Quote: Original post by curtmax_0
Sirisian:

For puzzle games and Myst there is no point then. The hardware to run games like that is trivial.

I have the same thoughts as that article. I think it's impossible on many levels, but I do hope I'm wrong and that it works, because it's a pretty neat idea. If it worked it'd be fun to take while traveling for something to do.

I believe Myst switched to using 3D worlds a while back. Also a lot of people are using netbooks recently. My friends at my university carry them around. They have a small resolution and I always see them playing really old games on them. They've mentioned that one reason they don't like them is because they can't play games that require actual system requirements. So something like On Live might work in this situation. Especially since my uni has basically no limit wi-fi.

Quote: Original post by owl
Maybe someone can put all that remanent analog equipment to good use. :)

The problem is twofold:

1. The analog spectrum will be re-aportioned to other uses once the "switch-off" happens, so there'll be no OTA spectrum for analog signals anymore, and
2. You can't multi-cast analog signals (that is, each client would need a dedicated piece of the spectrum, or a dedicated wire from the server to the client).

Besides, I'm not really sure which problem going analog would solve?

Quote: Original post by Codeka
Besides, I'm not really sure which problem going analog would solve?

Compresing the video signal, sending it through internet, decompresing it. I'm not really sure if broadcasting an analog video signal is faster than that.

Quote: Original post by ElectricVoodoo

Quote: Original post by Andrew Russell
The biggest question is - will anyone buy it? This reminds me of movies-on-demand that was big 1-2 years ago. How did that turn out?

Pretty well, actually. You can do this from your 360 now, but it's a noisy beast. My cable service, however, provides movies on demand with no problems at all.

The comparison however isn't great;
- video on demand means I select a film/tv show, send the request and the cable company stream it back to me (in the case of the UK after a bit of a delay while they show you a 'thanks for watching films with us' style thing, maybe to buffer up a bit). Easy, job done.
- this game tech however has one important thing; user interaction.

Without client side prediction/responce I'm just now sure how it's going to work; for any action you are going to have to wait.

I would say that something like Crysis is playable at 25fps, which is 40ms between frames, indeed I would say for any game 25-30fps is the lower limit for interactivity. So right away you need to be LESS than 40ms away from the server you are playing on, then you need to tack on rendering time and logic updates, the problem here being that you'd need a reasonable chunk of time to also carry out the updates and rendering as well and for a 720p or 1080p at, what I assume are going to be high settings, this is a non-trival amount of work in the order of 33ms if they want to maintain 30fps rendering.

The other problem is the hardware todo this is also non-trival; the machine I recently brought which set me back around £1,500 can support ONE person playing Crysis at 1080p and Ultra settings at 23-25fps, if I drop the setting to high then I get around 30fps. CPU isn't the problem here, it's taking barely 25% of my cpu time, the problem is GPU resources; doing this maxes an 4870 X2 graphics card totally and this thing is FAST.

Now, granted Crysis is an extreme example, but the numbers don't look that much different as you go down the scale and when it comes to XBox or PS3 games the problem of hardware crops up again.

And lets not even get started on the internet as a whole; dropped packets, iffy routing, dodgy connections. I don't see latency getting any better either as ISPs sell on bandwidth not latency. So while we might have 100mbps in the UK by the end of the year I don't expect the pings to have changed for the good by then either.

I'm pretty sure it's all the same hardware. I doubt they'd have lines of PS3s and 360s serving up content.

In a video I saw the guy said it took 2-3 weeks (or something close to that) to get a game to work with OnLive. That's a bit more than just pointing something to an executable. Sounds like each game needs some custom changes to the codebase.

Anyways, as I said before: I can't imagine this working, but I hope it does.

All valid points, however even in a 30fps FPS, you're average input latency will be around 45 ms due to the fact that your input can't be shown or processed until the next time frame atleast (30ms until next frame flush + 15ms on from the last frame where your input fell in the frame time on average), and this doesn't even take into account the buffering done by the video device which could be 1-2 frames (so it could actually be even greater than 45ms ). Most people can't tell if things are happening 100ms behind their inputs, there is an exception though for full camera rotation in FPS, for some reason if there is excessive latency on that, it's quite noticeable and disorienting (i guess due to the full screen movement and latency in screen stopping).

Full round trip pings of 50ms or less isn't really uncommon these days, if Onlive can distribute enough datacenters around to get their average round trip pings to ~50ms they can probably provide performance imperceptible from running a local game. They can use the same tricks FPS games use to reduce lag for players and run the servers at higher update rates (ie 100fps or more) to reduce the natural input latency( 30fps->45ms, 100fps->15ms, 500fps->3ms ) and floating framerates (whatever the video card can keep up with). This is only if the game supports this, probably not the initial ones, but eventually more will if they intend to support Onlive natively.

Bandwidth and computational power is getting cheaper, eventually it will become cheap enough for a service like this to succeed. Maybe this will be the service which does.

Since their technology is based on lossy video streams, dropped packets won't be a problem, it will just show up as a few block compression errors which are probably unnoticeable. They said they will do video scaling depending on bandwidth as well, so they can scale down the video stream dynamically and you probably wouldn't even notice.

-ddn

Quote: Original post by ddn3
Full round trip pings of 50ms or less isn't really uncommon these days, if Onlive can distribute enough datacenters around to get their average round trip pings to ~50ms they can probably provide performance imperceptible from running a local game.

But you're assuming they can process the inputs and render the frame instantly. The latency of the network is in addition to the "natural" latency of the game. So if their servers run the game at 30fps, and you've got a 50ms ping, then that's actually 45ms + 50ms = 95ms before the response to your input comes back. That might not be impossible, but that 50ms is very optimistic, IMO.

Yeah but there are a few things you can do to reduce that latency, you can run your local update rate at say 1000 ups ( so you only have 2 ms input latency locally before u flush the packet) and run your servers update rate at the highest you can ( they say they run the games at 60 fps server side ) which reduces the natural input latency to 2ms+22ms=24ms. Then there is the round trip ping ( lets say 50ms for argument sake ) and video decoding (lets say they only show video at 30fps playback ), that's 30 ms there. So all together from button press to bullet fire is 50+24+30 ~= 104 ms. That's within the threshold for most people, as to be consider instantaneous, maybe not hardcore FPS player trained on 25ms lag, but for average gamer who is playing Madden 09 it's good enough :D If they can do 60fps video decompression you can subtract another 15ms from that.

Most games will run ok within this system, I suspect.

-ddn

http://www.engadget.com/2009/03/25/video-onlive-streaming-game-demonstrated/

Agreed, there are issues here that don't apply in video on demand. The problem of processing/rendering is not trivial, but the average 360 is on for maybe one hour a day, server side hardware is always on, which splits the hardware cost some, and in an 8 player game you no longer have to run the simulation 8 times locally then merge the results, you just run it once on the server, another saving.

As for lag, they claim that their radical approach to video compression is what makes it a realistic proposition, and as long as you are within 1000 miles of the server you're ok. At around 44.30m in the video linked above they claim that the ping is sub 1ms!

Quote: Original post by ElectricVoodoo
..and in an 8 player game you no longer have to run the simulation 8 times locally then merge the results, you just run it once on the server, another saving...

surely that would require major code changes to each game?