Google Stadia hands-on: streaming analysis and controller impressions: eurogame

Eurogame:

We've talked elsewhere in our Google Stadia coverage about the fact that it's not a console, it's what Google calls a 'cloud native' system that leverages tight integration of gaming components in the datacentre to provide - in theory, at least - a system capable of delivering genuinely different gaming experiences, while providing a generational leap in processing power over today's consoles. That said, we should expect to see ports of multi-platform games and Google itself has already demonstrated one of them - Assassin's Creed Odyssey, delivered to the US public at the tail-end of last year for extensive beta testing. We had the chance to go hands on with the game again, running on the latest version of the streamer, and with access to Google's own controller hardware.

[crazyCoconuts]

Interesting to look at the lag measured for Google's 1080p/30 demo (about 160ms) compared to the lag measured by DF on PS Now a few years ago ( https://www.eurogamer.net/a...
Pretty comparable, even though PSNow was measured on a slower ADSL line...

[darthv72]

Your link had a couple extra characters in it, resulting in a page not found. This one is fixed.

https://www.eurogamer.net/a...

[rainslacker]

It's interesting, except it doesn't measure what is actually considered input lag. If you have to display the actual result of the button press, then it's not a measure of input lag...it's a measure of how long it takes to display that a button has been pressed. The game loop doesn't run like this. The game loop will take the input that is held in memory on it's input poll, and then use that to calculate whatever it needs to, then it will render the frame. That input could be sitting in memory waiting to be read for any given amount of time before the program uses it.

"Lag" would be represented when the program itself doesn't get the input command due to a failure of transmission to the memory buffer which holds the users input command. If that takes longer than expected, or the input poll skips reading the data for some reason due to other in game loop issues(typically frame timing in games that use it), then it may present as input lag.

It shouldn't take long to measure how much lag there is between a device and the server which uses it. DF of all people should know how to do this. The average person can do this on their PC.

In all this discussion surrounding input lag from various services over the years, and especially more now, the only conclusion I can come to is that people are using the term wrong, or don't know what input lag is. The input lag that is going to be a problem with game streaming is what I describe above. The input won't be registered in the programs memory in time to be read by the input poll. It'll always be behind the current frame more than you'd have with a traditional console due to having to transmit that frame to the user. Whether it's a perceivable issue will depend on the game, and the users internet/internal network.

[crazyCoconuts]

DF's measurement of input lag is inclusive of network latency, right? Button press to observed reaction = game latency + network latency. You could get network latency by measuring this for both a local and streamed game (same game) and subtract. I think this is a better way to measure the added lag because there's more going on than just the latency around a ping. There's potentially resource scheduling and other things that could add to the delay. The way DF is measuring makes sure it's all accounted for in the aggregate. If the point is that Google could make up for network lag by reducing input lag with really fast servers, I think that's a fair point.

[rainslacker]

Not sure I explained it right. The latency that DF is measuring is the time it takes to record when the input was pressed. It is like saying that input lag is based on when the button is pressed, and when that input press has a perceivable action on screen.

However, the game loop itself doesn't work like this, and input lag itself is merely the time it takes for the input to be recognized by the program.

So, lag can be caused by two things.

One, the time it takes for the input to be recorded in system memory by the program(or system) itself.
Two, the time it takes for the input to be registered by the program for it to be used within the program itself.

Generally, only two is relevant to the discussion when we're talking about traditional consoles. However, with pushing the input off site, both one and two become relevant, because the amount of time it takes for the input to be recorded within system memory is going to take significantly longer than what you have with traditional models.

I don't know if it's possible to record input lag 100%, but if one is relying on a visual cue, and high speed cameras(which themselves introduce lag), then it's not going to be accurate. It is possible to measure input timings using hardware monitors, but I'd imagine that sort of thing is well beyond the scope of this, and just measuring the time it takes for that data to be sent through a remote connection is more than enough to measure general lag time. A simple ping operation is enough to measure lag time, then you just remove anything that happens internally to the software your using, and anything that deals with the return data timing.

It's not really rocket science, and has been done for decades before DF and others made up all sorts of ways to measure it in ways it was never meant to be measured.

All that's important is if it's perceivable to the user. 160ms would certainly be perceivable. That's almost 2/10th of a second to register that a button has been pressed. that's 5-6 frames of data that are drawn before it even recognizes that you pressed a button. More likely, the actual ping was much lower for the purposes of determining input lag, as much of that time would have been transmitting the data back However, anyone that uses the internet will know that lag will vary not just from day to day, hour to hour, or minute to minute, but from packet to packet, as networks don't work by averages, they work packet to packet. Games work in a similar fashion, with each set of data being something to process into something on every game loop....a loop which runs every 16-33ms.

[crazyCoconuts]

"A simple ping operation is enough to measure lag time" - I think we'd agree that the perceived lag attributed to streaming is more than just the network ping time. The rendered frames have to be compressed, transmitted, and decompressed before the user sees it. So as primitive as high speed cameras are, I can't see measuring how effective a streaming solution is end to end any other way...

[rainslacker]

I think the measurement of such lag is wasted effort either way. There are just way too many variables involved in routing internet data, general ISP's reliability and services, and people's own competence/accessibility on their own internal networks or wi-fi networks they may use to access such services.

It's nice to say, "Oh, we measured this and we got this ping", but it doesn't mean much in terms of real world results. It's an interesting experiment that just doesn't tell you anything more than what kind of "lag time" DF recorded.

Unfortunately, any company can make claim to what they expect results to be on their end, or that they intent to achieve such results. But, they also know that they can make such claims for the layman, because the lay man doesn't understand how things work, and for the most part, they can't be held accountable, because it's very likely that people will achieve such results. Sony, MS, and Google can sit there and say, "Yeah, we can reduce lag on our end", and be 100% legitimate, because this is what network designers do. They aren't representing the truth if they say that they'll reduce the users lag though, because they have no control over that. Google has the advantage of controlling many internet backbones, but the backbones aren't the issue.

[AnthonyDavis]

That’s unacceptable. Google says they targeting 60-70ms extra latencey for DOOM. That’s garbage. That’s like playing on a big slow modern HDTV.

[Shikoku]

This is just a rehash of Digital Foundries YouTube content, nothing new to read here honestly.

[Baza]

Google’s claim of native 4K 60fps is a joke. 1080p and will drop down 720p when your connection lags. Didn’t realize we were calling 720p next gen

[ThinkThink]

1080p and 720p are certainly not next gen so if that's what google is coming out with in 2019, it's a problem. As long as they still offer the 4k/30/60 for those with faster internet, you can't really fault them for trying to push the internet speed forward.

[Tross]

Shhh...Google just wants you to look at the novelty of what they’re doing and ignore things like performance issues. Thinking for yourself is old fashioned in their eyes. Just sit back and let them tell you what the future is, because we should always trust corporations to tell us what’s right for us. /s

[AizenSosuke]

Lag just like XCloud's first demo oof this is going to hurt

[WelkinCole]

There is no magic in the world that will automatically make everyone that already are playing consoles in the wide world be able to get fast/stable enough internet for 4k gaming.

So companies like Google and MS should really stop trying to fool people in thinking that streaming is the immediate future for gaming. The media also needs to stop and be honest with people they are reporting these info to.

Its like the whole power of the cloud thing again with extra power sauce crap.

Gaming is not like movies or music that does not require input from participants most of the time. Its not just the bandwidth but distance from datacenters for lag that gaming has to deal with which is more critical

I am all for it as an option but people really need to honest with streaming. It will take a fundamental change in our existing global internet infrastructure for streaming to able to replace traditional console gaming globally.

Sony's approach to PSNow is the best. They make sure that traditional gaming is still the primary focus as thats what a responsible global brand should do that caters to gaming audience globally. Not lie to people

[ThinkThink]

Nobody, including Microsoft or Google is saying that everyone in the world will be able to stream 4k. Everyone understands that certain regions are behind when it comes to internet speeds. That doesn't mean that they shouldn't try. People were saying the same thing when Xbox live first launched. There wouldn't be enough internet bandwidth for multiplayer. We got through that together, we can get through this together. :)

[opc]

"Everyone understands that certain regions are behind when it comes to internet speeds. "

The people here think that in order for Google to make a profit that they need to sell this service to every living person on earth.

[ThinkThink]

@opc,
Exactly. The average global internet speed was 7.2Mbps in Q1 of 2017 and it's been growing at a 20-25% rate each year. No reason to think the average won't be 10Mbps by 2020. And that's just the average. Goog's can cater towards the high end of that and make bank while the rest of the world improves their internet.

[Realms]

No they don't need to sell it to everyone but the sensationalist media and those cheering google on need to check themselves streaming games is far from the ideal way to play games and that will continue to be the case for a while.