Post by Dan @ 09:24am 02/07/09 | 39 Comments
Not to be confused with On-Live, which uses it's own client hardware device, all Gaikai needs is a web browser and the latest version of Adobe Flash.
Gaikai is a revolutionary new technology that lets you play any game online in your browser. In the age of the cloud, when all your documents, email, photos and videos are instantly reachable online, it seems archaic that you still need to install gigabytes of game files on an expensive PC with an even more expensive video card. And even then you can only play from that specific computer!The video demo showcases a surprisingly smooth experience and is purportedly running on a client with a 21ms ping to the server (similar to the latency most here would experience to local gaming servers in Australia). David demos a variety of recent games including Spore and Need for Speed Pro Street, World of Warcraft, Eve Online and even a Nintendo 64 emulator running Mariokart.
Gaikai takes a radical new approach – we host the games, we run them, we worry about hardware and software updates, and we stream them to you. Full resolution, full speed, stereo sound, low lag, no compromise. The only thing you need is a browser and an internet connection.
.
If it's as good as it looks, this may well be the perfect way to play a lot of modern games without having to upgrade, but we're going to stay skeptical at this point because there's no way to judge what kind of input latency we're seeing there.
Most gamers can probably recall playing a first person shooter or other fast-paced game on a laggy plasma or early LCD monitors and how bad that lag feels, so 20 or 30 more ms from an Internet connection could only be worse right?
Nevertheless, it's still very cool tech and would be extremely handy for say, trying a game before you buy. Purchasing a 20 minute trial of a full game with no need for lengthy installs or hefty system requirements would be way better than any demo, input lag or no - even from a Mac or Linux machine.
Latest Comments
39 Comments
Show
Follow Us
Advertisement
Log In
Posted 10:16am 02/7/09
The fact they have wrapped some half assed concept around it is wank, sure they can maybe demo it on a s***** clicky game for a few execs / pubs / and venture guys, get some cash then spend the next 2 years realising the model is completely imposible to scale
grats2u
Posted 10:23am 02/7/09
I'm not expecting to throw my PC out, nor see this become a huge thing. But that doesn't mean it has no point.
What its potential lies at is offshored processing power of modern applications (and games). What would gain benefit of these things? Well underpowered ultra-portable laptops (eeepc etc). My mobile phone. Devices like these where they have a battry life would do well to just run a flash interface for the virtualisation program, pay a subscription fee for some programs like CS:3 and be able to pretty much do what you already do on your desktop, with perhaps even better results (bar the retardedly small screens etc).
I don't think they are the future of all program applications (although that would stop piracy) but surely it has some niche.
Posted 10:27am 02/7/09
last edited by thermite at 10:27:16 02/Jul/09
Posted 10:29am 02/7/09
Posted 10:33am 02/7/09
ISPs in the US are already complaining and starting to consider caps - so far they've been stopped by consumer nerd backlash but as soon as they figure out the right people to bribe so they can get laws passed to protect them from doing whatever they want, they'll change the laws. Australia-style caps have been suggested in a lot of places.
Posted 10:36am 02/7/09
Posted 11:09am 02/7/09
Posted 11:20am 02/7/09
It should be very simple to work out if the model can scale. If you can run one new game well on a $2k desktop but you can only stream 50x of the same game simultaneously on a $200k server - then it doesn't scale, and you cannot mass market it to the casual gamer.
EDIT: On this topic, I'm a member of Sun Developer Network because of my work and they are really getting behind cloud computing at the moment. They are putting an alarming amount of energy into it. They make you feel like you HAVE to keep up or you will miss the boat.
Posted 11:19am 02/7/09
Posted 11:24am 02/7/09
This is the same thing, except they've made it in actionscript :/
Posted 11:46am 02/7/09
You're right. Its exactly the same concept which actually seems to come in cycles. Its in the cycle right now we're towards the end of a huge era of beefy client PC's and it being popular. We're probably about to hit a cycle in the near future of virtualisation and distributed computing where everyone will be all into the chunky servers and dumb clients.
And then in X number of years after that, it'll find benefits for the chunky clients again probably.
But that doesn't mean that there is no value in what they have to say. Time has passed a lot in 10 years, the internet has exploded, acceptance of computers has risen dramatically. Its not quite the same place it was 10 years ago.
Although I think bandwidth would be an interesting hurdle to its implementation. You'd have to really get ISP's onboard with it at a general home population level.
Although I also think this could be scaled down to a company level whereby dumb clients may become (potentially) popular again.
Posted 12:29pm 02/7/09
Posted 12:36pm 02/7/09
Like: on the client side, you'd need flash installed on every device? or on the server side constantly encoding something live quick enough and sending it via stream to the client?
If you're more specific maybe my lack of knowledge would be filled in and your point grasped.
Posted 12:37pm 02/7/09
haha what
Posted 12:41pm 02/7/09
yeah in hindsight that was a very pointless statement :)
Posted 12:44pm 02/7/09
Posted 12:50pm 02/7/09
Posted 12:52pm 02/7/09
Posted 01:15pm 02/7/09
Yeah obviously a hurdle, I was thinking about that earlier too. I don't really understand what cost it would incurr but bandwidth would easily be 10x more if even 10% of the population started using this as their full time PC replacement. I don't really have answers. Be more something I'd ask a communications engineer to estimate out the cost.
Well that already exists - its just spread out in our homes. I guess its a communist way of thinking, but you'd gain a lot of efficiency in some regards from a large serverfarm. When your PC is idle it does nothing. When you're not connected to the farm what you leave behind someone else will gain. Economy of scale?
Countering that point straight away: Right now games aren't really designed to be server-farmed for the clients. IE: most of them work by 1 video card rendering 1 output. Or even 2-3 video cards rendering 1 output. They aren't really optimised for 10000x video cards rendering 100000x outputs. Neither is the hardware (from my limited understanding). Unless they were you probably wouldn't really see that efficiency of scale.. I could keep listing hurdles here but I don't probably know enough to say whether or not it'd be dead in the water :P and obviously smarter guys then me think they can make money out of it.
Posted 01:19pm 02/7/09
Posted 02:10pm 02/7/09
Well there's a crushing response..
As for the network engineer, I meant more along the physical network requirements to be able to transmit that data. But your point still works too from the deployment side.. (Communications engineer is just a job description.. you just design the deployment of a communications network and infrastructure from my understanding)
Yeah, obviously for very time sensitive things it can be. But for a majority of day to day stuff it might be fine, or hell anyone knows that wow has a lot of lag. Or if you didn't know online play for Warcraft 3 has a default of 250ms latency from the client setting before any actual network latency (150ms on lan). People didn't realize that for years.
Well I think I understand a lot better now, the streaming video makes it incredibly easy to implement but when you say it like that I see problems at a larger scale like you suggest. From thinking about that 'hosting costs' it comes down to a consumer probably that they will want to only pay for a service if it offers value. IE if they can access their programs for less money then owning the software and hardware outright. So therefore the cost per connection would have to be such that they could charge the client less then the cost of the software which he would have to pay for the physical servers for hosting, the hardware to process the clients programs and also the cost of network bandwidth
The reason this would be much worse then say a specialized client is the network bandwidth would be very high (or so youtube has taught me and the quality would need to be higher).
Hrm. Ohwell this is the best kind of thread for me, where i walk in thinking one thing and learn something and change my mind :P
last edited by skythra at 14:10:38 02/Jul/09
Posted 02:13pm 02/7/09
well, I just don't like this whole 'in the cloud' thing anyway - because a lot of the time it's not about convenience or making things easier for the customer, its about control and, as pinkman says, the ability to come up with payment models that mean they can charge you any time you want to play a game. I quite like the "old" model where I buy a game, and it's mine to do whatever I want with - play it, sell it, eat it, mod it, whatever. I find these sorts of things scary (technical limitations aside).
Posted 02:33pm 02/7/09
1) I think the cost of transporting the user actions to a server, where it then as to be processed, then sent on somewhere else, then back again makes this method unrealistic for any multiplayer game (though console type games might get away with it)
2) Something _has_ to process all that video, and no matter what its enormously expensive to do in terms of cpu / gpu and that wont change even if you could effectively virtualised the video component
3) As dave mentions, the sheer bandwidth cost is boggling, the author is over estimating his latency to the server for a reason, that reason is latency is incredibly important - you'd be reacting to things half a second before the real client has been told you’re dead heh. This sort of latency might be doable in the states - but not here. Unless they've figured out how to defy the speed of light its hogwash in my view
4) Cost, as mentioned above the cpu costs to generate the graphics on anything but the s***tiest of games is likely to be super expensive, so any models using such technology simply cant scale on any current technology I'm aware of to a degree that you could affordably virtualised the current generation AAA title.
5) Compression, you cant really do anything on compression using flash so the game has to go though a whole heap of post processing to render the video to get it to the client, this is doable to some degree but further adds to the overall latency to the user, and more importantly the expense to scale since compression == CPU power
I know for sure they could probably build something that allows this to work at a conceptual level with currently available tech, hell we could do it in a few weeks I reckon. But your talking a few clients max per really expensively configured pc, you'd need like 50-100 a server which is an order of magnitude so far removed it makes the whole thing a bit silly.
To me this is clearly a VC farm which any half arsed due diligence would show is clearly not doable in volumes that would make the concept sustainable for any foreseeable future. As someone who has worked on putting different DRM technologies involving games into Australia I have some idea on the numbers you'd need to make this happen and unless they've divised something magical that no one else is even talking about this is simply not happening for all the above reasons
Virtual Box 3 is looking like some interesting technology virtualising DX stuff http://forums.virtualbox.org/viewtopic.php?f=1&t=18820
Posted 02:37pm 02/7/09
Jim Farrelly and Termite would agree with pinkman on that I think.
Posted 02:57pm 02/7/09
I think cloud computing is currently built on the principle "ignorance is bliss"
Posted 03:24pm 02/7/09
I like the idea of dumb clients and a huge massive server. I don't mean like one central network cloud with the rest of the world, i mean a internal server that is secured away from the external network.
I mean - at my work currently my PC uses 5% of its CPU about 5 hours of the day. the other 2.5 hours its stuck there processing one or two jobs. My workmate next to me has a similar routine which is just slightly offset, my PC is idle while theirs is working overtime. If we could pool the resources together it'd be great.
From a management perspective in the IT department I'm sure it'd be much easier to manage it if it was like that. Currently every user having near free will over their own PC, and always having to respond to broken PC requests etc. Upgrading the server if it was managed virtually would maybe be a lot easier without neccessarily going offline. You'd not need to get nearly as much processing power as not everyone uses their PC at once, just enough to cover whatever the bussiest time would be.
Or at least i like to think it'd be that way..
Posted 03:37pm 02/7/09
Posted 03:39pm 02/7/09
Playing with yourself is a favorite past time!?
Posted 03:45pm 02/7/09
I did and its already been implemented, just press that button on front of the case :)
It doesn't do much just a black screen. I'm trying to think of a way to tell that to our clients "When we process your data, we'll use this revolutionary technique called Off Switching".
Its all about the sales pitch :)
Posted 03:54pm 02/7/09
Posted 04:10pm 02/7/09
I agree that there's a lot of applications though where dumb clients make sense, but gaming isn't one of them - you just need too much processing power, low latency, high bandwidth, near-realtime feedback, etc, etc.
Posted 04:24pm 02/7/09
Yeah. What you are describing is how most engineering companies are configured for virtual engineering simulation (computationally expensive). It's not uncommon to have a large server (often Linux) and then all clients have some XWindows emulation client (e.g., Hummingbird) and remote that way.
However the trend in engineering now is clusters and batch queuing systems anyway, so that's different again, but very scalable and would suit gaming - but the basic principle of scale is still the same.
Posted 06:34pm 02/7/09
Posted 09:36pm 02/7/09
Posted 01:06pm 03/7/09
Q.I don't mean to troll, but whats the distance between USA and here?
A. 12065000m
Q.And what is the speed of light?
A. 3x10^8 ms-1
Now that you have realized that your statement was stupid, and that most lag happens from processing on routers and switches, we can talk about the real problem, scalability, most server farms would have blade servers, they are getting cheaper, What would be really cool is selling this software suit to companies or even tieing it in with steam, but with a monthly fee. The main blocking point is getting local ISP's to take on a server rack for this in their local datacentre to serve the people for a decent price (from memory the blocking point from telstra to host WoW).
Posted 01:20pm 03/7/09
Posted 01:20pm 03/7/09
Ctrl + F, "lag" - only one mention of lag on this page, and it's in your post. So, I'm not sure who you are arguing with. But by all means, troll away.
I don't think the real problem is scalability so much as supply and demand economics.
Is there any decent demand for this? If you want my opinion, I think not at this stage. It's an emergent technology, they have to try and create demand. Videos like above aren't doing it for me.
I think demand will come if game sizes increased dramatically. If you had a 60GB/month ISP account and a computer game was so massive in content that it was a 500GB game - then streaming games would really be in the mix of delivery modes. Storage is very cheap, bandwidth expensive - so there's a viable business model in there somewhere.
Posted 01:43pm 03/7/09
A much better idea however is not to have a central hub of processing power, but simply to distribute the power across client computers that are being underused.
Posted 02:41pm 03/7/09
I think a lot of proof of concept is using games etc because games are both intensive to a machine and it gives a kind of idea of what it could be (more then just virtualising text pad).
But also I think that the very things that this should be marketed at - like portable devices - probably won't be able to use them very well by default. I'd think things like PDA's using 3g to connect to a remote server for applications would result in the most terrible experience ever. And even wireless occasionally may get packet loss depending where you may be at the time. I think the market would be those who have those under powered devices and want portability.
Yeah the problem is you need to market something that appeals to a person. You need to teach them that they need this thing. Right now all they can do is what you already do on a daily basis anyway. If a tech demo comes out with like world of warcraft running on some guy's phone, while he catches the train and listens to MP3's and alt-tabs to his university assignment in the background then yeah that would be way cooler.
last edited by skythra at 14:41:41 03/Jul/09