Hacker News Reader

Comment by ManuelKiessling | Hacker News Reader

Honestly I just don't get why every single time there is discussion about game streaming, the topic of lag as a potential show stopper comes up, and every time it's a hypothetical discussion, which doesn’t make sense at all - the solutions are out there! Don’t assume - just verify it.

And when you do, you will find out that there isn’t any actual issue. I’ve played games like The Witcher 3 via streaming over three years ago.

Input/output lag simply isn’t an issue, and I’m your average German VDSL user, nothing special.

From my point of view, there is no sense in any further discussion - the empirical data is already in.

View on HN

You've specifically chosen a game that isn't sensitive to lag.

It's like saying "Switching from a car to a bike just isn't an issue - I once rode all the way to my neighbour's house!"

Try playing a multiplayer first person shooter on the same service, you'll find latency is an issue.

By all accounts, the Doom demo that Google gave was plagued with lag.

Yeah, speed of light is actually a lot faster than people think. Honestly things such as controller and display lag are probably bigger than the round trip latency. And experimentally, anyone who has tried Stadia at GDC or I/O has said that latency wasn't an issue.

The bigger limiting factor I think will be bandwidth, especially in America with the awful internet provider situation.

"Yeah, speed of light is actually a lot faster than people think"

This would only be relevant if you had a direct fibre link in a straight line from your house to the service. It's packet switching\routing etc. that introduce latency. And even if you DID have direct fibre, the worst case scenario for distance - server on the other side of the world, but with no other latency accounted for - is about 200ms, even half of which would be noticeable in many games.

I'm not saying it's impossible to do game streaming, but you can't just divide the distance by the speed of light and say it's low. That's not really the issue.

Most of those hops would be on ISP equipment with 10G or 100G interfaces, modern routers forward packets in tens to hundreds of microseconds and switches can do it in a few microseconds, so as long as there is a bit of distance it will account for most of the delay, and SOL is 2/3 of its speed I'm vacuum. Switches can even sends out packets on a port before it has been fully received on the incoming port

> modern routers forward packets in tens to hundreds of microseconds

Only when network is underutilized, and there is no throttling/bufferbloat.

True, though routers with 100G interfaces do not have a lot of buffer space. But most ISP's make sure that isn't happening as it its pretty bad for their customers experience. I know some ISP's do not put enough effort into that at all but every single one I worked for did (Europe). In this case Google can put effort into direct peering relationships to make sure that their customers have a good experience.

Right, which is exactly big Cloud companies such as Azure and GCP have a huge advantage over most other smaller companies, as they have their own dedicated backbone, allowing them to minimize the number of hops.

> server on the other side of the world, but with no other latency accounted for - is about 200ms

In practice, it’s a lot worse than 200ms. 500-700ms are typical for servers on the other side of the planet.

That definitely varies greatly from provider to provider. I usually see <300ms for world-wide pings from Texas from a FTTH residential connection.

Average Pings from Digital Ocean regions: BLR1: 251ms FRA1: 129ms SGP1: 206ms

Other sites: yahoo.co.jp: 179ms rakuten.co.jp: 138ms belarus.by: 167ms

As most things go, YMMV.

> Honestly things such as controller and display lag are probably bigger than the round trip latency.

The problem with latency is that it adds up, lag from the server, lag from the TV, lag from my brain, lag to my fingers, lag from the controller to the machine, lag back to the server, lag in processing and then start again.

Client-Server lag might be acceptable in isolation but not as part of a system.

There are press articles talking about how latency is an issue when they tried it out: https://www.pcgamer.com/i-tried-googles-stadia-and-latency-g...

So not really anyone. And that's in an extremely latency friendly environment where the servers are just up the road.

Yep, I played a couple of the Arkham games on streaming services, the first one way back on OnLive before Sony bought them. It was fine.

I now play games with some friends and while I'm using a local PC, they use Shadow PCs (ie a VM with a GPU in a datacentre vaguely near them) and it's fine.

Maybe it'd be worse on an LTE/5G connection, but my phone is getting 25ms to the nearest Speedtest.net instance right now, which doesn't seem substantially terrible.

I would be much more concerned about the pricing of all of these things, than the technical issues. How many gaming subscription services am I going to need to add to my video subscription services?

Mind sharing what the solution you referenced?

From a hard theory network perspective, having higher bandwidth, lower latency, lower error rate on transmission path combined with closer geographical proximity to servers seem to help mitigate increases in latency.

However, if you argue the problem is solved by your empirical evidence gained by playing Witcher 3 of all games, I disagree with your point. Play a real time strategy game, a MOBA, or a first person shooter. 70ms~ (a report for Stadia reported 70ms-130ms, may be stale data) of lag is noticeable to people. On the other hand, I don't mind streaming a turn based game.

I had rolled my own setup with a GPU-powered AWS EC2 instance, streaming the video signal via NVENC etc.

Witcher 3 is a single player game and the users who care about lag by far the most are multiplayer gamers.

Certain games like rhythm games would also be pretty miserable with jitter...

Reply on news.ycombinator.com