OnLive - Why Linux Gamers Should Take Notice

[Kevin]

http://image.examiner.com/images/blog/wysiwyg/image/OnLive.jpg


Some of us Linux users have been waiting for a very long time for Unreal Tournament 3 or for that matter, any big name titles coming to the Linux platform.

GDC (Game Developers Conference) is happening right now and there has been one huge announcement that has taken the internet by storm accross every major gaming/tech news site.

A new startup has announced a new product that allows gamers to run games on their computers or televisions without actually running the game on your hardware. Basically, how it works is that you get a very well set up Xbox Live/Steam like community page in which you can rent or purchase games and it will run the games on their servers and stream them to you, the computer user.

Sounds like a concept that would work horribly? Well according to IGN and others who got to play Crysis with max visuals on a Macbook Air, it runs nearly perfect and is very impressive. IGN is already calling this the death of PC gaming as we know it.

Check out what they have to say here:

http://pc.ign.com/articles/965/965535p1.html

Anyways, it's supported in Windows and Mac but no word on any Linux version (as always). However, apparently it only requires a simple web browser plugin to work. I don't know about the rest of you but I certainly plan to convince OnLive to support the Linux platform. Their website goes live later today and can be found here:

http://www.onlive.com/

Videos:

http://kotaku.com/5181625/see-onlive-in-action

This is a huge deal and could mean that Linux gamers finally step on Microsoft's face and finally get the gaming fix they have waited so long for. I've been reading up on this technology and reactions from those who used it and to me it sounds like it's going to be a huge thing just as Netflix is taking movie rentals to the next phase.

[Zhick]

So yeah... instead of the users onLive purchases the Microsoft Licenses... The games still run on Windows-Boxes.
Sooo... what did you say again about stomping into Microsoft's face?
This is just the same thing as a VirtualMachine or DualBoot and has nothing to do with Linux-gaming. I'm not the least bit interested.
BTW I think I've read about another service doing pretty much the same somewhere here on the board.

[Kevin]

Well if they got it to run on Mac then I don't see why Linux gamers won't be able to use this.

[RealNC]

This is BS. You can't play through the net. Lag anyone?

IGN is full of hype.

[Kevin]

^ Others have reported that it worked as well. It's playable at GDC. Looks like everyone here is closed minded.

[Zhick]

Well if they got it to run on Mac then I don't see why Linux gamers won't be able to use this.
Yeah but at some point the games will have to run on windows-boxes (on onLive's servers (ugh windows on servers)). So it's still windows-games running on windows-machines and thus imho is unrelated to linux-gaming.

[Kevin]

It's related to Linux gaming if you can play the games without owning a copy of Windows.

At any rate, I thought Linux gamers who have been waiting for some Linux games forever, would find this to be a rather appealing service. I guess I was wrong.

[monraaf]

^ Others have reported that it worked as well. It's playable at GDC. Looks like everyone here is closed minded.

Maybe in a controlled environment, where the server is situated close to the client, yes. But in the real world, sending data through a series of routers all over the world takes way too much time for interactive gaming.


$ traceroute www.onlive.com

...

16 * 72.21.197.33 (72.21.197.33) 120.300 ms 120.823 ms

...


I'm skeptical.

[Melcar]

Yeah, I'm skeptical as well. Never mind that IGN has become a big ol' hype machine. Anyway, I fail to see how this would help Linux gaming at all. For the people that just want to play games, any games, on their non-Windows boxes this may be the start of something great, but when it comes to Linux gaming itself things like this are actually setbacks. If you really care about Linux Games, what you really want is native games being developed, or ideally, platform independent games.

[RealNC]

^ Others have reported that it worked as well. It's playable at GDC. Looks like everyone here is closed minded.

Close minded, no. Internet connections that lag during peak-hours, yes. So with this system, I would only be able to play when there's no heavy traffic limiting my 2mbit DSL connection?

Er, how about "no, thanks" :D

[Dragonlord]

I commented on this already on another place so I make it short: Huge pile of shit. Another hype-bubble to burst in the face of the people. On a LAN this is barely possible ( anybody who did Video Streaming knows what I talk about ). Furthermore to get this data across the wire you had to compress it a big time. So a small resolution and all riddled with compression artifacts? That's not gaming, that's a joke.

As mentioned, the solution looks different... and Cloud-Computing is not part of it ;)

EDIT: Oh and another fine one. Somebody hates game XYZ... let's go and DDoS the servers. Whoops... nobody in the world can play the game anymore. Old rule of system maintenance: POF... Point Of Failure. If all goes through one bottle neck it requires only this bottle neck to break down and your entire system breaks down.

[Svartalf]

^ Others have reported that it worked as well. It's playable at GDC. Looks like everyone here is closed minded.

It has absolutely NOTHING to do with closed-mindedness...

1) It doesn't HELP Linux gaming. Seriously. It doesn't help WINDOWS gaming either, if you want to get down to brass tacks. It's a walled garden much like a console, but it brings all the negatives of the vapor Phantom console along with latency concerns.

2) If you've never dealt with latency or scaling on something like this, you'd think they'd gotten a good answer. The problem is, all they're doing is providing a "slick" remote framebuffer and user interface device communications protocol. Seriously. Ever try doing VNC over high latency links? If you've not done so, you're in for a shock if you think that this will scale at all well or not have serious latency issues when the userbase gets larger.

3) There's always a lot of snake oil in the game dev industry. And all the big players jump on board "just in case". This isn't any exception to the rule.

Before you discount my remarks as being part of the "competition", I, not LGP, as a consultant to MANY companies over the many years, have been at doing development for quite a few massively distributed client/server systems in my day. This isn't the way to make gaming work.

[Jonsul]

I joined just to chime in my opinion.
Seems like it's too big an opportunity to miss to me. Imagine if anyone is wrong, and this works amazingly.
Also the game isn't streamed, but the video output is. Technically it's netflix with controls. Also they appear to have developed 'revolutionary' compression technology that near gets rid of lag. Yeah it's alot to take in, and may be crap. But if it's true it could be amazing.

Imagine all the netbook users who can play Prince of Persia, Crysis, or Burnout on their computers! It could change everything.

Plus it can't be garbage. EA, Ubisoft, Eidos, Take2, and Atari must have signed on for a reason. It's a big step to take, supporting a new system, and the fact that so many have signed up with their big name games means their must be something about it.
Honestly I've checked there isn't anything at all like this. No games are stored on the computer at all. Everything is done in servers and the output is sent to you while your input goes to them. It's completely windows and platform independent, all they need to do is have a program that connects to the stream and send out input.

And have you seen the microconsole for it! My MP3 player is bigger then it! Seriously this thing that plays Crysis at full speed is smaller then my hand. It completely takes out the hardware from the picture 90%

You all should really research it. Quite amazing if it's compression technology is as good as they say.

[Wyatt]

...Imagine if anyone is wrong, and this works amazingly....EA, Ubisoft, Eidos, Take2, and Atari must have signed on for a reason....

You know, I feel like you completely ignored Svartalf up there. See, look at 'em, I think he's gonna cry! You might want to apologise.:rolleyes:

Also, This cnet article (http://news.cnet.com/8301-10797_3-10202688-235.html) provides some decent insight: "...no lag, so long as their Internet connections meet minimum thresholds. For standard-definition play, that would mean a minimum 1.5 Mbps connection, and for high-def, 5 Mbps."

Now, I don't know if you've all seen what passes for "high definition" these days, but we've had computers, monitors, and games that have been able to do it for more than a decade now. Colour me cynical.

[Pfanne]

if it works, then: hell yeah!
if it doesnt: who cares?
no need to get all aggressive.

[Dosfish]

why not use http://www.streammygame.com, it even has a linux client, if you want to go down that road.

[kUrb1a]

Yea,I even tryed SMG and it's pretty impressive.but it was long time ago when i tried it.And why everyone making suck a fuzz about it.I remember when I saw this http://www.youtube.com/watch?v=Y4cUl9WFv7g for the first time!Nothing is changed.

[Dragonlord]

Revolutionary compression? My laugh. There's a certain threshold you can't go below and this is mathematically proven. It's vapor-tech but they know how to wrap it up in honey. You know the .COM bubble also had huge names on board and it burst because it was hype-shit. This is technical and practical nonsense. You can NOT nullify lag. And for those not in the know lag means:

Lag = Static-Latency + Dynamic-Latency.

Static-Latency is stuff like round-trip-time and processing overhead on the routers in between. This can not be avoided, is always there and is already quite a large number. Then on top of this we have Dynamic-Latency which is all sorts of unpredictable slow-downs which can happen all along the way. Sum this up and double the number as your input has to travel down ( 1-way ) and then the image back up ( 2-way ). Try to play a shooter with that much latency.

Note: Some might object that you can use video compression tricks like motion prediction and such. Problem is only that games are a highly in-coherensive experience. In game-dev we call this frame-to-frame coherence. Some stuff changes slightly over time but many things don't... and graphics do NOT belong in this category.

[RealNC]

This is pushed like crazy not because it works (it's over the net, it CAN'T work), but because of people not being able anymore to get warez games over p2p. That's what the companies are after. This ensures that you pay to play. If it works or not doesn't really matter, as long as they get more money.

[Jonsul]

You know, I feel like you completely ignored Svartalf up there. See, look at 'em, I think he's gonna cry! You might want to apologise.:rolleyes:

Also, This cnet article (http://news.cnet.com/8301-10797_3-10202688-235.html) provides some decent insight: "...no lag, so long as their Internet connections meet minimum thresholds. For standard-definition play, that would mean a minimum 1.5 Mbps connection, and for high-def, 5 Mbps."

Now, I don't know if you've all seen what passes for "high definition" these days, but we've had computers, monitors, and games that have been able to do it for more than a decade now. Colour me cynical.

I read, But what makes it different from the phantom is the fact that is has all the backing it has. It has funding, and a great library of games.

And it could be good for gaming on all platforms because it opens any game up to be played on any platform regardless of how it's programed. So not only does development cost go way down, but Linux user can finally get in the picture as well.

Well it's not just that these big names are supporting it, it's that they're putting their high profile games on it. There's brand issues, putting their top games on a platform that fails could hurt their profits and name bad. It's a big risk, so there must be something to it to risk their big name products.

Eh, I don't really care about high definition play. as long as it's at least Standard quality I'm good. Plus my connection is way more then 1.5 mbps so it's no big deal.

Hey all I'm saying man is that if it's true it'd be amazing!
I'm skeptical as well, but I'm hopeful this works. I know there could be latency issues, but as long as I'm playing Prince of Persia and Wheelman on my linux box I could deal with a little lag.

If it works, I'm set^^
If it's garbage it's garbage, I'm out what? $20-$30?
I'll chance it :D

Plus I've signed up to Beta test. I'll hopefully be able to test it out before I have to buy anything. And if I don't I'm sure we'll hear some leaked info on how fast the Beta is^^

[Svartalf]

You all should really research it. Quite amazing if it's compression technology is as good as they say.

Heh... I've researched stuff like this for decades- seriously. The Internet doesn't even remotely work the way you think it does and each connection adds to the picture- and it's not even close to a pretty one at that.

They're flogging this not because it works over the Internet. They're flogging it as an "antipiracy" play- and the game publishers are the same crowd that're doing the same insanity over "piracy" in the Music and Movies/Video space.

Here's a hint: Unless you've got the sort of link I have at my house, this will plain-flat NOT work because the speed and latency will not even remotely be there for you.

Forget Cable.
Forget DSL unless you've got a business link.
Forget consumer FIOS or similar. You're sharing the ATM backhauls with other people.

You will need to shell out $100-200 a month, minimum, for a business service, fixed IP set link to FiOS or similar at 5+ MBits to get the thing to have the same performance that they're showing at GDC right at the moment- and that's for the low-fi play. You'll need the 20Mbit symmetric link I happen to have just to realistically do the Hi-Def one over the Internet- and that is if you don't have issues on the backbone you're sitting on.

And this doesn't even get into what kinds of pipes the game company is going to need to provide service Add it up. It's not magic. An OC-12, which is "godlike" bandwidth in most people's eyes, can only sustain 415 simultaneous sessions to and from the server pool at LOW RES, it's roughly 83 of them at Hi-Def. That's over 600MBits/second. This sort of crap breaks companies REGULARLY, even in this day and age.

I've researched video compression tech for a long time.
I've researched network performance and latency for a long time.

I've held down jobs collecting the data from 8 primary securities trading markets daily- the full trading volume brutalizes OC3's effortlessly. This would savage an OC-12 like nobody's business.

I've been the CTO of a set-top box company trying to do on-demand video, etc. for years. 1.5Mbits of "low-def" video compression's nice, but it's not much different than MPEG4 rate with a few other things tacked on. I wish I had it back then- would have made a store and forward caching system for that sort of thing viable back a few years back- I would still be the CTO for the company, now effectively dead, as they'd have had quite a few customers in hand.

This is someone trying to make a play for investor money and bilk them out of it. Seriously. You can discount what I'm saying here, but in the end, what I'm remarking on is the brutal reality of what they're claiming and everyone's twittering about here. It's not magic. It won't do what you think it will.

[Zhick]

Has anyone paid http://www.onlive.com/ a visit lately?
Not Found
All your base are belong to us.
:D

[Svartalf]

You know, I feel like you completely ignored Svartalf up there. See, look at 'em, I think he's gonna cry! You might want to apologise.:rolleyes:

Feh... It'd take a bit more than that to "hurt my feelings". ;)

[nukem]

This isn't all that impossible. HP has a technology called RGS which was made to stream graphics over a LAN or WAN to a thin client or other low powered machine. This includes 3D and it currently working on both Linux on Windows. I've seen it run a few times and it works very well, you can't tell the difference. Of course this was on a LAN but still this idea isn't that far fetched.

Of course this won't be the death of windows gaming or Linux gaming. Most likely they will have a client which runs on Linux Windows and Mac while their servers all run Windows.

[Aradreth]

I'm keeping an open mind about it, if it does work a linux port would be beneficial although I'd much rather play the games on my PC and not stream them. That said I wouldn't be surprised if this disappears quietly like Valves PowerPlay (http://en.wikipedia.org/wiki/PowerPlay_%28technology%29) did being unable to over come latency issues.

[Aradreth]

Of course this won't be the death of windows gaming or Linux gaming. Most likely they will have a client which runs on Linux Windows and Mac while their servers all run Windows.
You can put money on them pushing for linux games if it works to cut costs of running the service.

[Dragonlord]

Uhm... LAN and WAN are two entirely different things. On a LAN you are mostly connected to each other over one switch if everything goes well and using fiber. But on the internet you have tons of switches, routers and what the hell else in between. It's like saying you can drive a sports car on a shoddy overland road. Sure you can but it's a bumpy ride and slow unless you like picking your exhaust pipe from the latest larger stone on the road you hit.

[Jonsul]

Yeah I already know,
It's near impossible
All I'm saying is if it works it's be awesome.

Which is true, so I'll waste my $30 bucks and do the Beta,
If it sucks I'll let you know. If it works I'll let you know
too.

[DoDoENT]

I've just opened a www.onlive.com and while watching their intro movie, it stopped cca 7 times, because of too slow video stream (and right now I'm sitting in LAN with other 5000 computers in student campus behind 20+ Mbps internet connection). So how would it look like to play NFS on 1024x768?

[BlackStar]

Guys it won't work, simple as that. Want to know why?

Open a terminal and type "ping phoronix.com".

The human brain can easily detect latencies in the order of 10ms. Compare that to the number above. See now?

Read Svartalf's answer for a more technical explanation.

[igor]

But when I ping any of the sites local to me the latency is in the 10-20 ms range... Some of you seem to think that onlive are suggesting that their service will work regardless of server' location which is not true; at the demo they said they tried playing from australia and couldn't because of lag.

Provided they've really made a really low-latency codec, and with much fanfare, they say they have, only problem left is the dynamic latency...

It's quite possible for them to strike deals with the ISP to insure against that, a bit of money goes a long way.

Just saying it's not theoretically impossible, personally I'd consider a delay of up to 100 ms playable.

[Jonsul]

From what I hear the delay at GDC was 1 ms. Impossible for a human brain to notice. If it's already doing that good, then with lag it gets to the 10-20ms rang I'd still be able to play comfortably. Even at if it got up 100ms, I think I'd still be fine, and I'm sure it'd take a lot of lag to get up that high from 1 ms.

I don't know they may just have pulled that number out of their ass though. But if it's true, it is amazing and would definitely be doable.

[Dragonlord]

Nope. Don't compare pears with apples. On a LAN you have at best one switch in between. A switch is fast, it just duplicates the data packet on the appropriate port. On a WAN though you have routers all over the place. Each router processes data packages which yields delays. Striking deals with ISPs doesn't help much since their routers are not the only players in this game. Latency builds up at various places so this is utopian. As mentioned above they even said themselves it didn't work around the globe and if they can't do this, which IS the heart of their idea, then it's just bull from A to Z.

Besides the resolution is crappy and with heavy compression comes heavy artifacts. If you compress lossy ( and they have to if they want to claim such a thing ) then you have visual declination of quality and if you compress even more heavy than the best video compressors out on the market which already yield not too blistering quality with small videos then their quality is going to be more shitty than shit.

Besides why pay for a broken system where you are cut off playing your games whenever the Internet is overloaded again or your ISP line ( which is shared by the way unless you pay BIG buckets ) is crammed full? Or you want to play on a laptop? Always looking for a huge backbone since WiFi obviously isn't useful for that kind of usage at all? God the problems and restrictions are so numerous how can any sane person even remotely consider using that trash?

[igor]

As mentioned above they even said themselves it didn't work around the globe and if they can't do this, which IS the heart of their idea
That's not their idea at all, the servers are supposed to be local, and therefore link quality more under your ISP's control and generally much better.

I'm getting consistent 20 ms pinging a local site, and that's over wlan (which undoubtedly adds some lag)

[Svartalf]

That's not their idea at all, the servers are supposed to be local, and therefore link quality more under your ISP's control and generally much better.

I'm getting consistent 20 ms pinging a local site, and that's over wlan (which undoubtedly adds some lag)

Heh... I'm more than familiar with "local" and how painful doing that will be for them.

For them to have the "locality" you're describing, they'll have to sit on every ISP or sit on the backbone. I know of a company that tried this sort of play with content delivery. They experienced the LARGEST flameout of the dot.com boom at the tail end thereof. 85 million US in less than 1 year's time. The name of the company: epicRealm. I did part of the software that resided in the 50 data centers worldwide. It really did work- burn rate doing it, though, was hideous. As high as 3 million US per month in data access and sever power operations. And that doesn't get into what this bunch will need to make it really, really work.

When I say it's unlikely, there's a reason. It's because I've been in that space repeatedly and it's not cheap and typically, unless you've got really, really deep pockets (IBM deep or GE deep...) you can't afford to play in the first place.

[igor]

They've made standard rack mounted systems if that helps anything :)

It's really not as centralised a platform as you dudes perceive it, you should check out the videos where they speak about it, they claim both hardware and software advances in order to achieve this kind of latency.

I believe the public beta starts soon, so we'll just have to wait and see

[Svartalf]

they claim both hardware and software advances in order to achieve this kind of latency.

Claims are one thing.

Delivery's another. Unless they've got the sort of money epicRealm had in hand, they're not going to last past the beta.

Everyone keen on this keeps missing the math.

It's a brutal reality and software and hardware "advances" do NOT offset the way the Connected Internet actually works.

1Gbit links will only be able to handle about 500 people tops.
Keeping it local means you're going to have to handle thousands. 10Gbit links will give you about 4000 or so (Most stuff can't COPE with 10Gbit stuff full-on, not to mention the packet overheads will trim about 1/4 of the peak "speed" on the links... We're still struggling on how to best do packet analysis at near wire speeds at Tektronix, my current day job, in the 10Gb realm....)- that's 4000 or so for a major metripolitan area's main ISP. So you could probably serve MAYBE 12k people within a town the size of Dallas, TX. Seriously.

Realistic? Not on your life. :D

[energyman]

this will go nowhere.
a) Latency
b) Latency
c)Bandwidth
d) cloud is just a buzz word.
e) remember the disaster called 'thin clients' in the 90s? Same stuff
f) remember 'smart terminals' in the 70s&80s? same stuff

this is made of failure to rip off investors.

[me262]

I don't have much to say here since you all already got my points.

It seems like quite an ambitious project, however the 1 thing that will perhaps be the unwinding point of this.
Networking
Think about where they demoed it! at GDC! Which means this was done in a hall with a very fast dedicated connection to the internet. Even if they had that, it's still the fact that it was basically an in-house demo of the product, which means most likely GIGABIT LAN speeds.
Years ago I tried streaming games over TightVNC using 8-bit color, Hextile encoding (by trial and error, this was the fastest I came up with), I got 4 boxes that were never in sync and updated every 1 or more seconds. Imagine that, at least 10 times slower.
The very logistics of it seem to boggle the mind. I would have loved to have been there (and I get an excused absence (even credit!) from school for going) to see how much stuff these demo'ers knew about the product they're selling. They better have had a lead tech there!

EDIT: Discard my foresight, it seems to betray me.
I foresee this as probably a separate dedicated server and a separate client on the same network (slingbox anyone?). Server does all the legwork, streaming it to the client, that's the only way that they could combat all the lag the internet produces. At that point, you might as well save a few hundred bucks and build your own box.

I do see the need for them to get industry support, after all, what good is a product that has no titles at launch?
I for one will hold back on developing for this until it's proven and reliable.

... I guess I had more to say than I thought.

EDIT: Okay, I didn't really look at the site all the way through before writing, now I have. I just don't know how this is really going to work over the internet, games are all about instant command and reaction. Ever try to shoot someone in Quake3 on a laggy network?

EDIT 2: God, can you think of the bandwidth that ISP's would have to push if this service became popular? I shudder to think of at least 20 people doing this in an area simultaneously, or even 2+ people simultaneously on the same connection.

[djack]

This is a huge deal and could mean that Linux gamers finally step on Microsoft's face and finally get the gaming fix they have waited so long for.

By that logic, I have been playing Burnout on Linux for ages - my PS3 is plugged into a TV card- I play on Linux using tvtime!

[tball]

It would work great in Denmark :-)
Constantly < 14 ms lag from every single place in this little country.

~$ ping x.x.x.x
PING x.x.x.x (x.x.x.x): 56 data bytes
64 bytes from x.x.x.x: icmp_seq=0 ttl=250 time=7.220 ms
64 bytes from x.x.x.x: icmp_seq=1 ttl=250 time=7.403 ms
64 bytes from x.x.x.x: icmp_seq=2 ttl=250 time=6.895 ms
64 bytes from x.x.x.x: icmp_seq=3 ttl=250 time=7.790 ms

[Zhick]

By that logic, I have been playing Burnout on Linux for ages - my PS3 is plugged into a TV card- I play on Linux using tvtime!
Haha, great example/analogy. :)

[Dragonlord]

It would work great in Denmark :-)
Constantly < 14 ms lag from every single place in this little country.

~$ ping x.x.x.x
PING x.x.x.x (x.x.x.x): 56 data bytes
64 bytes from x.x.x.x: icmp_seq=0 ttl=250 time=7.220 ms
64 bytes from x.x.x.x: icmp_seq=1 ttl=250 time=7.403 ms
64 bytes from x.x.x.x: icmp_seq=2 ttl=250 time=6.895 ms
64 bytes from x.x.x.x: icmp_seq=3 ttl=250 time=7.790 ms

No, you won't. This is time measured for a couple of bytes. We talk here about entire fullscreens 30-60 times a second.

[daveerickson]

I must me missing something here. I have played many games over the net, and I know that lag sucks, but is generally tollerable.

I frequently fire up Diablo 2 these days and I float around an 80-150 ping and the game works fine. How is this different from the lag onlive would experience?

I also remember back in the day with Quake 2 and pings over 200 and still being playable.

The point is I can see needing more bandwitch since you are essentially streaming video from their servers to your screen. And if you have lower bandwith you would probably end up with a crappy looking stream, just like streaming from Netflix to my Roku box. It just seems to me that the games are already handling the fact that the games are being played by many humans in different areas and compensating for the lag. Why can't this service?

[MartjeB]

Like someone said.. closed minded soules here..

Do you really think this service will be ready in 2010? Do you really think the whole Internet will be upgraded in a year? No of course not! This will take years. Here in the Netherlands it will be (in a few years) perfectly possible. In a short while we'll all have fiber.

And who cares what OS the servers run on? It's all about the clients.

[Dragonlord]

I must me missing something here. I have played many games over the net, and I know that lag sucks, but is generally tollerable.

I frequently fire up Diablo 2 these days and I float around an 80-150 ping and the game works fine. How is this different from the lag onlive would experience?

I also remember back in the day with Quake 2 and pings over 200 and still being playable.

The point is I can see needing more bandwitch since you are essentially streaming video from their servers to your screen. And if you have lower bandwith you would probably end up with a crappy looking stream, just like streaming from Netflix to my Roku box. It just seems to me that the games are already handling the fact that the games are being played by many humans in different areas and compensating for the lag. Why can't this service?

Game state updates are a a small package of data every now and then. Here though it's huge amount of pixels to send over the wire. It's like comparing squeezing a bug or a Brontosaur through a wire. Obviously it takes a lot longer to get the Bronto through if the line isn't much fatter than a bug is.

And streaming of some video at crappy low resolution can not be compared to a game screen. In a video it is acceptable if the resolution is low and compression artifacts are visible. In a game though you want 1680x1050 at pixel precision so any trace of compression artifacts is not tolerable. Furthermore as Svartalf mentioned above ( and I can confirm this doing this in my company too to do support over the net ) try doing 1680x1050 through VNC over the internet and then we talk again about how feasible streaming such screens is. Hell even over LAN ( if I sit in the building not at home for example ) this is damn slow and totally unusable to play a game.

[Svartalf]

Like someone said.. closed minded soules here..

No, you've got several professionals telling you WHY it won't work the way you guys think it will.

I've been dealing for the last decade plus with the sort of stuff you and everyone else that thinks this is a nice idea keep glossing over.

It's part of my current day-job even.


Do you really think this service will be ready in 2010? Do you really think the whole Internet will be upgraded in a year? No of course not! This will take years. Here in the Netherlands it will be (in a few years) perfectly possible. In a short while we'll all have fiber.


Heh... Riiight. Let me run the numbers again so it'll soak in.

1.5 Mbits/s per SD level link.
5 Mbits/s per HD level link.

An OC12 consists of ~622MBits/s worth of data rate. THAT is still going to cost a pretty penny. Your fiber you remark on is to your house which translates into ~5-20Mbits usually, with a few notable exceptions around the world at 100Mbits. Business service, though, is framed in in terms of OC3, OC12, and so forth to OC192.

An OC12's worth of link will set you back ~$3-5k/month.

You can do the math on the OC192- and finding someone that'll feed you just shy 10G/sec worth of data directly to the Internet will be...interesting.

Now...

At 1.5 Mbits/s you're going to be able to service only about 414 clients simultaneously with a smidge of headroom to spare on that OC12.

At 5 Mbits/s you're only going to be able to deal with at most 123 with that same headroom, probably a little less- probably more like 80-ish if you're lucky.

Now, that presumes you don't have anything WRONG with your pipe at all on either end. If you knew anything about the way IP networking works, you'd realize that it's not going to work nicely and you're going to have a hell of a lot less people able to use the OC12's worth of link.

Why do I make this remark? If you use TCP, the congestion algorithm that is applied to each and every system, from the desktop machines, to the servers serving the data, to the routers and switches delivering it to you and the server, will cause random delays in packets once any one of them senses congestion. This injects latency into your game play- BAD latency. If you use UDP, the congestion algorithm will likely drop some or ALL of your traffic because that's how it does things on UDP. You go into low levels of congestion mode at 30-40% utilization of the link. It gets worse as you jam more crap through the pipes. SCTP ameliorates some of this, but not all routers grok SCTP traffic (there's you a hint...) and it still doesn't remove the issue completely.

Adding more bandwidth makes it worse.

No, no closed minds here. Just hard brutal reality- and one that won't change magically like you seem to think it will in just a couple of years' time. No, it doesn't matter what server hardware it runs on as far as the users are concerned. But with the Internet the way it is, it's just never going to work in your or my lifetime in a manner that will be commercially viable (here's a hint, you would have a 5-10 MILLION US per month burn rate to be close to making it commercially viable with the infrastructure costs where they are right now and will be for at least 5-10 years from now- and that's to service a quarter of a million people...)

This.
Is.
DRM.
Snake.
Oil.

The reason why the players bought in is most of the publishers don't listen to the people who do know better and they hear "all server side and nothing on the client of their assets" and they'll shower money on anyone claiming to make it happen.

[BlackStar]

A simple calculation shows that an uncompressed 1360x720@60Hz consumes about 1793Mbps of bandwidth. Assuming their HD stream is close to this resolution (if it's lower, it won't be HD, duh), they'd have to compress their stream about 350 times in order to fit their 5Mbps target.

a) There is no current technology that can compress a video stream 350 times in real time.

b) Even if they have created such a compression scheme, the quality will be atrocious. Ever seen youtube fullscreen? Something like that.

Also, keep in mind that every time you hit a key or move the mouse, you'll have to wait a full roundtrip before you see the results. Ever used a computer where the pointer reacts 60ms after you move the mouse? I have (VNC) - it's unusable.

You still think this may work? I have some money in a Nigerian bank that need a recipient - help me and half is yours. Interested?

(Yes, I'm implying this whole thing is a scam.)

[Aradreth]

This discussion is just going to go in circles until the company disappears or becomes a success. I'd put my money on the former but there is no harm in approaching them for a Linux port.

The reason why the players bought in is most of the publishers don't listen to the people who do know better and they hear "all server side and nothing on the client of their assets" and they'll shower money on anyone claiming to make it happen.
This is why I hate management (and politicians...), they make decisions they generally know bugger all about. When iD stands behind it I'll be sure it'll work, Carmack wouldn't support them otherwise.

[Dragonlord]

I don't want to step on John's toes buuuuut... Quake based games never had that much of good netcode as other applications. I would therefore not bet my money concerning network stuff on him ;)

[Aradreth]

I went with iD because out of the large game developers its the one when a programmer has the most pull (to my knowledge at least but I'm sure someone will enlighten me ;))

[deanjo]

f) remember 'smart terminals' in the 70s&80s? same stuff



70's and 80's? Hate to break it to ya but that is still alive and flourishing, they call them thin clients nowdays. Hell the vm market lives off of it.

Back to latency, etc etc etc. Yes I'm doubtful too but you can be guaranteed that the net providers are going to give packet privatization to the service if they can pull in another $10-20 a month off the casual gamer in a day and age where people are trying to cut their cable bills down during the economic meltdown. Any way they can soak extra money out of ya they will give it a shot.

[deanjo]

I went with iD because out of the large game developers its the one when a programmer has the most pull (to my knowledge at least but I'm sure someone will enlighten me ;))

Not sure how much that applies now days. Carmack does not dictate the future of gaming as he once did. If he had the same amount of clout as he once did pretty much all games would be using openGL which was his "holy crusade" for years. That gave way to DX.

[Dragonlord]

70's and 80's? Hate to break it to ya but that is still alive and flourishing, they call them thin clients nowdays. Hell the vm market lives off of it.

Back to latency, etc etc etc. Yes I'm doubtful too but you can be guaranteed that the net providers are going to give packet privatization to the service if they can pull in another $10-20 a month off the casual gamer in a day and age where people are trying to cut their cable bills down during the economic meltdown. Any way they can soak extra money out of ya they will give it a shot.
That's also a bit something else. A thin client ( or in layman terms an internet browser with or without a plugin displaying a website ) does still all the rendering but the processing is on the server. This is different from this problematic here since there the server does the calculations and send parameters to the client which then does all the rendering work. So on a thin client you have separation of number crunching and rendering which works well over the internet and has been done successfully since ages. In this example though number crunching and rendering is on the server and the entire result is send to the client. Can't be compared in my opinion.

[Svartalf]

Not sure how much that applies now days. Carmack does not dictate the future of gaming as he once did. If he had the same amount of clout as he once did pretty much all games would be using openGL which was his "holy crusade" for years. That gave way to DX.

That has less to do with his "holy crusade" and more to do with the ARB sitting with their thumbs up their backside (and as you observed with Khronos, they did it again with the mainline GL standard to some degree...) for too many years.

It IS worth observing that DX doesn't do so hot in the embedded space and if you want handheld stuff, you're going to HAVE to use OpenGL ES. It's so close to OpenGL (1.3 for ES 1.1 and 2.0, sans the fixed function path in the case of ES 2.0...) that more often than not, you don't have to do much to get an OpenGL renderer to work on the embedded system.

As for this... Heh... I don't envision them being able to pull the server side together enough to be credible within the next 5 years because of expense reasons. Tech-wise, you can make it work. Tech-wise, you might even have a solution to the latency concerns that the nay-sayers point out. Tech-wise, though, you're not going to remove the expense of handling multiple OC-192's worth of data per server location to be credible.

[Arceliar]

If by some miracle, this turned out to be more than vaporware, then, lag issues aside, I ask myself this:

"How could it possibly be worse than Flash?"

[Dragonlord]

Flash is not streamed. You can have full quality flash animations as it is vector graphics. Unless you mean streamed flash movies. They are totally crap.

[energyman]

and 'Streamed' flash videos are saved on your harddisk anyway (and they don't have to be crap).

[deanjo]

(and they don't have to be crap).

Ya, you can get very watchable streams nowdays. When I'm out at my farm I usually capture the football games from the stream later on. Granted they are no 1080p but very watchable and better then a VCR recording. Lately I've been using vReveal to "fix them up" afterwards. A great app that really needs to be ported to linux.

[Duo Maxwell]

What I'd like to know is how this is even feasable? I mean think about the amount of bandwidth this would have to be sucking down toplay a game like Crysis or any other fast paced, super high detail game at a high resolution and 30+ framerate without lagging. We're talking Gigabit fiber optical connection minimum for anything I can think of.

I don't know about any of you but I play my games at 1920x1200 with a minimum frame rate of 60 fps, but usually vsynced to the monitor's 85Hz refreshrate.

I'm sorry, I just don't see how this could possibly work on a large scale unless they've found the holy grail of compression algorithms, allowing them to compress gigabytes into kilobytes.

[Aradreth]

What I'd like to know is how this is even feasable? I mean think about the amount of bandwidth this would have to be sucking down toplay a game like Crysis or any other fast paced, super high detail game at a high resolution and 30+ framerate without lagging. We're talking Gigabit fiber optical connection minimum for anything I can think of.

I don't know about any of you but I play my games at 1920x1200 with a minimum frame rate of 60 fps, but usually vsynced to the monitor's 85Hz refreshrate.

I'm sorry, I just don't see how this could possibly work on a large scale unless they've found the holy grail of compression algorithms, allowing them to compress gigabytes into kilobytes.
The amount of data (compression + slight quality reduction to get to the required level) needed to be sent isn't really the problem it's the lag that'll most likely kill it.

[Dragonlord]

Furthermore you can not compress below entropy. Therefore no matter what super algorithm one might find in the future there is a hard limit below which you can not go.

[Milyardo]

I think everyone has missed the bigger licensing issue here with Onlive. If you buy a game with Onlive, do you own that game? If you cancel their service what happens to the games you've bought? Games you already own? Can you install a game you bought off of Onlive at home on your own machine?

World of Warcraft is one game where I think this system could work. The latency issues aren't a big deal, and you don't have the licensing issues I've mentioned above. If reasonably priced I wouldn't mind being able to play WoW or any other MMO from anywhere I'd please.(I'm the gold farmers would love this service, it'd make it difficult for blizzard to ban them by IP)

How about a quick game of Starcraft or AoE2 while you're on a buiness trip or in a cyber cafe(Cafe owners may not like that 1.5Mbs video stream however XD)?

If anything Onlive's marketing strategy is why they're doomed to failure, not the technical limitations of Video on Demand.

[Aradreth]

A little update if anyone gives a damn it's now under public beta (http://kotaku.com/5351295/onlive-public-beta-its-on) Intel Mac's and Win only, apply 'ere (http://www.onlive.com/beta_program.html) (USA only)

[gforum]

so you ppl arent satisfied with DRM? we need to go a step further with a system where its not even installed in your pc? crazy talk.

This goes in the opposite direction of current 'development' i say, while the new world tomorrow brings more of linux, companies like gog, woflire,Frictional Games, and of course S2Games(independent, Open Minded), games like in the old days, where u just push and run(and have fun -hassle free-).
These guys are going back into extreme proprietary levels...

funny, but i dont think so. maybe... out of touch?

the funnest game-time i every had was with friends, over LAN(that's L for local). I'll certainly be one to never give up on that, I'm not ready to finance our demise.


cheers

[pvtcupcakes]

Gaikai looks much better, and AFAIK it supports Linux.
It was started by David Perry, the guy who did Earthworm Jim.

http://www.dperry.com/archives/news/dp_blog/gaikai_-_video/

[Svartalf]

How about a quick game of Starcraft or AoE2 while you're on a buiness trip or in a cyber cafe(Cafe owners may not like that 1.5Mbs video stream however XD)?

If anything Onlive's marketing strategy is why they're doomed to failure, not the technical limitations of Video on Demand.

Oh, no, we're not missing the licensing issues... There's just no way they can supply enough resources now or in the next 5 or so years minimum to handle more than a small percentage of the entire potential client base.

The technical issues kill it dead right out of the gate.

I've already spelled those out- you're not going to make a lot of money if you're talking about only a couple of thousand per major metropolitan area- and that's all they're going to manage.

It's snake oil. Largely any of the companies selling this are selling that. It's another step in the folly of the "media" industries trying to "protect" their revenue stream.