Google VP says Stadia could be more responsive than your PC in two years

Polycount

Posts: 3,017   +590
Staff

Game streaming has the potential to change the way many gamers experience their favorite titles, but the technology is still young at the moment. Many companies that offer game streaming services are still grappling with issues like latency and excessive data usage.

Given that game streaming functions by sending your inputs to a separate server and feeding back the results to your screen (whether it's on a smartphone, low-powered notebook, or TV), some amount of delay should always be expected. However, Google believes it has mitigated the problem with its Stadia subscription service -- early testers say that while the lag may be noticeable for more hardcore or competitive gamers, for the average user (with a decent internet connection), it's unlikely to be a problem.

Simply reducing input lag to nearly-imperceptible levels isn't enough for Google, though. In a recent Edge Magazine interview (first reported on by PCGamesN), Google VP Madj Bakar said Stadia's response times could be better than local machines in "a year or two" due to what he calls "negative latency." Apparently, Google is aiming to use its machine learning tech to predict your inputs, effectively making games feel even more responsive than they would on a high-end PC.

Some skepticism may be warranted here. Yes, Google has more than proven itself in the realm of AI and machine learning, but when it comes to the world of video games -- particularly competitive ones -- unpredictability is often the key to success.

Wildly zigzagging, sliding, jumping, and ducking behind cover are all common (and fairly random) tactics employed by shooter players. It will be tough for an AI to predict which of these tactics a player wishes to employ at any given moment, and if it happens to guess wrong, that's not exactly going to endear Stadia to users.

However, as PCGamesN notes, it's possible that Google is merely planning to prepare for potential inputs, and won't actually use them until it receives "confirmation" in the form of the player's real input. Think of it as a local diner preparing one of their regular's favorite meals in advance. The individual can change their mind when they arrive, but if the prediction is correct, all the better for the diner and the patron.

Either way, it's an unusual way to tackle the latency issue. It's tough to say how end-users will respond to the idea, but we wouldn't be surprised if serious gamers simply choose to take the latency hit in exchange for full control over their actions (and a bit of freedom from Stadia's watchful AI eyes).

With that said, we have no idea whether or not this "negative latency" feature will be optional, so gamers might not have a choice in the end. We'll simply have to wait for Stadia's official release next month to say for sure.

Permalink to story.

 
From the linked article that this write-up is based on

"This term describes a buffer of predicted latency, inherent to a Stadia players setup or connection, in which the Stadia system will run lag mitigation. This can include increasing fps rapidly to reduce latency between player input and display, or even predicting user inputs."

Big problem with this, as large variances in FPS is worse then slightly more input lag. The most important thing for gaming is a stable FPS number.

"So does that count as the fastest system if technically some clever algorithm is anticipating your actions for you? We’ve received a heads-up (thanks!) that negative latency, powered by a datacentre’s worth of compute silicon, may offer future cloud gaming systems flexibility to anticipate the likely action of a user, and ensure a speedy response ready for that potential eventuality. Whether or not a player takes the anticipated path or another entirely remains dependent on local player inputs."

So if the player doesn't take the anticipated path then what? They get an increase in lag again? Great, so you'll have frequent stuttering. Unless they machines can consistently predict what you'll do next 100% of the time, this is going to mean uneven lag (which is very bad for gaming).

I also don't see why an AI that can predict input couldn't be released to run on Nvidia's tensor cores either. A model could be created, running an AI to do something as simple as predict input should be doable for Nvidia. It's only the training part that could not be done real time (as far as I know). If there were benefits to doing it on the cloud, those benefits would be multiple times better local, where you don't have 36 ms plus latency to talk with some remote server. Of course, I'm extremely skeptical of this tech will even come to some benefit to begin with.
 
You know what? People will believe it because google said so. Just think of the ISP providers saying that you need "more bandwidth" for "better" gaming, when it's ping times that make the difference. They have made a ton of profit off of swindling people. /facepalm . They are preying on the unknowing and their wallets.

As krisby said - they are essentially implementing an aimbot cheat. Console games already make the click zone bigger to compensate for poor controller reaction times compared to mice.

They are pushing this so we are renting games, and they get to yank whatever games from subscribers at their whim. They also can make the price whatever they want. Just like movie and music rental services like Nutflicks and Hulu.
 
Stadia is nothing like Netflix and Hulu. With the latter two, you pay one price to watch as many videos as many times as you want for one month. If you choose to pay for subsequent months, what you are paying for, in my mind, is that there will new videos to watch. The monthly price is typically low relative to buying even one or two titles or renting a handful. If you stream a lot of videos it's an easy deal to appreciate.

Stadia, meanwhile, wants you to pay them for server power, while also paying for each game -- and in the process locking that game to Stadia's servers. If you stop paying Stadia, you lose the investment you made in that game library. And if you keep playing Stadia, you need to keep buying new games (assuming you're like most people who want new titles from time to time.) The price of a game library built up over time is not trivial so it's not like you'll want to just start again on some other service. You'll be locked in.
 
A Game-Streaming-Service with predetermined Games is useless.
I want to install the Games, that I like to play.
 
There is no possible way that Stadia would ever have better latency than a PC.
The PC itself still will have the same latency it had with a native game and then you will add on internet latency and internal processing latency at google.

Therefore this comment made by the google exec is nothing but pure marketing hogwash!
 
This is actually possible. If my CPU, RAM or GPU in two years develop a serious problem that freezes it every 5 seconds, network delivery of compressed pics may be faster.
Oh, wait, I'm still gonna need CPU, RAM and GPU to decompress the received data. Damn, then it will still be slower.

 
Lol is google planning to start rolling out google fiber across all of the USA I will believe this when I see it.
 
Lol is google planning to start rolling out google fiber across all of the USA I will believe this when I see it.
I've had GF 1000+TV since the beginning. Overall upset with the direction the experience has gone.
In the beginning the TV menu was very responsive and you could use the DVR drive to put your own content on and play from any connected tv box. Now they changed the menu UI and it has become rather laggy. Also personal content is now locked and completely gone when the UI update came. YouTube is also a preinstalled app that Google owns and it runs horrendously.

With that said I really have no faith in this project from Google to be a pleasant and lag free experience.
 
Back