Onlive remote gaming isn't new
ONLIVE IS MAKING waves talking about streaming games, and some are falling over themselves wondering how the magic happens. It is not magic, has been done many times before, but this is by far the most accomplished version ever shown.
The concept is simple, we told you about it yesterday, and it doesn't do anything more complex than render 720p rez games remotely. This isn't exactly an easy thing to do, but AMD's Fusion Render Cloud did the same thing at CES, and there are several LAN based ones like the Zotac boxes shown at CeBIT which do it more locally.
Basically, Onlive took several existing concepts, redid some of the parts for better performance, and put out the most polished package to date. It is something even a console gamer could potentially use, it could really be that simple.
The problem with running games remotely is always latency, if you are driving and turn your car, if the car doesn't react for three or four seconds, you will eat wall quickly. The generally-assumed latency that a human can perceive is 85ms. Anything less, you probably won't notice, so that is the bar Onlive has to hurdle.
First up, they made a proprietary compression algorithm that they won't talk much about yet. It compresses things sub-frame, and adds under 1ms of latency. Things are good so far. The lag between controller and computer, basically how long it takes for a button press to register, adds about 5ms more. The killer, ISP latency, it is between 20-50ms if you are on a good link.
That said, the worst 'official' case is under 60ms, 75 per cent or so of what you can perceive. So far, so good. Then again, with the Net, there is no such thing as 'worst' case, your mileage will vary, especially if you are on a shared connection or a cable modem. We won't talk about Comcast's packet shenanigans either, curse their black eyes.
On the cloud side, Onlive runs the game, and has a bunch of hooks into the system. The company won't be very specific about how they do this, it is their secret sauce, but the game itself is unmodified code. There are currently two US data centres, east and west coast.
The stream itself is then sent over standard TCP/IP to the user. A high rez 720p stream will need a 5Mbps connection, it maxes out at 4Mbps data rate plus ISP marketing safety margin. A standard def stream will only need a 2Mbps link, 1.5Mbps stream plus ISP exaggeration factor.
On the receiving side, you can use either a browser plugin or a 'microconsole'. The plugin is currently for PC and Macs with Linux under consideration if the market warrants. The microconsole is a little box with a USB port for the controller and HDMI outs. It looks like a smooshed pack of cigarettes with a PS3 controller wired in.
Onlive microconsole
The silicon in the console is proprietary and there are no specs to be had on it. One nice feature is the compression algorithm is flexible and heavily server loaded. Compression is intensive, decompression far less so, and it can be done serially or in parallel. The light-duty parallel bit strongly hints that the console chip is a multi-core part, and is said to have custom hardware compression blocks.
There are actually two streams sent out at once, the one you play and a higher-rez version. The play stream is adaptive and varies compression and data rates on the fly, based on network transit conditions. The high-rez version is not playable or really interactive, it is meant for 'brag clips' or filming of the session. Think of it as what the spectators see on the big screen at a Lan party. It is multicast for efficiency whereas the play stream is unicast.
Pricing isn't set, nor is it going to be really done by Onlive. The company will OEM the boxes and run the cloud or let you run your own should you want it. You are much more likely to see an Onlive box with your cable company's branding than theirs, and the pricing will (shudder) also be done by the cable company.
In the end, Onlive has the first end-to-end complete package for streaming games that looks viable. Nothing is particularly new, nothing is particularly exciting, but it is all done seemingly right. Without playing with it under live Net conditions, you can't say how well it really works, but for now, it looks to be not only functional, but best of breed. Keep an eye on it. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
Something tells me my 8 meg BT connection, which struggles with BBC iplayer, would simply fail to run this smoothly. Well, it might manage it at 3 am in the morning.
Guess what happens when you take an old xbox controller and write live on the buttons? It rhymes with smawsuit.
But Onlive will die 3x before that happens. 2 words: Ageia, Ageia.
I remember that sega tried this back in the day. And failed.
I've seen lots of stuff explaining the ports on the microconsole but nothing about how it connects to a network. Is it wired or wireless? Hope it's WLAN and supports WPA.
It definately has a LAN port (shown in some other pictures online)...Don't see why they wouldn't have WLAN, but wouldn't that be shooting themselves in the foot by adding even more latency?
great news!! until i read the bit about needing a decent internet connection which rules out most of this country..and this doesnt sound like its gonna be cheap either ...i can also imagine your ISP targeting you for downloading over your allowance
maybe when virgin media rolls out there much touted 150mb fibre optic !! it will all be fine ..yeah rite!!
LIKELY: Download-based games, episode by episode plus add-ons, where the sum of parts costs more than current games. Expensive and fail-prone consoles to run it locally.
NOT LIKELY: Streaming games, charged by playing time, that run/clog on a central CPU and extremely dependent of consumers' own infra-structure quality. Relatively cheap console.
Yeah right, dream on.
Muds
Moos
Heck, X11 Netrek in ~ 1990! Yet another example of why UNIX is better.
First of all, this service will be limited in popularity until the U.S. get's the infrastructure (broadband) of a first world country where big internet pipes are provided at a reasonable cost.
Also, I find it idiotic that they're not introducing the system with immediate Linux support. One of your biggest user bases is obviously people who are sick and tired of being forced to use Windows because of their gaming needs/desires. Something is "fishy" about this IMO especially when all that is needed is seemingly a browser plugin. Can they not quickly compile one for the Linux browsers?
Lastly, I'm also concerned about the bandwidth requirements of this and how a large adoption of this service could potentially slow down the current internet infrastructure. The is especially true if ISP's get involved with packaging deals for profit and install some kind of QOS packaging priority for OnLive packets to make the system work as advertised or expected.
I think the service has a lot of potential but it seems to be a "cart before the horse" issue especially as ISP's are increasing capping limits on data rather than expanding and growing their infrastructures. Maybe the stimulus plan will put this behind us but until then, this company should initiate this service in "modern" countries that can handle the implementation of revolutionary technologies.
cant deliver on the promises. no video compression technology is that fast, lag is inevitable when you take a realistic compression time. 1 millisecond is utter nonsense. it just doesnt exist im sorry. compressing 60 frames per second HD video along with audio inside a 5 megabit stream also means macroblocking, even if you can compress it fast enough to be playable no compression algorithm exists to make such a video clean and sharp. if it did there would be far better (and more lucrative!) applications than this. then you have the hardware, so you are saying that each user that wants to play a game is somehow going to have a dedicated 'in the cloud' machine to serve them their games, with a modern graphics card for the 1280 x 720 res, and a capable multi core processor, and compression hardware? thats ridiculous. think of the space required for just a few hundred thousand users, think of the heat, the cost to purchase the hardware or even run such a vast warehouse of machines!!! its like providing everyone who visits a popular website with their very own server, each and every one. seriously - explain how that is possible?? it just isnt. what about latency, the killer. it all comes down to this at the end of the day, and it cant be done. all the algorthims and hardware you want their end, it comes down to the connection from your isp to the cloud. no one can guarantee a coinstant stable, ultra low latency connection all the time. common sense says so. this. cannot. work. end.
Besides all the obvious problems with this system ie internal networking, external networking, guaranteeing service for all users at any given second, etc.. and even the less obvious ones like in order to actually show a profit you're probably looking at atleast a $200/£138 a year service fee per person plus the cost of the things you "buy" but don't actually own and the possible rental fee they are gonna charge you on the little box needed for the service I don't think anyone really has payed attention to what the guys running the service have said. They point blank admit that even if you have the best connection in the world it won't be as good as playing the game on a box right in front of you because of how the video stream works. I'm not sure what streaming compression system they are using but I bet it will probably be some variation of the flash video codec which is the popular website format these days and it has horrible quality/compression ratios plus even at the best quality it's framerate tends to be crap in comparison to every other codec on the planet.
Oh there is also the fact they admit that not every game released is gonna appear on their system so odds are you're still gonna have to buy a gaming pc and consoles anyway. I mean it's pretty obvious that Sony, Nintendo and Microsoft exclusive games aren't gonna appear on this system unless they were already planned on releasing them for PC. God then there is Capcom games which are almost universally horrible on PC but great on consoles which even if they get released on this system you'd rather own the console and play them without requiring an internet connection. Anyway I could continue to go through the list of developers which won't be releasing all their games, if any, on this system but I think you get the point.
4Mb sustained? You're having a laugh. It's possible on some UK ADSL connections, but only when connected to companies with vast amounts of bandwidth.
I can't believe people will pay the money required for onlive to supply the necessary bandwidth. The games will be old and low res in order to meet the necessary targets.
It might be interesting technology on a high speed LAN, but not over a WAN. Perhaps they're looking to be bought out..
does charlie even know that nvidia was
a development partner for onlive and
that they are currently using nvidia
GPUs(8800 GTX)? looks like Jen-Sen
managed to swoop some ass full of
green GPUs under charlies bonet.
i think we will see a new fight
between nvidia(onlive) and DAAMIT
(AMD fusion cloud) and guess on who's
side charlie will be.
vulcanproject that is kind of the point, prior to them creating this no video compression technology existed that was that fast. Now it does exist (they invented it). With regards to providing the computing hardware to run games in a cloud, it is possible and they are doing it (and they claim to have designed custom server boards specifically tailored to this application which minimize cost). I agree with you regarding the final weakness being the ISP connection, but they claim that their system is fairly robust (I interpret that to mean that it probably doesn't work on a crappy connection, but it probably does work on an average one). Also, Peter Kay, the speed requirements (1.5Mbps or 5Mbps) are peak reauirements, not sustained. Their biggest technical challenge will be multi-player homes (1.5Mbps peak is fairly average, but for multi-player in same home you would need 2 or 3 times that peak which is not a mainstream connection) and their biggest marketing challenge will be pricing (it has to be less than the cost of buying the games, otherwise why bother since there will also probably be subscription fees) and availability of titles.
problem is tavi if the compression worked and they invented it, they are wasting their time with this service because of the other insurmountable problems. even if you take the compression in isolation, that sort of claimed quality heralds a new era in video compression technology for the world. it would be incredibly more lucrative outside of games, dont you understand? its like inventing an engine that runs on water then only using it to power lawnmowers, not licensing it for anything else....as for these custom boards, you do still realise each custom board would have to have massive processing power, a decent gaming pc these days is a half teraflop+ machine. support, room, runnning costs, networking for a just a few hundred thousand half teraflop boards? you kidding!!?!? thats possible you think? its a dream. its simply impossible at this moment in time. but most of all the limiting factor is latency, and most of the connections in europe and the united states couldnt hope to have low enough latency so that its as smooth as that presentation. lets be honest here, this is aimed squarely at the western world. that presentation was a joke im sorry - they obviously had total control over so many variables, and their servers were just 50 miles away, no doubt on an uber dedicated optical connection to the show hall. seriously you think thats going to be representative of most peoples connections to servers maybe 250 miles away, or 2500 miles away? hahahaha. thats the killer right there, and thats why the whole thing cant work for most people
You are overestimating the worth of the technology they have created. Their one-direction specialized server video compression is certainly fast but it probably does not have the greatest compression ratio out there. The compression is specifically tailored to this domain (extreme one-direction speed with dedicated compression hardware at the server side) and not necessarily to two-way or bandwidth limited or space limited or server CPU limited requirements so I doubt its application is as wide as you say, because there are cheaper 'good enough' alternatives, although it probably is potentially wider than just video games. Each custom board does have massive processing power but I agree that it is hard to believe they can price this competitively and maintain the level of computing power necessary as their customer numbers grow. Yeah your comment about the network is the biggest one from the technical perspective because it is safe to assume that GDC had a monster internet connection and 50 mile distance from their server proves nothing. They need to do a demo from a residential level connection at greater than 200 miles to prove the network piece.
so now you say it doesnt have the best compression ratio? yes, which i covered earlier. that means macroblocking, pretty average video/audio quality on the claimed 5 megabit connection at best. even if the compression is as fast as they say to reduce latency (very unlikely), its almost certainly low quality not possible of coming anywhere close to locally rendered HD 1280 x 720 60 frames per second. you cant have it both ways. this compression technique would have to be out of this world to deliver their claims - ultra fast AND super quality. more likely its not fast enough and decent quality where you get awful lag, or fast and pretty poor image quality. chances are its both, not fast enough and average image quality. there has been a lot of fanciful ideas and claims from companies over the years, and carefully controlled demos of all sorts of products. i seen them all. but this takes the biscuit. it isnt going to happen near the level they claim
This isn't off the shelf hardware or software so I doubt it is the worst case that you mentioned 'slow and average', I bet that it is probably fast and average at the bandwidth they claim, but maybe average is good enough if the responsiveness is there. I haven't seen that many demos, so perhaps I have been duped but according to articles I have read here and elsewhere the professional reviewers have been duped as well, maybe because we all want this to be true? I am still not ready to say it is impossible, but given your arguments it does seem unlikely (but I still hope you are wrong).
Scientists are great at making up new ideas; marketers are good at selling them. Together they still manage to fail epically.
Remember that failure is when a product is completely retarded and millions of dollars go towards spin. "Cloud Computing" is so much BS and who the hell wants to watch a movie of them playing laggy games?
Welcome to "Phantom 2". 70 Million Dollars later their CEO will have a good laugh from his Swiss Chalet.