Sean Hollister on the corporate demise of OnLive

The Verge ran an excellent story about the financial collapse of OnLive, which operated high-end gaming computers in the cloud and streamed the gameplay to distant customers.

Two Fridays ago, Steve Perlman told the 200 employees of cloud gaming company OnLive that it was all his fault. He thanked them for their hard work, and then he had HR show them the door with no severance pay. Then, through a legal insolvency tool, Perlman transferred all of OnLive’s assets to a brand new company and took over as CEO, hiring back only a skeleton crew to keep the ship afloat.

It's amazing to see such an in-depth and well-reported piece so soon after the event: this all went down days ago.


  1. OnLive was doomed from the start.  Latency matters in gaming, so much so that HDTV manufacturers even include bypass modes for their fancy processing just so people can play games on them, saving only a few milliseconds.  On the internet, latency is measured in tens or hundreds of milliseconds, each way.  It gets worse when you look at the whole picture:

    1. Player presses a button on the controller
    2. Signal is relayed to the Onlive STB, this is fast, nanoseconds, no worries so far.
    3. Onlive STB has to send the button press packet to the server, tens of milliseconds have passed. 
    4. Onlive server has to input that button press into the game state.  Almost instant.
    5. Onlive server has to render the next frame, assuming a 60 FPS for the game, this takes up to 16 ms or so. 
    6. Onlive server has to compress the frame (meaning it must finish rendering the frame first) taking more time.  How much I don’t know, but compression isn’t free, and at the very least it has to wait for the whole frame before it can finish compressing.
    7. Compressed video buffer is streamed back to the STB, tens of milliseconds pass again.
    8. Video buffer has to be decompressed, and then played on the screen.  That’s another 34ms in the worst case. 

    All told, 100ms delay between your button press and the action on the screen is about as good as your average person can hope for.  100ms is a lot in twitch game terms.  That’s 1/10 of a second, well above the threshold were people will notice the delay.  People complain about the ~34ms delay added by HDTV processing already. 

    This means shooters are out.  Fighting games are definitely out.  Rhythm Games are mostly out, although people can adjust for the latency sometimes.  Platformers are out.  Even RPGs where latency isn’t so bad often have minigames that are very timing dependent.

    So you have to target casual gamers.  Fine, people make tons of money on casual gamers all the time.  The problem:  Reaching the casual gamer is hard.  They don’t read Kotaku.  They have never heard of nor care about the GDC.  It takes real advertising on mass media to reach them, and how much did you see?  None.  Because that kind of advertising is expensive. 

    1. “Latency matters in gaming…”

      Not really no. It matters in SOME gaming, like FPS yes. If I am playing chess online with someone, no so much. It all depends on the types of games hosted.

      Also it depended on how “cloud” is the actual cloud. There are plenty of games, particulary older ones that have been converted to pretty much browser games, but still multiplayer.

      Anyway you are always going to be limited after a fashion by trying to do everything in the cloud, which will limit what you can do with it. However for a certain limited selection it is entirely doable. If that is really a profitable business plan is something else entirely.

      1.  If you’re playing chess online with someone you can do that on a 10 year old laptop without their cloud service. OnLive’s market was “people who want to play cutting edge modern games but don’t want to spend $300 on an Xbox or PS3”. That market is non-existent.

    2. That was the real problem, not the parts the article goes on about.  onlive had a great interface, was easy to use, but the lag just stunk.  I don’t see it being as unsolvable as you put it though.  Listening to carmack in the 2012 quakecon talking about it, he appeared to say that it was solvable, but they just hadn’t put the work into fixing it.

      1. Solving it might be just focusing heavily on games that don’t require fast reflexes.

        It might be possible with some games to move some of the game logic to the thin client somehow to avoid anywhere from 1/3 to 1/2 of the lag (basically the packet transfer time between your onlive box and the servers), but I wouldn’t expect that to work outside of little minigame type things, and it would be a ton of work on Onlive’s part and would have to be replicated for every game that needs it.  Probably not viable long term.

        Ultimately, I feel like this service was aimed at a customer that doesn’t really exist:  The person who wants AAA titles and can afford a fancy cloud service, but doesn’t just want to buy a $200 console. 

        The real kicker is that outside of a few edge cases, games that aren’t lag sensitive are also the kind of games that run just fine on netbooks and other limited computing devices.  Civ 5 is an exception, but if you want to play online chess you don’t need a fancy cloud server streaming you graphics.  Literally any computer will do. 

        Onlive failed because it was a bad idea, not just because the CEO was apparently a bit batty. 

  2. Just because you can put it in the cloud doesn’t mean that you should put it in the cloud.

    Also, the article says: “Streaming games make sense: There’s no ability to pirate titles, and manufacturers don’t have to build up an install base by selling expensive hardware to their customers at a loss”. Those are advantages for producers, not for consumers. I’m not exactly sure what the advantage to the consumer is (I doubt that they’d get their gaming fix any cheaper in the long run), and there’s a very obvious disadvantage, in that your gaming experience is dependent on the performance of your network connection. If that performance is less than flawless (as it will be), your experience will suck.

    If your consumer-facing product offers only advantages to producers and none to consumers, you’re going to have a bad time. As you should.

    1. Well, the advantage for customers is obvious : not having to pay for a top-of-the-line PC. Just get a good screen and you’re done.That said, jandrese’s comment @1 is spot-on ; unless they manage to get perfect and lossless data transfer in less than, say, 30ms, this won’t work for hardcore gamers. I’ve tested the service a few months ago : horrendous video quality and small-but-noticeable lag. I guess it could be a niche (work as a demo, or for people who can’t afford gaming computers), but I don’t see it taking over the market. Neither did customers, apparently.

      1. “We provide the same thing you can get elsewhere, only for cheaper and lower quality!”
        Hey, it’s a business model.  And in some markets it’s well received.  But video game fans are generally looking for “more” and “better”, with less emphasis on “cheaper”.  The target audience for OnLive is people who would be very excited about:
        a) Being able to play the latest AAA game at max detail on a netbook
        b) Having instant access to a huge library of games

        I think that almost all those people are already willing to spend the money on their own gaming machine, and aren’t particularly attracted to a laggy, low-res version of the same experience.

        The portability and convenience of OnLive could appeal to a much broader casual gaming audience, but casual gamers don’t need OnLive – those games already run natively on smartphones and netbooks.

        1. For a gamer who spends a lot of time traveling for work I could see a potential market. “Don’t haul around your gigantic gaming laptop through airports, just carry an OnLive set-top box!” but the problem is that hotel internet is so atrociously spotty that you’d almost be guaranteed a worse than usual gaming experience.

  3. I’m amazed at how poorly marketed this service was; I remember hearing about the idea a while back, having the same thoughts as jandrese about the insurmountable latency problem, and then nothing. I assumed that this was one of those ideas that died before it was brought to market.

    And now the next thing I hear about it, four years later, is that it was available for three years and has just now gone bankrupt. I don’t follow the game industry obsessively, but I follow it enough to generally know who’s offering what.

    Were they hiding this thing under a rock? Apparently the first rule of OnLive club is you don’t talk about OnLive club.

  4. i used this actually and purchased one game with them for the offer of 1 pound. was fun, and defo playable. no lag if you had decent internet connection. bit angry there was no warning email!!! but hey i only paid a pound for one game that i played at least 20+hours. 

    i think their main mistake was making demos available for 20mins sometimes longer. i often played a lot of games demos 2-3 times after which i was bored of it anyway. so i didnt end up purchasing. they should have logged demo plays and only allowed one offs…

  5. OnLive has a critical flaw, the input latency. Online games all have some network latency but the difference is they have almost ZERO latency from the button press to action on the screen. With OnLive the network latency also affects control inputs and for me at least it was a complete and total deal breaker. That 50-100ms delay every time you wanted to do… anything was just horrible from a usability perspective.

Comments are closed.