Stadia and the Ghost of OnLive

Stadia and the Ghost of OnLive

Tech News game streaming gdc google stadia internet caps onlive

So yesterday at GDC, Google stood up on stage and told the world it's going to put every game inside a browser tab. Phil Harrison walked out, showed off a controller with a little glowing dot on it and a dedicated "capture" and Google Assistant button, ran a demo of Assassin's Creed Odyssey streaming into a Pixelbook, and called the whole thing Stadia. No price. No release date beyond "later this year." Just a name, a controller, and a promise.

I've been sitting on this for about eighteen hours now and I keep coming back to the same place: I've seen this movie before.

Anybody remember OnLive? I do, because I actually paid for it for about four months back in 2011 or so. Same pitch, basically word for word: the game runs on a server somewhere, gets compressed, gets shot down your pipe, and your local box just has to decode video and send your inputs back upstream. No console, no download, no install. It was genuinely magic the first time it worked. It was also unplayable the second time your neighbor started torrenting something, and the company folded in 2012 and got picked apart for parts by Sony. I don't say this to be a downer about Stadia specifically, Google obviously has more server infrastructure and more money than OnLive ever dreamed of having. But infrastructure was never really the whole problem. The problem was the last mile, the actual wire running into actual houses, and that hasn't magically fixed itself in the years since.

Which brings me to my own connection, because this is a blog and I get to make it personal. I'm on a plan with a monthly data cap, and streaming a game at anything resembling a decent bitrate for a few hours a night is going to chew through that in about a week and a half. I did rough math on my phone during the keynote livestream and got somewhere around 15-20GB an hour depending on resolution, which is completely unworkable for me and, I'd guess, for a whole lot of people who don't live somewhere with real fiber. Google can build the fastest data centers on the planet and none of it matters if the pipe between the data center and my living room is the bottleneck, and for most of the country in 2019, it still is.

The other thing nobody on stage really addressed is input lag, and this is the part that actually worries me more than bandwidth. Local rendering means your button press and the pixels on screen are separated by one frame, maybe two. Streaming means your input has to travel to a server, get processed, get encoded, get sent back, and get decoded, and every one of those steps adds milliseconds you can feel even if you can't always name what's wrong. Fine for something turn based or slow. A genuinely bad idea for anything twitchy, and Doom Eternal, which they also demoed, is about as twitchy as games get. I'd want to actually hold a controller and play the thing before I believed any claim about latency, because "it felt fine on our optimized demo network at our own conference" is not a sentence that has ever once translated into "it felt fine on my actual internet."

None of this means Stadia is doomed, to be clear, I'm not trying to make a prediction here. Google has the money to subsidize this for years if it wants to, the way Amazon subsidized shipping money-losing packages for a decade before Prime made sense on a spreadsheet. Maybe the compression tech really is better now. Maybe 5G rollout closes the gap faster than I think it will. I just remember writing enthusiastic things about cloud gaming on this very blog back in like 2011, and I remember how that turned out, and I'm not in a rush to write the same post twice and be wrong about it twice.

I'll believe the input lag claims when I've got a controller in my hands and not a second before. Until then it's a nice stage demo and a controller with a screenshot button on it.