There are coding techniques that will reduce the "feel" of input lag, but I doubt they are using them. For example, they could render and transmit each screen larger than needed (IE add a bit of overscan). By pan & scanning in real time on the client you "may" be able to fool some players that there is less lag. You'd have to have sophistacted routines to tweak the scaling of the image, but it's the same sort of techniques used when created panoramic photos or correcting wide angle distortion so nothing complicated really. You could also zoom in & out of the image to create the impression of immediate feedback for W & S keys. A&D straffing would be a simple pan, parallax would go out the window but again, I'd say it could improve the feel.
The hardware debate is a genuine question. I've seen a few articles now questionning their ability to recieve input for and render a million games of Crysis at the same time, whilst simulataniously encoding the output to a coded video and transmitting it across the globe. Not only would their bandwidth costs be astronomical, is it even remotely possible to build server farms that can process such an insane amount of graphical horsepower without literally buying a new graphics card for every new potential gamer? I know the power of economics of scaling but this is something else.
Good point Bauul, i had thought of it, and it is a serious question, what computers can render Crysis at great resolution? And how many versions of crysis are they running? One. So thats a computer each...
But then most people who play games like crysis at max also aim for uber high resolution, what kind of processing power do you need to run crysis at max settings but 720p resolution?
Resolution is less important than which of the engine's features are turned on with something like Crysis. If you're rendering hugely complex shadows, shaders, physics, lighting, textures etc. etc, that you need to output fewer pixels at the end isn't going to improve things much, especially when 720 isn't exactly a tiny resolution. We're not talking 320x240 here. Besides, any reduction in resolution output is completely eclipsed by the need to encode every single frame to video as you go along.
Bauul, this isn't going to be sent global, it's going to be set ups of servers in Data centres all over the world, they have a radius of 1000 miles, and if this does go massive, I am guessing it will be far less than that, and there will servers all over the place for this. Found on Google they received funding from Warner Brothers etc a couple years ago of $10m+, and that was from one funding, they have plenty of money, and I don't see why companies would fund a company with a product that doesn't work.
That still doesn't answer the question of how they are going to fund the hardware. Regardless of how many or few server farms they have, unless they've invented some crazy new hardware us mortals know nothing about, if they want to run Crysis on maximum for each player (which I've heard they do), they will need about £400 worth of core consumer priced components (at the very least) for every single player. And in three years time (at the very most) all that will be redundant and they'll have to buy it all over again. My idea of server-sized hardware may be wrong, but as far as I can tell, the sheer cost of maintaining a decent gaming rig for every player who ever wanted to use them is going to be absolutely huge. Regardless of how much investment they have, if they want to break a profit they'll have to earn enough money to buy every player a new high-end gaming rig every few years, even before covering the cost of the games themselves. I'm sorry, it just doesn't seem to add up for me.
I think it is good that this is not going too hyped up, that way when it is released people will be surprised, much like the PS1 release and the start to sony's gaming division. Who knows if its going to work or not only time will tell. As for hardware, if you already own a computer you dont need the little console thingy to stream the games, you only need it if you want to stream games on you TV.
They claim this adds negligible overhead as they have developed bespoke hardware to encode. Crysis probably will need one GPU per player, but they will be banking on not everyone being simultaneously online. Their press conference video mentions virtualisation so multiple instances of lower end games can be run on one physical server. I guess this might be built into their pricing model, with e.g. Crysis costing more to buy/rent than say World of Goo. I guess one big problem is that the need for proximity to servers means that even after a worldwide roll-out they won't be able to benefit from time zone differences by repurposing US servers to cover Europe during the European evening peak, then repurposing Euro servers to cover the US during the US evening / Euro middle of the night. They must have a pretty chunky data pipe to stream 5 Mbit / sec (actually a bit less - they say that's the minimum rated broadband connection you'd need for 720p, but actual peak throughput will be a little less, and average may be less again) to thousands of gamers simultaneously, but that is doable. Latency, scalability, hardware costs and bespoke software / hardware design are FAR bigger issues for these guys. @Jenny yes there are techniques to fudge the input lag issue, but they claim their microconsole and PC / Mac plugins are effectively dumb terminals, so I doubt they are using them. Also transmitting an overscanned image to allow scanning would increase their bandwidth issues, and that sort of thing (e.g. pan/zoom instead of true parallax movement) looks pretty horrible and smeary - gives me headaches, like the oldschool Doom 1/2 games with their skewy parallax adjustments when you move. Can give a reasonable approximation of camera rotation though (i.e. turning, looking up/down). Anyway, it is only a hack - it doesn't really fix the issue.
I share your scepticism, but the hardware cost is mitigated to an extent by the fact that not everyone will be playing simultaneously. To give a good experience, though, they will need to ensure sufficient server capacity to allow every player to play what he wants, when he wants, which means all that hardware sitting idle during quiet periods just to ensure they can cope when they get slammed during the 6-9pm peak.
re: Bauul I think current MMO's plan for peak load to be %30 of their customer base at any time. This might be comparable then. They don't need enough gaming rigs for every customer, merely enough to service whoever is currently online. Plus of those online, only a certain percentage will be playing the most demanding games. Even despite all that, the costs still seem pretty heafty. Especially when you take into account bandwidth. I wonder if they might use an "On Demand" / "pay as you play" style pricing model/option? Since their costs seem to be tied based on how much time you play (bandwidth + timeslices on the hardware) it would make sense to charge accordingly. I still don't see this as more of a novelty for the high end games though. If you want to play Crysis to beging with, or play twitchy FPSs anything more then casually, having your own box is likely the way to go. This is something that would seem better suited to the more mainstream titles with wider appeal. And it would make it easier on them as the performance requirements would probably be lessened. Less Crysis and more Starcraft II?
You can't really compare OnLive to a MMO, though, because the MMO servers aren't really doing anything more complicated than object handling - what monster is where, what map you're on, what treasure is around, etc. The rendering of the actual world (i.e. how you visualise those objects) is done by the client PC not the server. An OnLive server has to do everything - management, rendering, streaming, everything. And that effectively means the equivalent of one (high powered) PC per connected user. They also need to price themselves at a sensible price point. After all, if Joe Q Public works out that it'll cost them $100 to complete Crysis, then they'll just go out and buy themselves a copy of the game. Of course, they'll probably forget about the $1500-worth of PC they'll need to play the game, but that won't be a part of the "gut reaction" to the pricing model. I'll be surprised if this (a) lives up to the hype; and (b) actually gets any paying customers, playing more than just Tetris at 3pm - but I'm willing to be surprised. Let's just say that I'm not holding my breath
just been reading this article reason 1 is very annoying problem in UK, and as Gadget Show on Five said, 99% of ISP has it. reason 2 has been discussed here. this also raised point of 720p video is being compressed, so major quality lost if it is being scaled up to say 24 or 30 inch. reason 3 is of most concern, almost everyone now uses a Router, so the what if your sistor starts steaming video on your 5Mbps connection? or what if that Folding just completed, you'd have to live through 1minute of laggyness? reason 4 is quite sad for those such as Garry, but im sure offline version should keep it going. alternatively, they may find a way to provide it, always keep mods up-to-date, similar to Steam. reason 5 meh from me. reason 6 depends on personal views, i prefer to backup all my savedgames, so this might make it difficult for me. it all comes down to the internet connection and whether you are willing to accept quality loss. it's great for console crowds, but may not look so good for us PC gamers
This is going to be a bandwidth hog. Unless people are running QoS on their home connections, say goodbye to multiple users using the internet. Not everyone has powerful connections either. I only have a 1.5Mbps download speed and 500Kbps upload. This is sufficient for Xbox Live and whatever else I do. However, sending all the video data, sound data, and playing with other multiple users. Am I going to have a lag free gaming experience comparable to my 360? Yeah I doubt it. Netflix Instant view is a good example of where this type of technology is at. I've tried the trial for 2 weeks, and to be honest, it didn't work that well for me. My connection doubled the minimum requirements and I couldn't get more than "2 bars" and most of the time it was at "1 bar". To give an idea what that looks like, imagine a Youtube video being blown up to fill the screen. Artifacts everywhere! It leaves a bad taste in your mouth. A way to get around this problem for Netflix is to allow a buffer override, where I could let the video buffer for a longer time to allow a better quality playback. Instead I got the video instantly, 2 minutes into watching the video the playback would stop and it would throttle the quality down to "1 bar". Overall I wasn't happy with my experience. If I had a faster connection, say 5Mbps or so, I could see this OnLive service working for older titles. Say Xbox & PS2 era games.
@ sui_winbolo: It doesn't change a thing if it's old games or new games. The video stream remains 720p.
Well a solution that would help save bandwidth is run older games at their native 480i/480p resolution. That's what I was getting at. (coming from a console game point of view that is) Sending a 720p video stream and the capability to provide a decent gaming experience seems to be wishful thinking. On a LAN this would be easy because everyone would have at LEAST 10Mbps connection and only a few hops between devices at MOST. I'm currently a Network Engineering student, so I'm just rattling thoughts around my head of how plausible this could be. I do not know much about compression though and how much actual bandwidth this would all require. I'm finding it difficult to imagine. Especially since Netflix has left me sour towards instant streaming media.
