Hacker News new | ask | show | jobs
by darth_avocado 1359 days ago
I’m surprised, but I’m also glad they are doing this. It could be to avoid class action lawsuits. I used mine for a total of 5 minutes before throwing it in trash. It is a very unfinished product they shipped thinking they’ll solve it. But the reality is, even with the best internet in the country, the games were barely playable. I’m talking 600mbps download and a 100mbps upload speed.
6 comments

I've used Stadia for the past year exclusively and it's been fine 99% of the time. I guess I'm relieved from defending Stadia duty now though, sigh
The problem with such statements is that game streaming services are INSANELY dependent on literally a century of cruft and how it was handled on a house to house basis. You can have great performance in your house, but your neighbor across the street could have utterly useless behavior.

Like this product literally depends on which godawful modem your ISP sent you when you first got service.

Speed alone isn't what matters here - latency and jitter are more important. A 100Mbps speed test over 30 seconds is meaningless.

I've played multiplayer FPS games on a home-made setup with an AWS VM with GPU and Steam streaming (using a VPN to make both machines appear to be on the same LAN so Steam streaming would work).

This worked well, but only because it was on an enterprise-grade leased line with consistent 1ms latency to the AWS datacenter, and all wired (good wireless gear might've worked too, but forget about trying that on garbage consumer-grade hardware like your typical router or mesh Wi-Fi setup).

Is it technically possible? Yes and it works well under optimal conditions.

Is it possible for the average user who doesn't have good equipment nor the budget for it? No chance - it's a recipe for disaster. Those who do have the budget are better served by just buying a gaming machine and running the games locally.

Games streaming can be a value-add to a good ISP (such as Google Fiber) whose network actually permits this, but don't expect it to work on the majority of residential connections. The vast majority of them suck (whether because of the ISP's network or the customer-premises equipment), people don't know they suck and have no easy tools to test that, so they'll end up blaming the game streaming provider when it inevitably doesn't live up to expectations.

Until good networking setups become commonplace, game streaming will remain limited to a very small niche that have serious networking setups but for some reason don’t have a local gaming machine.

Game streaming is great for casual gamers. A lot of games are perfectly playable even with 200ms tacked on, actually.

It's unacceptable even with a 1ms link (because of the extra 2-3 frames of latency that get buffered in) for hardcore players in some genres. Even if they can't see the difference, they'll feel it when they miss shots in FPS games and links/confirms/parries in fighting games

Unfortunately, most of the people here and in the industry making these streaming products are adults with real lives who don't understand how bad game streaming is for hardcore players

200ms?? it's really frustrating in my experience for every game (I had 180ms when Xbox Cloud Gaming connects server over pacific ocean for unknown reason).
Yeah, a lot of people play games with their TV not in game mode, which is a ticket straight to +300ms City

Not only that, but a lot of AAA games nowadays have super long animations, tons of post-processing slapped on the tail end of the rendering engine, etc, so you end up with 300+ ms from "button pressed" to "something happens"

I suspect most of those games played on TVs would be running on consoles which are much more forgiving as their games are optimised for that use-case and there’s built-in assistance for inputs (aiming with controllers is much more difficult without it).

My understanding is that Stadia and most other game streaming providers run PC games which are developed with the assumption of precise mouse/keyboard inputs where there’s no assistance.

Anytime I see an asymmetric upload bandwidth like 600/100, I assume the ISP is just advertising temporary burst speeds and does not actually allocate enough upload bandwidth to the neighborhood for people to sustain usage at 100Mbps.
Upload bandwidth for something like Stadia is tiny. Only thing you need to send are user inputs
However, you need consistent latency, which isn't guaranteed in a highly-oversubscribed network.
It actually just means they're using DOCSIS to carry the signal, which has asymmetric bandwidth allocations for upstream and downstream. 600/100 is a standardized allocation too.
In practice, it is always a heavily oversubscribed network that never delivers sustained bandwidth for either up or down.

Contrast to whenever I have used a symmetric fiber connection that advertises 1Gbps/1Gbps, I can actually sustain close to both of those and at sub 5ms latency. Whatever the theoretical promise is, I assume non fiber non symmetric connections are simply low quality (in the USA).

You could make the same argument regarding upload speeds. They simply have asymmetrical link, and overprovision on both download and upload.
I assume you mean same argument regarding download. In my experience, the download is always far less over-provisioned than the upload.

For example, Comcast over-provisions their upload so much they cannot even advertise what it is. They will sell you 2Gbps download and never tell you the upload. Which I assume, based on experience, is 20Mbps split over a neighborhood of 500 houses.

Bandwidth isn't that important with game streaming after ~40-70 Mbps, latency and jitter (essentially latency consistency) is.
Consumer-grade Wi-Fi is also a major problem when it comes to latency & jitter. It doesn't even have to be game streaming, any real-time application such as calls suffer from it as well, despite not actually requiring much bandwidth at all.

Unfortunately there is no user-friendly tool to test for this. Most tests focus purely on speed, which can be tricked by various packet-loss-compensation algorithms, so you can score a "perfect" 1Gbps speedtest despite the connection cutting completely for a second.

speedtest.net used to have a sibling "pingtest" site that measured your jitter. I'm not sure why they don't exist anymore.
I remember it using a Java applet. I think the reason none of the online test sites support it is because it’s hard to test latency & jitter in the browser as the lower layers try hard to compensate for it.
Oh. It was Flash.

Sometimes I forget that was ever a thing!

Speedtest for random server (servers listed on Ookla is quite random) is useful but ping for random server is a bit useless. Just ping for targeted server that runs service you use.
I'm somewhat surprised the 4 sibling comments as of this writing don't even mention the latency/jitter issue-- to me, that's always been one of the obvious biggest flaws with game streaming. Your average consumer has little to no awareness of it, it's beyond Google's control, and it has a very noticeable impact to anyone experiencing it. Not a good combination.

Edit: Nextgrid showed up as I was typing this and set the record straight. My faith in HN is restored.

It would be great if they could somehow open up the API of the controllers, they are nice.
I'd be shocked if their contracts/EULA wasn't structured to avoid risk of suit around something like this. Shutting down a live service feels pretty defensible as not a crime or tort, and they could almost definitely fight the lawsuit for less than this costs in refunds, which makes it all the weirder.
It's the most likely reason. We've seen plenty of cases were EULAs were declared void and that won't hold in a place like the EU. You can't sign away your rights as consumers here. They might be able to fight individual lawsuits in some places, but it might eventually escalate into an investigation by the EU. There's significant legal risk there that is being avoided by just refunding a few millions. It's the sensible move.
Honest curiosity, has that actually been proven in EU court? The sort of "licensing as a service" that Stadia did doesn't seem that different from the business models of something like Audible, or even iTunes in the DRM era. I totally agree that it's a predatory and anti-consumer model, but I wasn't aware that anywhere in the EU had successfully argued that removing access to something you were essentially "renting" access to was a violation of consumer protection laws.

These sorts of EULA arrangements are essentially the foundation of almost all modern media consumption - if anything Stadia is on better ground than most since it's not just a DRM layer like Steam or iTunes. If Steam disappears, I have a bunch of entirely playable game files on my computer that I can't use. When Stadia shuts down, you have a client for nothing. You're not paying $60 to own a copy of a game, you're paying $60 for an unlimited term license to play the game on Stadia's servers. Legally, it feels odd to claim in court that that should be the same as a purchase of the game in some other method. If Google had turned off Stadia, but transferred everyone's purchases to Steam or EGS, that wouldn't be providing the same service you purchased from Stadia.

It's also just a good marketing move. "they made people pay full-price for games and deleted them shortly after" is the kind of association that sticks around and even Google has an interest to avoid.