Discussion in 'Article Discussion' started by bit-tech, 10 Oct 2019.
"More responsive"? I've been playing MMORPGs since early 2007 (WoW first) and sometimes network latency can be up and down like a trampolinist, so I don't believe they can deliver on their claim, certainly not all the time. Too many factors, beyond Google's control, affect latency between host and client, possibly thousands of miles and many nodes apart.
Did you read the bit of the article which points out that not only is it possible in theory, Microsoft literally did it in practice in 2014 - and even let you download the binaries to try it yourself?
Yes and my experience of MMORPG playing is that latency can be very erratic and sometimes go to 2 seconds plus...
Whatever Google, or others do, they cannot control things end to end.
The claim is frankly silly: the same "render potential outcomes and then select the one that matches future inputs" technique can be used just as well with local rendering as with remote rendering. The frame-resampling portion is already implemented for VR, in a more advanced way than the MS paper (allows for compensation not only for camera angle changes, but for camera motion and in-scene object motion due to the use of motion-vectors for forward predication in addition to the depth buffer).
Does this mean you could use doing something unpredictable like a lag switch?
That's rarely network latency - if things were that bad, you'd never be able to have a decent Skype call. When MMOs hitch like that, it's usually 'cos the server is overloaded - an issue from which, you'd hope, Google with ALLTHEMONEY wouldn't suffer.
This 'ere measures latency and jitter - the latter being the amount by which the latency varied during the test. It says my connection to a server all the way in New York has 80ms latency (well with DeLorean limits for being entirely removed) and 3ms of jitter (which wouldn't take it outside said limits.) At no point during a few repeated tests did the jitter ever go higher, nor did the overall latency ever hit 2 full seconds(!)
Long story short: don't blame network latency for Blizzard's unwillingness to invest in infrastructure.
All sorts of ways to exploit lag-compensation.
One example: give yourself 100ms of pseudo-lag. Snoop the incoming predicted frame variants. Because the frames must arrive before you perform the input that selects which frame is 'real' (because that's fundamental to how the latency reduction system works) you now have 1/10s lead time in which you could have a NN sit and analyse incoming frames to identify the ones that show (e.g.) a headshot from the ones that merely are a near-miss, then perform the input that selects that image.
I don't see how the amount of money google has matters for this.
Look at it this way:
The basic versions of the Xbox / PS4 are around £200 new (and the Xbox could even be considered a 4K blu ray player with free gaming functionality).
Now if Google chose to throw extra money at Stadia to reduce the latency problem... you can see where this is going, they would inevitably raise Stadia prices to recoup the investment making the value proposition compared to "normal" gaming even worse.
Google probably already has the spare capacity in GCP to not have to throw a great deal of money at this, at least initially.
This obsession with streaming games down our phone lines/fibre has to end.
...because it's not going to have to scrimp on the servers to boost its bottom line? Especially as...
I'm not sure that's the case. There will be hefty profit built into the current subscription - remember, companies love subscriptions because only a fraction of your paying subscriber base is going to be active at any given time. 's how gyms stay in business. That and making it really awkward to cancel.
Should be 'its Stadia platform', right?
Yup - fixed, ta!
Separate names with a comma.