80 ms latency @ 60 fps is about 5 frames, which is a significant latency on an FPS.
I guess if you wanted to play a game on max graphics on your underpowered work laptop, this is a (pretty cool) solution.
But I don’t think a lot of people have this problem.
I’ve played Fallout 4 on AWS using the methods described on the blog previously:
I live about 4km away from an ADSL2+ exchange in Melbourne and am able to get about 30ms to Sydney’s AWS region. 30ms of latency in a single player game like Fallout is pretty good to play with.
I do think 80ms might be a bit high for a game like Overwatch, not that I’ve ever played it.
How does it compare to playing locally? I’m mostly interested in artifacts introduced by compression. 1080p30 is already pretty bandwidth intensive. 1440p60 (or 2160p60) would require considerably more.
I’ve never played it locally. I don’t have Windows nor a computer powerful enough to play it. Living 4kms away from an exchange means I don’t have much bandwidth. I get a lot of compression artefacts.
[Comment removed by author]
I’m certainly not good enough to compete at a level where a few FPS would make a difference.
Don’t think of it as losing out on 5FPS, but losing out on 1/12th of a second of time. It’s not really FPS frame skip (a.k.a. “lag”) the way you’d normally experience it if you had, say, an underpowered GPU that couldn’t keep up with game rendering.
Essentially, after every input, you have to wait 1/12th of a second before the game visually responds to that input. Depending on your age (the older you are, the less susceptible to these delays you become), this can be very obvious, or not. It can really screw with your brain and its expectations, and in some cases induce nausea.
For interactive games, it’s highly recommended to keep input response latency under 33ms, so 80ms is really pushing that boundary.
Recommended by whom?
John Carmack was talking about 100ms for VR, which has extremely hard latency requirements.
1) Empirical research. A lot of time and effort has been put into latency and visual perception as it relates to HCI. Here’s a good meta-study of over 150 papers.
Studies on human performance in VE show that people are generally able to detect latency as low as 10–20 ms.
(Section F. Time Delay; Page 1236 [Page 6 in the PDF])
2) John Carmack was not talking about 100ms as being acceptable for VR. He was saying that existing headsets prior to Occulus Rift had 100ms+ latency, and that it was completely unacceptable. Carmack himself has set a bar of 50ms as barely acceptible (and 20ms as desired). Source.
As I understand it, Overwatch is very simply a multiplayer shooter. So the value of actually owning it and playing it offline or in a LAN seems minimal. It makes me wonder if the makers of Overwatch could turn completely-remote play of Overwatch into a service.
(There is no LAN or offline play, so you can probably change ‘minimal’ to ‘none’ or ‘undefined’.)
Interesting, but I’m pretty sure many of these tweaks are unneeded. If you can use RemoteFX, you should.
Hhm definitly trying this out, I’d really like to be able to try out the witcher 3, but I really dont have the specs
It was pretty easy to set up, but unfortunately my internet cpnnection does not seem up to the task and while the game was beautifull, it was extremely choppy due to latency.