emulating code of the "state" and not rendering the frame, so that helps too, but still need entire game "state" emulation per branch.)

 

So much so, you missed that, it already exists!!!

 

https://arstechnica.com/gaming/2018/04/better-than-reality-new-emulation-tech-lags-less-than-original-consoles/

 

But also, yes, Stadias option was impossible, and just a "follow on" of controller input, so perhaps spin 2 games up, 1 taking only player input. 1 following the direction of the joystick if it don't move. So "turn left" carries on, and "turn right" carries on... but as you say, when it has to correct (fall back to the new input stream/game) it is jarring as now you get 2 or more frames "reversed" back to a previous state, then "sped up" to a new different state. Like errors in lag compensation in MP games!

[vanished]

2 minutes ago, TechyBen said:

No. You render the 8 possible states. If it takes 100ms of *original* code, due to timings/programming design (so not removable if emulating original system) then "future branch prediction" on the CPU would, as you say, be 8^12 or more.

 

However, as you know the user can only make 8 inputs, you can run 8 versions of the game simultaneously with no branch prediction. Then kill the 7 threads that are irrelivant, and keep the 1, then branch that one off, each frame. If the game is running 60fps, you have 16ms per frame time allowance. If its' 5ms to render the frame (retro game, super "light" on GPU/CPU) and 10ms to spin up/cull the threads/branches, that's giving you true 60fps 1 frame accuracy, to a game that previously had 100ms latency because of the way the code worked (it's still a game running 100ms in the "future", but you can remove that limit to it now). I realise doing 3 or 4 frames ahead then needs 64 or 4096 threads/branches, but this is again, much more manageable and selective in code than the 8^12 you mention.

 

Think small, think little steps, but also think the big picture. You just went off the deep end, and missed some obvious optimisations.

 

This of cause is impossible on anything past Street fighter 2, or Super Mario. We are talking needing to be able to emulate at least 8x for 1 frame, or 100s of times for multiple frames (some code could shorten this, if emulating code of the "state" and not rendering the frame, so that helps too, but still need entire game "state" emulation per branch.)

 

So much so, you missed that, it already exists!!!

 

https://arstechnica.com/gaming/2018/04/better-than-reality-new-emulation-tech-lags-less-than-original-consoles/

Clearly we're not on the same page here.  You're only describing the elimination of 1 frame of lag, not anything significant enough to actually matter.  I guess I misread your original post thinking that you were talking about more.  Regardless, this still does nothing to actually reduce the network lag that's a problem for Stadia.  You can come up with ways to render ahead all day long but they're all useless without the right tech driving the decision between which of those paths to send - that is the challenge that exists today and what google was hoping to solve, and what I see no possible way to achieve at the moment.  Another big factor worth considering is the type of old games that work with what you're describing are so small and so easy to run that a vastly better solution would be to just send a copy of the game + emulator seamlessly to the "steaming" device and have it run it locally.

[TechyBen]

Just now, Ryan_Vickers said:

Clearly we're not on the same page here.  You're only describing the elimination of 1 frame of lag, not anything significant enough to actually matter.  I guess I misread your original post thinking that you were talking about more.  Regardless, this still does nothing to actually reduce the network lag that's a problem for Stadia.  You can come up with ways to render ahead all day long but they're all useless without the right tech driving the decision between which of those paths to send - that is the challenge that exists today and what google was hoping to solve, and what I see no possible way to achieve at the moment.  Another big factor worth considering is the type of old games that work with what you're describing are so small and so easy to run that a vastly better solution would be to just send a copy of the game + emulator seamlessly to the "steaming" device and have it run it locally.

No. The game emulator can do about 3 frames IIRC. And I *gave numbers for that*. It also might only be emulating the game code/state not the actual GPU pixel perfect frames (as the emulator can push them out faster than 16fps and local so little latency).

 

Stadias implementation would not work. But in other areas, this tech totally does work (and is part of CPU branch prediction already).

[Taf the Ghost]

3 hours ago, Derangel said:

People have "spread misinformation" because Google has done a piss poor job of actually explaining anything. Nothing was clear before launch. They didn't even talk about all the promised features that wouldn't be at launch until directly asked about it in an AMA. If not for that AMA they never would have revealed their lies.

 

Stadia supports every device? BULLSHIT! Now, you're the one spreading misinformation. The app works on a lot of devices, but you can't stream games on them. It works on Chrome, but only at 1080p with no surround sound, no HDR, nothing but plain 1080p. Outside of that you can play on ONLY Google Pixel phones or the singular Chromecast Ultra that ships with the bundle. Nothing else, until some nebulous point next year. Oh, and, you can't even play it over mobile data at the moment, unlike xCloud. The xCloud beta might currently be limited to Android devices, but at least it works on more than three of them. I have it installed on my Shield and Fire HD. Stadia is a beta product being called a full release, I really wouldn't put it that much a head of xCloud or Geforce Now. Also, you seem to have conveniently ignored Playstation Now, a fully released service that tends to do things better than Stadia.

 

Most of the US has a poor connection. That may not be Google's fault, but it is their problem. They make all those grand promises that simply cannot be delivered on current connections, at least in the largest consumer market for this stuff. That's without getting into Google's blatant lies about connection speed requirements.

This is actually incorrect. I know it's a meme, but US Internet is, on the mass scale, near the top of the world and the best of any large country by a margin. It's the smaller European countries + South Korea that have invested billions of taxpayer money into their Internet where it's very good. (There's an exception to New York City, but that's due to local politics. But also probably why the chatter is always around.)

 

That said, Stadia works fine on Business Class fiber within a few miles of a Google server farm. Having chatted with someone in that situation, the service is actually solid. For the entire rest of the world, it's been pretty terrible. The QoS on Business Class connections actually seems to matter a lot more than the actual bandwidth. 

[goodtofufriday]

I have stadia. Its honestly nothing to write home about. It works well enough, but paying full price for games ive already played, or own, is just a loss.

Fighting games also just arent meant for this. 

 

The one exclusive is too expensive.

[Kisai]

4 hours ago, TechyBen said:

No. You render the 8 possible states. If it takes 100ms of *original* code, due to timings/programming design (so not removable if emulating original system) then "future branch prediction" on the CPU would, as you say, be 8^12 or more.

 

However, as you know the user can only make 8 inputs, you can run 8 versions of the game simultaneously with no branch prediction.

Just stop there. That is not what they are doing. Google is not going to spin up 8 sessions of the game and run them simultaneously, dropping paths not taken and spinning up new vm's every 200ms. 

 

You're vastly oversimplifying how to predict anything. There isn't 8 states, there are 19 buttons with an on and a off state, so 19^2, so there's 361 minimum possible states to track for the buttons alone. Add in the analog controls, and you see pretty darn quick why this is barking up the wrong tree.

 

https://www.w3.org/TR/gamepad/

 

This is the standard "17" button gamepad. Stadia has two more buttons than this, and two of the shoulder buttons are analog buttons like on xbox/playstation controllers. 

 

From playing games thus far with it, it seems like "negative latency" is simply as "press and hold" technique that MMORPG's have been using since inception. More newer PC MMO games, the server keeps the player in "auto attack" mode when the player drops s connection, so a player can't just disconnect to avoid a loss. But the reality of how this appears to work is that there is no AI involved. Google could, in theory, have AI trained off user data to see what the most optimal controller state is when the player is x minutes into a fixed stage, so that it if you have something like Mario running across the screen, the state with a local device would be if the battery died in the controller, mario would just stop moving, dead, stop on a dime. So taking that into account, a Stadia "Ai" learned off player input would know this, so that if the AI knows the player is running and should be running at this point in the stage, and the player drops, then the AI could keep holding those buttons until they reconnect. But it's not going to play the game for a player for more than 200ms. You're not going to let the AI play the entire game by dropping the the controller only.

 

But that's only half the problem. The other half is how do you keep streaming a video like that input is being AI fed? If the player drops, so does their video being received. Most people's reaction to dropped video is to release the controller.

[Otto_iii]

On 12/13/2019 at 4:25 AM, Eniqmatic said:

This is good news because the actual hardest challenge (getting the games to a playable latency) seems to be great.

A lot of people have forgotten how pronounced latency is, with exception of people who care about being more competitive in games, because many older displays whether in PC space of Televisions for last 10 years have had very bad input delay themselves, OLEDs likely to become more popular and brings us back to CRT levels of input delay even on large television displays, so any layman who has avoided noticing will do so even more then before if they simply try a xbox or playstation on their buddies OLED TV. 

If you have ever taken time to try to get games at about 20-50ms total latency, its very very hard to go back, it feels god awful, forgetting about being competitive the funfactor plummets, it feels like sludge.  At best Stadia will double this latency, its simply not likely or possibly for it to get better unless they literally built thousands of servers blanketing the world so everybody was within a few hundred miles of a host for stadio. 
 

6 hours ago, goodtofufriday said:

I have stadia. Its honestly nothing to write home about. It works well enough, but paying full price for games ive already played, or own, is just a loss.

Fighting games also just arent meant for this. 

 

The one exclusive is too expensive.

I was thinking about what Linus said on the WAN show, i could see streaming games be interesting if like in his theoretical analogy you made games specifically for it which could only be played via streaming services because of the level of details going on being too much for anything besides a huge datacenter/server like what stadia runs on.  Stuff where you theoretically have like thousands of people at a time in a game, something no computer could handle and most modern servers would struggle with right now, but potentially a single dedicated cloud with streams broadcasted from it could handle.

My favorite theoretical for this would be something like VRChat, but since input latency is a huge issue in VR that can actually make you feel sick, likely something more like WoW, more strategy and story based where a huge arena environment would still be cool

[Eniqmatic]

14 hours ago, Derangel said:

People have "spread misinformation" because Google has done a piss poor job of actually explaining anything. Nothing was clear before launch. They didn't even talk about all the promised features that wouldn't be at launch until directly asked about it in an AMA. If not for that AMA they never would have revealed their lies.

 

Stadia supports every device? BULLSHIT! Now, you're the one spreading misinformation. The app works on a lot of devices, but you can't stream games on them. It works on Chrome, but only at 1080p with no surround sound, no HDR, nothing but plain 1080p. Outside of that you can play on ONLY Google Pixel phones or the singular Chromecast Ultra that ships with the bundle. Nothing else, until some nebulous point next year. Oh, and, you can't even play it over mobile data at the moment, unlike xCloud. The xCloud beta might currently be limited to Android devices, but at least it works on more than three of them. I have it installed on my Shield and Fire HD. Stadia is a beta product being called a full release, I really wouldn't put it that much a head of xCloud or Geforce Now. Also, you seem to have conveniently ignored Playstation Now, a fully released service that tends to do things better than Stadia.

 

Most of the US has a poor connection. That may not be Google's fault, but it is their problem. They make all those grand promises that simply cannot be delivered on current connections, at least in the largest consumer market for this stuff. That's without getting into Google's blatant lies about connection speed requirements.

I don't disagree with that, their communication has been abysmal. I 100% agree.

 

OK, "supports" was perhaps not the word I should have used, official support has been limited so far but it does work on many many devices by simply opening chrome on any other phone. Seen people running it on Galaxy's, OnePlus's etc by just using Chrome. The 1080p Chrome thing was actually communicated beforehand. However I find it interesting your saying Chrome is limited to 1080p when the entire xCloud library is limited to 720p? Again, Playstation now limited to 720p. 

[TechyBen]

8 hours ago, Kisai said:

Just stop there. That is not what they are doing. Google is not going to spin up 8 sessions of the game and run them simultaneously, dropping paths not taken and spinning up new vm's every 200ms. 

 

You're vastly oversimplifying how to predict anything. There isn't 8 states, there are 19 buttons with an on and a off state, so 19^2, so there's 361 minimum possible states to track for the buttons alone. Add in the analog controls, and you see pretty darn quick why this is barking up the wrong tree.

 

https://www.w3.org/TR/gamepad/

 

This is the standard "17" button gamepad. Stadia has two more buttons than this, and two of the shoulder buttons are analog buttons like on xbox/playstation controllers. 

 

From playing games thus far with it, it seems like "negative latency" is simply as "press and hold" technique that MMORPG's have been using since inception. More newer PC MMO games, the server keeps the player in "auto attack" mode when the player drops s connection, so a player can't just disconnect to avoid a loss. But the reality of how this appears to work is that there is no AI involved. Google could, in theory, have AI trained off user data to see what the most optimal controller state is when the player is x minutes into a fixed stage, so that it if you have something like Mario running across the screen, the state with a local device would be if the battery died in the controller, mario would just stop moving, dead, stop on a dime. So taking that into account, a Stadia "Ai" learned off player input would know this, so that if the AI knows the player is running and should be running at this point in the stage, and the player drops, then the AI could keep holding those buttons until they reconnect. But it's not going to play the game for a player for more than 200ms. You're not going to let the AI play the entire game by dropping the the controller only.

 

But that's only half the problem. The other half is how do you keep streaming a video like that input is being AI fed? If the player drops, so does their video being received. Most people's reaction to dropped video is to release the controller.

I never said that's what Googles doing.  

I said to be careful on saying it's all bunk. Because the system exists and is useful for something that does both 1) emulation on limited small codebases (retro games) and 2) local emulation (so render frames/delivery is very quick and within a frames time of 16ms or so).

 

Then said it won't work for Google and Stadia as it's both too computationally intensive, and you still have 100-200ms round trip (of update cycle even if just confirming the frame is good/bad).

 

 Read what I said. I even said this is probably just "press and hold". I SAID EXACTLY THAT!  [Gives up on the human race and the internet, drops mic, exits stage left, and blasts off to Mars]

 

 

[goodtofufriday]

20 hours ago, Derangel said:

but you can't stream games on them. It works on [...] the singular Chromecast Ultra that ships with the bundle. Nothing else

 

I dont disagree with your arguments but I think its worth noting you can update current chromecast to play stadia. I did that myself. I only purchased a controller and got a buddypass, and updated my chromecast that I bought last year. 

[SPARTAN VI]

Got an invite so I booted up Destiny 2 over a wired connection. I've been using GeForce Now on and off for the past few months, so Stadia has its work cut out.

 

Visually, it was a soupy 720P mess. GeForce Now has way more options as far as tailoring the visual experience, not to mention it retains all of the in-game settings. Input lag was not very noticeable, and not nearly as bad as the OP animations. However, I noticed some sporadic mouse movements that was game breaking.. basically every 10-20 seconds, my mouse look would jerk dramatically in a random direction. Could be in a middle of the fire fight, or just jogging down a hall, then suddenly be looking at the ceiling/floor/behind my character. If that alone were fixed, this would have been a playable experience, even at soupy 720P. 

 

My speedtest results:

 

And my jitter test results are A+ across the board.

[BroliviaWilde]

I dug out my old steam link the other day, and even after updating it - and we are on a home wired network, the latency was slight, but ti was still there, and that is over about between 50' ~ 100' of network cable.   I didn't care for it and put it away.   I then compare that to Stadia, and just how much worse Stadia would be and I can only come to one conclusion:

 

Ooof

[Derangel]

10 hours ago, Eniqmatic said:

I don't disagree with that, their communication has been abysmal. I 100% agree.

 

OK, "supports" was perhaps not the word I should have used, official support has been limited so far but it does work on many many devices by simply opening chrome on any other phone. Seen people running it on Galaxy's, OnePlus's etc by just using Chrome. The 1080p Chrome thing was actually communicated beforehand. However I find it interesting your saying Chrome is limited to 1080p when the entire xCloud library is limited to 720p? Again, Playstation now limited to 720p. 

xCloud is still in beta and neither service tout "4K, HDR, surround sound" as something you get for paying for the service. Having Chrome not ready for launch is a pretty big black mark on the launch.

 

4 hours ago, goodtofufriday said:

 

I dont disagree with your arguments but I think its worth noting you can update current chromecast to play stadia. I did that myself. I only purchased a controller and got a buddypass, and updated my chromecast that I bought last year. 

Did they finally roll the update out? I thought that was going to be another "sometime in 2020" thing like most of their botched promises. Good to know they at least got that done this year.