OnLive Is Dead

I'm not sure how dead you can say something is when it was never really alive and is most likely just fraud, but let's give it a try.

This is the claim for OnLive's technology that is supposed to let you play AAA game titles without an expensive computer as long as you have a fast Internet connection:

Latency through the algorithm is just 1-ms instead of the 0.5- to 0.75-second lag inherent in conventional compression algorithms used in corporate video conferencing solutions, for example.

Now, let's ignore the overlap between people with expensive computers and people with very fast Internet connections. Obviously this business only has an audience if those two don't overlap too much, as ideally you want people with very fast Internet (5Mbps or so) but either don't have PCs capable of playing new games or simply don't want to invest in upgrading CPUs and GPUs as needed to keep up with new game releases.

So, for that to be an attractive proposal, the subscription fee would have to be less than what those upgrades would cost over a similar time frame. If you don't want to be cutting edge, but want to stay current, you could easily do so buying one new video card a year and paying about $200 for it. So that amount, plus the cost of however many games you'd be playing, has to be more than a year's subscription. So that proposition largely comes down to how many games will be available on the service, compared to how many games people plan on buying (or, rather, people's perceptions of how many games they are willing to buy, compared to how many they are prepared to give up-- presumably the service will offer a subset of all available games).

So the more games you'd be willing to buy in a year at $60 a pop-- three, four, or more-- it starts to look a bit better. Except you don't own those games; like subscription-based music and movie services, once you stop paying the fee, you lose everything and you've nothing to show for what you've spent so far except your memories.

Of course, if someone was willing to spend $300 a year on new games I'm not sure why he'd balk at the video card or be willing to shell out for something he can't keep. In the conversations I've had with people about similar services-- namely music subscription services-- the primary reason people justify spending cash monthly on something they can't keep once they stop paying is that they simply couldn't afford to consume the content they want if they had to purchase it outright. Someone might have an interest in a new console every 3-4 years or a gaming PC upgrade every 9-18 months, plus new titles, but if they don't have the cash then their desire is moot. So the audience that this service is targeting is the portion of the gaming market that does not have the cash to participate fully-- they can't afford the games they want to play and the hardware required to play them.

OnLive talks about two service tiers-- standard def on relatively slow connections (1.5Mbps) and high def on faster connections (5Mbps). The problem with the first option is that standard def games, or 640x480, are not particularly taxing anyway. In order for that service tier to be advantageous, it has to be significantly cheaper than high def content, and probably targeted at users without any PC at all, as even very cheap new or quite old equipment can play games at that resolution.

Let's get back to the above quote, the one that compares the latency in OnLive's compression technology compared to the latency in corporate video conferencing solutions.

First of all, the claim strikes me as bogus. Half to three quarters of a second in lag is 500ms - 750ms. That is not the delay inherent in the compression routines of any system in commercial use for voice-- not for VOIP, not for videoconferencing, not for anything. That might be the total lag in those systems, under less than ideal connections-- such as satellite-- but certainly not the lag in the compression alone. If it were, there are less than ideal connections that would be wholly unusuable for videoconferencing, but in reality, are entirely acceptable.

So the comparison is apples to oranges; they are apparently comparing the total latency in a videoconferencing solution (compression plus transport) with the latency ONLY in their video compression, without transport.

However, there's a bigger question here, which is: if this company, after seven years of development, happens to have come up with a video compression routine that is, as claimed, 500 times faster than currently used in the corporate videoconferencing market, why not just use it to take over that market? Why not use it for delivery of streaming movies and music?

There are several possible answers for this.

One might be that they have not actually developed any such technology. The demonstrations to date have shown compression artifacts even under ideal, controlled circumstances, meaning that there will very likely be a perceptible difference in rendering quality between a game hosted on their service and a game hosted locally on comparable hardware. Their system, as demonstrated, may actually require more bandwidth than their current claims, although that isn't really relevant to their value proposition, which is not trading cost for bandwidth but rather trading cost for the speed of compression. (Actually, looked at one way the system is trading latency and bandwidth for CPU/GPU cycles at the cost of latency, with the apparent latency cost very low, but the bandwidth cost very high.)

After all, 5Mbps is also more than adequate for watching high quality video from remote sources and this is easily done; any variability in transport performance can be eliminated by aggressive buffering. Buffering is of absolutely no use in a gaming situation as it just causes more perceived lag rather than eliminating it. Since unlike in video, you can't compress the next frame until it's rendered and you can't render it without user input, the system will be playing hurry up and wait. Presumably their system can compress that frame in only 1 millisecond, so the total lag in your system will only be whatever you normally get when playing a PC game, say, 50-150ms depending on your connection, plus that 1. The amount of bandwidth required will be much higher than to just play an online session of a AAA title game like Crysis, where only player actions and world state go over the wire. World state data stays on the server now, but a compressed audio and video stream, as well as player input, goes over the wire.

Another reason may be that while such a drastic decrease in the speed of compressing a given frame is advantageous to an interactive session like gaming (or videoconferencing) it is of little value for streaming video or music, as content can easily be compressed in advance, decompressed on the remote end, and buffered in between. Whatever licensing fees OnLive would want for their technology would most likely not result in any increased revenue for those services as it would not offer any appreciable benefit to the client or cost savings to the service operator/content provider.

Whether the tech exists or not, the gaming market may have been targeted on the assumption that it was easier to pitch the gaming industry, as a relatively broad based consumer service, as a more attractive target to potential investors. Marginally better videoconferencing does not warrant the kind of financing OnLive has sought (and received) from places like Maverick and Warner Brothers. However, a gaming service doing something no one has done before, that purports to be able to expand the gaming industry far beyond those hardcore dedicated who upgrade consoles every generation and PCs almost yearly, is attractive enough.

OnLive is either a solution in search of a problem or vapor in search of money; take your pick.