Dave Perry on OnLive

To get 60 video frames per second of gameplay streamed to your home (like OnLive needs), the Internet would have to deliver your control inputs to their server farm, then OnLive would have to process the game response, calculate the new display (based on those control inputs), render the frame, compress it, return the results through the Internet, and decompress it locally, all in under 16.67 milliseconds. Is that even possible? Well, if I ping (test the speed of a connection) just 50 miles to Los Angeles and back, with just 32 bytes of data (no gameplay, no video data, no compression), it takes between 7-15 milliseconds. So on my network (which is great by US standards), once the gameplay is included, I will end up dropping frames, unless the servers are REALLY close by, or on special high-bandwidth networks like OnLive used on their Game Developer’s Conference booth.

And he continues... (I love this kind of scribble-as-you-think stuff). Dave has been around for years and years, and made MDK, which was major and got everyone's knickers in a twist as it stands for Murder Death Kill. Heh.
 
OnLive.

Comments

[Adam]

I'd be more interested in Dave's opinions here if he didn't start by spouting inaccurate pseudo-science technical mumbo jumbo for the first half of his commentary.

From: Reporting a ping to somewhere random (ignoring what I and the other commenters who appear to have a clue have said about what you should be measuring here), to: saying he needed a "higher bandwidth" network when he meant "lower latency" ... this is not impressive.

To then repesatedly claim his argument is based on bandwdth, when in fact he's *completely ignored* the bandwidth issue (couldnt' be bothered? I dont know why he ignored it), again is unimpressive.

I suggest instead seeking out articles / blog posts from people who actually still program games or have online games experience, or even - best of all - are actually network programmers. Not a game designer.

So ... to my mind ... this smells of a silly personal PR stunt for his next company, rather than any serious attempt to engage on the subject.

[Mr Tom]

There is a good article by Richard Leadbetter on Eurogamer about Why OnLive can't possibly work

It just doesn't sound plausible at all!

[Space Toast]

I think there is a midrange future for systems like this, but we have to remember that they'll have to crawl before they can dance in crosswalks.

Think of the evolution of CD ROM games in the '90s -- where bandwidth was also the primary limiting factor. We saw a lot of non-interactive video interspersed with relatively low-bandwidth interactive parts. Burn:Cycle was revolutionary at the time, but its interactive portions were basically sprite-based minigames superimposed over static backgrounds and (slightly less successfully) mpeg video. See also: Obsidian and the Journeyman Project games.

The real question is, what can a distant server do in a bandwidth-limited situation that a home system can't? All I can come up with are scenes with a high degree of computation but relatively modest rendering requirements. Declare martial law: Click "End Turn": Watch Flash video of 3000 citizens storming the parliament as 500 riot cops try to stop them, watching closely for clues to the opposition's real leadership: Issue arrest warrents: Click "End Turn": Repeat.

A better place to start might be bringing NES/Genesis/Saturn-era games to any tv at Netflix prices, over a free-with-subscription ($15) controller/console that's basically nothing but a wifi card and video decoder. But wouldn't most people interested in that already have a computer powerful enough to run an emulator?

[bob_d]

What the Eurogamer article pointed out was that the real issues aren't so much technical, but logistical and economic. At some point we will (or could, rather) have really, really low-latency internet and real-time video compression of the sort being claimed. The real problem is that you need the equivalent of one fast pc (plus the video compression hardware) for every (concurrent) gamer, located quite near where they live, if you want to compete with Xbox/Playstation, which is OnLive's stated purpose.
Of course, if you go with slower, more "turn-based" games, and games that are a few generations old, it becomes economically feasible, but you also wouldn't need OnLive to do it- you can play turn-based games over the web now, and older generation games can easily be played directly on the sort of hardware that would be needed for an OnLive system (i.e. a really cheap set-top box or older computer).

[bob_d]

Looking around, I am surprised by all the enthusiasm the publishers are showing for this system, despite the obvious problems. Actually, I take that back, I'm not surprised- publishers really, really want this to be true. This is the holy grail for game publishers- metered games on a set platform. (The end of piracy! The end of second-hand games! No extra costs for developing for multiple game systems! The more someone plays your game, the more money you could make!)
Heck, I think we're going to see something like OnLive even if it *doesn't* work (logistically OR technologically).

[Space Toast]

Some good points, Bob. A better option for the short term might be high-speed LAN-based gaming. Say you have a hotel. At one point does it become less expensive to install a beefy, high-maintenance game server and lay fibre to each room than to buy an XBox 360 for each unit? Consoles are generally sold below market price for their hardware -- does it ever become cheaper?

What probably has the publishers squirming is that Sony has voiced their belief that this is the future of gaming. Which begs the question of whether Sony is the future of gaming.

Maybe we should just learn to take gaming advice from "Lands End" models with a proper grain of salt.