Join 3,553 readers in helping fund MetaFilter (Hide)


Paradise by the console lights
August 28, 2012 4:54 PM   Subscribe

OnLive Lost: As founder and CEO Steve Perlman departs, The Verge looks at how OnLive failed and what remains of the revolutionary, restructured cloud-based gaming company.

Bonus from 2009: Why Onlive can't possible work (previously).
posted by 2bucksplus (22 comments total) 5 users marked this as a favorite

 
This week's Joystiq podcast discusses the shadiness of the restructuring in some depth.
posted by griphus at 5:04 PM on August 28, 2012


"Technical genius"? It requires a physical machine for each player. And you add in the network latency, which means that it's slower than if that machine were right in your own house.

I've still never seen a convincing argument why this stupid system would work.
posted by DU at 5:07 PM on August 28, 2012


Everyone I know who has tried it says it is surprisingly good. While I was skeptical about latency as well, it apparently is way less of a problem than you'd think.

That said, their problems appear to have been more of a business than a technical nature, which is not what I would have guessed initially but makes sense to me now.
posted by wildcrdj at 5:09 PM on August 28, 2012


Someone here doesn't understand how virtual servers and networks work.
posted by Burhanistan at 5:20 PM on August 28, 2012 [1 favorite]


Essentially they needed a critical mass of paying customers who never materialized - same story as many a startup with a good idea ahead of its time. OnLive worked; I was skeptical when I tried it at that GDC and remain skeptical of the image quality and latency to this day, but the fact it is it was well within tolerance levels. But it wasn't quite a Netflix for games, and in the end I don't think that's something people want just yet.

They launched too early, misjudged the demand for their (perfectly workable) product, and then Perlman screwed up the rest. They should have taken the money a year or so ago; everyone would have walked away rich and happy and satisfied that they'd nudged the world of gaming in the right direction. Oh well!
posted by BlackLeotardFront at 5:20 PM on August 28, 2012


Anyway, the OnLive desktop app is fun if only to have a Windows instance displaying on an iPad.
posted by Burhanistan at 5:21 PM on August 28, 2012


I've still never seen a convincing argument why this stupid system would work.

If NVidia had, as hinted at in the article, been willing to virtualize the GPU, then that'd be plenty.

Network latency vs. physical latency is not the simple question you think it is - there are good reason to believe that it's often faster to request something from the network than it is to request it from disk, even on performant machines.
posted by mhoye at 5:22 PM on August 28, 2012


I was surprised how playable this system actually was, but I don't think it ever really found much of a market. Most casual games run fine in-browser or on mobile systems and "hardcore" gamers are performance junkies to whom "pretty good" latency is not good enough.
posted by atrazine at 5:24 PM on August 28, 2012 [2 favorites]


If NVidia had, as hinted at in the article, been willing to virtualize the GPU, then that'd be plenty.

Any technical advance that could have speeded up OnLive could have also speeded up local gaming. And there would be no network delay for local gaming, so it would always be ahead.

...there are good reason to believe that it's often faster to request something from the network than it is to request it from disk...

Generated graphics are, by definition, not read from disk.
posted by DU at 5:32 PM on August 28, 2012


Steve Perlman ... I just remember years ago when he was hired into the department I was in at Apple. Boy Genius was the implied job title as he immediately jumped in and started doing all kinds of stuff. Don't remember anything that ever shipped but he did have loads of ideas...
posted by njohnson23 at 5:52 PM on August 28, 2012


I've still never seen a convincing argument why this stupid system would work.

I assume it's similar to the reason Comcast doesn't require multiple petabits of bandwidth?
posted by Talez at 5:54 PM on August 28, 2012


I really liked the service--or at least the promise of it--but as the article notes, lack of games and high prices were real problems. Lots of popular games and developers completely missing. Prices were high too, certainly much higher than Steam, and why risk $50 on a game-as-service that could disappear? They had a couple good gimmicks and sales--like $1 for a new Saints Row game, the only money I ever gave them--but that's it.

The bit about the service not being VMed is surprising to me--I would've thought that bit had been figured out before launch.

Also their Arena and demo services were very neat, and latency while playing actual games was within tolerable limits.

Somebody will make this work, sooner or later.
posted by aerotive at 5:55 PM on August 28, 2012


Any technical advance that could have speeded up OnLive could have also speeded up local gaming. And there would be no network delay for local gaming, so it would always be ahead.

I think you might be missing the point of the product, that you could get a premium gaming experience on top end hardware without having to secure the hardware yourself and you could get it consistently for a far lower cost and on just about any device that had an internet connection.

That's why the service was compelling, the speeding up you're referring to was handled by peering with major regional ISP's directly, without going through an ISP. It really worked, in fact I believe they colo'd with comcast/etc to further reduce network latency. That part of the product really worked btw.

The GPU thing nvidia had available meant you could reduce the machine footprint in the datacenter to service customers, not that it made the service faster or more better, it was a cost reduction thing.

Perlman aside, what they were trying to accomplish and mostly did accomplish was pretty staggering, it was the equivalent of building a national ISP and cloud computing provider with direct relationships with every major carrier in the US and every major ISP in the US - in just a couple of years.

I've done this before, I do it for a living, and it's a hell of a thing to do it as a start up, so you can shit all over onlive as a product and as a gaming relationship company but the technical aspects of what these guys pulled off was pretty and is pretty goddamn amazing and it's a damn shame they didn't get the content end of things worked out because they had the other bases pretty well covered.
posted by iamabot at 6:01 PM on August 28, 2012 [7 favorites]


DU: ...add in the network latency, which means that it's slower than if that machine were right in your own house. I've still never seen a convincing argument why this stupid system would work.
You understand what latency is, and you therefore have a more sophisticated understanding than the average person about the nature of time itself, probably. You overestimate the ability of the average person to notice or care about such things.

Modern television sets, with their digital filtering, commonly introduce several frames of latency into every signal that's piped into them, TV and video game alike. The viewing equipment itself is messing up the timing of video games that require fast reactions, but most people don't notice.

There are plenty of serious video game nerd people who do understand latency and object to it, of course, but the serious video game nerds were never the target market for a system like OnLive.
posted by Western Infidels at 6:19 PM on August 28, 2012


So good product, not so good boss.
posted by Kevin Street at 6:22 PM on August 28, 2012


An important thing to remember is that, largely because of consoles, the PC Hardware Wars are over.

Gaming performance comes mainly from the video card. A 9800GT 512Mb GFX card from 2007 will do a very credible job of playing nearly any game out there -- because that game will have been written so an Xbox 360 can do a credible job of playing it. And while you need at least a dual core, these days everything with a cpu has one of those.

Given that, anyone who gives any kind of shit about PC gaming has a tolerable machine, for not much money. So the market (people who care about gaming but don't want to spend craploads of money on it) is much smaller than it was.

I can see this working much better when keeping up with the Hardware Joneses was more of a deal. There's probably also room at the steam sale end of the market - a buck or two to play five or six hours of Modern Warfare Blargedy Blarg might be tempting.
posted by Sebmojo at 6:34 PM on August 28, 2012 [1 favorite]


I've still never seen a convincing argument why this stupid system would work.
Well, everyone who tried it said it did work and why wouldn't it? If the servers you played on were local to your city, the latency could be in the tens of miliseconds range, a lot lower then screen lag.

According to John Carmak it actually takes less time to send a packet to the other side of the world then it does to update a modern LCD because there's so much buffering and crap going on in the system.

Whether or not it was going to be commercially viable, well, I guess not. But even with just one server per player, unless you have super hard-core players you can still get some efficiency out of it, since not everyone plays 24/7
posted by delmoi at 9:47 PM on August 28, 2012


Oh one other thing, and I think it probably wouldn't have bothered many people. But go watch some 1080p footage of starcraft 2 games online. It looks pretty good, but it's compressed and there are obvious compression artifiacts.

Then try playing the game on max settings with a top-end graphics card. The resolution is the same, but the difference is incredible. The game is breathtaking. It just looks beautiful without any compression artifacts at all.

If you played and watched compressed video of your game, it would look like crap. It might be inperceptable to most people, and you might not even notice it consiously, but it would be there.

Video compression works well for "traditional" movies, like videos of people talking and maybe moving around, which comprises most of what we watch on TV and in the movies. It's slightly blurry naturally and you're looking at people's faces and bodies.

In video games though you want to see a lot of sharp details, text, graphics, icons, etc. When you apply video compression to those things, it doesn't look as good.

But, most people won't notice it. They would probably prefer the locally rendered image in an A/B test but they won't notice it without a direct comparison.
posted by delmoi at 9:54 PM on August 28, 2012


I've still never seen a convincing argument why this stupid system would work.

It did work! I played through Arkham Asylum for .99c on the PC client, and pre-ordered arkham city to get the controller/console.

The controller is great! Its one of the best I've ever used, the hardware and cabling and everything was nicely packaged, and had everything you needed. The box was a tiny set-top box, and the PC client was about 1Meg.

The graphics were good, there were issues of latency, but for something with an auto-save system, it was mostly ok.

There were a bunch of problems, though:
1) On initial launch, after sign-up, it would frequently just give error codes with no explanation for what they meant. I tried it at launch and the best I could do was spectate, I couldn't actually play anything.

2) The playpack games were old and pretty crappy. Nothing high end that you needed a supercomputer piping to you, more like 10-year old tomb raider games that most any system could play.

3) The client was _not_ forgiving with network drops, or even pausing. If I paused Arkham City and had to leave the room for 5-10 minutes, it would disconnect me, and I would have to spend a few minutes just reconnecting. Fortunately my game was autosaved, so I never lost anything, but it was still a hassle.

4) The big name games lagged, outside of Arkham and Saints Row, it was always too little, too late.

5) DLC was not an option. (they might have fixed this later, but I missed out on some challenge packs). Neither was modding.

6) Anything that wasn't in that specific mold would be unsuitable, meaning twitch FPS, RTS, MMORPGs, etc just weren't tenable for any combination of reasons listed above.

But, like I said, I made it through both Arkham games and found them very enjoyable (i'm going through them again on steam, because I liked them so much, but damn their Games for Windows clients. Onlive had them beat on that!).

I went from hating the service, to trying it a few months laters, to subscribing to the playpack, to having the console, and then finally just going back to steam and my regular consoles, since the incentive just wasn't there.

Its a shame it didn't work out, though. Seeing high-end graphics streaming through a 1mb client on my old and busted computer was pretty damn cool. I feel for the employees who are another casualty of silicon valley egos.
posted by lkc at 10:47 PM on August 28, 2012 [2 favorites]


I discussed this with some people from $InternationalSatelliteBroadcaster a few years ago, they wanted to implement it in their set top boxes.. told them it might work in the US, maybe korea and japan, but network speeds in most of europe just aren't high enough. Told you so, $InternationalSatelliteBroadcaster !
posted by 3mendo at 12:53 AM on August 29, 2012


It is indeed surprisingly doable, specially for non-shooters. Give it a try.
posted by CautionToTheWind at 3:17 AM on August 29, 2012


The games industry really does seem to attract the assholes into management positions for some reason.
posted by GallonOfAlan at 6:51 AM on August 29, 2012


« Older The Museum of Computer Adventure Game History in T...  |  Matte World Digital, the visua... Newer »


This thread has been archived and is closed to new comments