Cathode Tan - Games, Media and Geek Stuff
logo design by man bytes blog

Thursday, March 26, 2009


As of yesterday morning, I had never heard of OnLive. As of now, I've had like three conversations on it and read umpteen blog posts, several comics, etc.

Buzz like that. It's slightly intoxicating, pretty much no matter what.

So, let's preface this with the obvious: I'm not at GDC. I haven't seen this demo. I'm not in touch with OnLive. I have zero technical background information on the product. All I have is hearsay, rumors and third party reports.

It's pretty obvious as to why the world wants to believe in OnLive, but I have to say I'm not really in the excited bunch.

Course, we've been here before - the promise of a console-busting console, all downloadable data, hassle free play, so on and etc. It turned into one of the great vaporware - the wildly aptly name Phantom.

It's entirely possible that OnLive is a similar sort of shenanigans. I'm not saying they're necessarily liars and such, just that it's one thing to produce a decent looking demo based on sound theories and a completely different thing to make a market ready product. So we could easily see a year from now people still poking the burning embers of this red hot hype, but have nothing to actually buy.

Unless OnLive can prove some real world testing, I have to say this is likely. The promise of "zero lag" is something which is really, really hard to deliver because the Internet is, by design, not controlled by anything which can keep that promise.

The premise of OnLive is pretty straightforward. Virtual Network Computing is something like a decade old now and I can run it off my iPhone if needed.

The problem isn't the mechanics, it's downloading frames per second. Let's remember:

So what's not to like? For one thing, the service relies ...

... on a broadband connection with at least 1.5 megabits per second for standard-definition images, or 5 megabits per second for high definition. OnLive estimates that about 71% of U.S. homes have enough bandwidth for standard definition, and 26% have enough for high definition. But effective speeds are often less than advertised.
-- LA Times, GDC: OnLive promises near instant online access to high-end console and PC games

I just did a speed test here at it never hit 2mb/s. So, I'm certainly not playing any games in high def even by OnLive's theoretical standards.

Is there any gamer out there who is going to trade the convenience of cloud-based gaming with significantly lower graphics? I highly doubt it.

But let's play even deeper into the hype and theorize that somehow OnLive is capable of delivering liquid gold into the intertubes and high def video delivery is no longer a problem (which, as an aside, if were true they could be making a killing without bothering with games - just video aside). Would I still want it?


Look, I'm all for digital delivery, but at the end of the day I like the idea of having software installed that I can play whenever I want. I had enough troubles with Steam, and that's nowhere near this side of spectrum. OnLive would only interest me as a rental, pay as you play, kind of model - possibly a monthly subscription if could deliver the goods. But don't think for a moment that I'm paying anywhere near retail prices for software that someone keeps on the sly and allows access by online means. It took one bad experience from Valve, and I'll never go down that rabbit hole again.

If we're talking convenience, here's the convenience I get through my current means - I pay retail price, I can get access to the title on release day, if anything goes wrong with it, I can return it for a refund or at worst, go to a Best Buy and exchange for store credit.

Cloud based gaming, should it ever really take off, will still need to actually solve a problem I have, be it performance or cost or something. Otherwise it's just a clever parlor trick.

Right now the only problem I could see it solving for me is Windows based gaming on my Mac. If it offered a hassle free solution for that, we might have a deal.


Thomas said...

Eurogamer's got an column up about it today. They run much the same numbers, and verified them with a streaming expert. Plus, they talk about the problems of scale--you basically need one PC per user for anything 3D-accelerated, since it can't be virtualized (if it could, you'd just be gaming with VMWare on your Mac).

Then imagine what happens when the game crashes. I'll defend PC gaming to the death, but even I will admit that's gonna happen, particularly with stuff like GTA IV that's really been badly ported.

Obviously, it won't work. I can't imagine what they were thinking.

sterno said...

The trick here is about time sharing of equipment. It's just like what you had back in the modem days when you'd have 100 modems and serve 500 customers. The ratios would have to be much lower because getting a busy signal when trying to play your game would be a huge problem.

It's possible that they've modeled this out, but I should think the big problem is that gaming has very strong peak demand problems. People generally aren't playing games at 10am, but a lot are playing at 7pm.

Having said that, if they've got a cloud of raw processing power, they can maybe sell time on their servers to keep those idle cycles in use.

It seems like the basic physics of it is plausible, I'm just wondering how feasible the economics of it are. Licensing costs, bandwidth, support, etc, could be a big mess. Presumably they've thought this through, but hell maybe they got Madoff to run the numbers for them :)

Josh said...

Yeah, the biggest problem I can see is that you can't scale processing like you can scale data or bandwidth. Even removing GPU issues, which you can't - it's still very difficult to have one computer running at 50% for one person, another at 70% for a second and then have and then have the remaining 80% left over for a third.

You can do it, for some things, sure. But now tack back on all the specialized needs gaming requires as a software.

Maybe this is apples and oranges, but my daily work is "in the cloud". And the constraint the put? Processing. Salesforce will let you pull 10,000 records, sure, but only some many *at a time*. That way they know that the person next to you, and the person next to them is capable of so much for so long.

Multiply by the number of users, and you have a reasonable method of scaling. Without that governor, though, I don't see it working.

Or in other words, when OnLive says "any game", they mean "any we can afford to host". Which is the Achilles's heel here. They could totally do this, it's not even that technically difficult.

I just don't know of any gamer who'd be willing to pay for it.