GDC: Why OnLive Can't Possibly Work Comments by Richard Leadbetter

@rYZE

http://news.bbc.co.uk/2/hi/business/7456...

look it up.

Hydrogen plus oxygen = Water (give or take).

Also when I buy a WiiWare game the idea is that I own it. Sure it may not be in a box or a physical product but I don't have to return it next week so for as long as the service exits (and who knows what happens beyond that) I own it outright. That doesn't mean I own a physical copy, it just means I don't have to pay any more for it at a later date etc.

That's a hydrogen car, you cock. Water comes out of the exhaust.

Yes and it uses Hydrogen and Oxygen in the first place you fuckin dick, which is what creates the electricity the car runs on. Which like I said, is give or take, water(which is just H2O anyway) to run a fucking car!

Dick head

No, it's the water that runs on car :P

@Darthus
I think it is techically impossiblie to get video compression down from ~500ms to ~1ms in one jump. Maybe they have done it, I don't know but new technologies tend to progress in stages. This seems like going from stage one to stage five and missing out two, three and four. Why would they use that for a gaming system first? Surely they would license it out and make a huge amount of money, then develop a gaming system with the small change in their pockets from the vast amounts of money they would make in other areas. Lets wait and see the meat, because ther seems to be only bones so far.

Yawn Freeetardlol,

So my car runs off of Carbon Monoxide!!!! We're saved!!!

7 years is a pretty long time when it comes to any kind of computer technology.

If they had made the process visible then it would have evolved over that time in a way that you could suddenly accept?

Just because you didn't see it though then it's not happened...

@Darthus:

The problem with server hardware is that it is much more expensive than home PC hardware, which in turn is much more expensive than mass-produced console hardware, if you look at it from a performance/cost perspective. You get the most "bang for your buck" if you buy a low-end computer. A single server with ten times the processing power of an xbox costs a lot more than ten xboxes. A single server that could virtualize 10 gaming PCs at once is going to cost a lot more than 10 gaming PCs, and you also have to pay for the overhead you need to compress the video, stream the video over the internet, and handle virtualization itself.

I can see how you could do this technically, but the thing that really makes me doubt this is how they're gonna get the economy behind this to work. They're gonna have so much economical overhead in terms of staffing, hosting centers, bandwidth costs, keeping spare servers for peak hours etc, that they simply cannot offer consumers a service that is cheaper than buying new games and gaming machines now and then, while still offering the same quality that dedicated local hardware can give you.

Something's gotta give, and my guess is that it's gonna be quality.

Why would you know about it?

Er, because companies announce big deals like that as it raises their share price, and insiders leak details of said deals to newspapers and web sites?

They are going to be running gaming specific servers with Quad SLI top of the line GPUs, 32 gigs ram, multiple quad cores and SCSI hard drives. Then they are going to split those into a number of virtual PCs for the game. That server alone I just mentioned can probably run 16 instances of GTA4 at 720P resolutions.

Have you ever tried to run a game through a virtualised system? They run like crap. You'd be doing well to have a single game playable, 16 is fantasy land. As for those system specs, they best have come up with some fucking fancy-pants software to make the GPUs and CPUs run 16 games at once without reducing it to a stuttering mess. GTA 4 is a fucking pig to run native anyway, let alone 16 instances virtualised.

They need to release some technical details before we start speculating to that extent.

Why would they use that for a gaming system first?

Never mind the massive technical question marks, this alone justifies the scepticism. Are we expected to believe that they've invented this groundbreaking new compression and streaming technology and the first thing that popped into their head was playing videogames? It'd be much simpler to prove your tech worked by streaming high-def video, as adding user-input into all that just complicates things.

Do you really think the original guys that claimed they could run a car on water meant just pouring it into the tank and off you go?

Retard.

All that's happened now is the basic principles of the idea have evolved and they have figured out it's the base elements that make water that they have to use in a certain way to create the effect they were proposing back in the day.

It's not like they claimed it ran on tomato sauce and suddenly they switched to Hydrogen mixed with Oxygen and claimed that's that same thing.

@baranga:
"Again, why would they create something like this if it doesn't work? Do you really think they fooled all those companies? Do you really think this is a scam, while the world is watching them?"
I want to be very clear on what you're saying -- do you think that a company would not make public claims they can't back up? Or promise something that they can't (yet) deliver on?
That's whatI'm reading, and it seems a bit circular. They wouldn't do it because they'd get called out on it, so the people calling them out on it must be wrong because they wouldn't risk getting called out.
You point to the companies that have been "fooled", saying that they're too smart to get suckered so they must know its good. First, tell that to Trip Hawkins' 3DO partners. Second, depending on the "agreement", the other companies might not give a damn if it works or not:
"Hey Activision! I want to pay you money to play your games over my vaporware. You don't have to do anything -- just let my buy and use your existing games for a fee. There's no risk of the games being pirated. No cost, no risk, potential for profit? What do you say?"
"Uhhhh. . . sure. Why not?"
The only thing that seems sure from their model is that hte games can't be pirated, so the companies IP isn't at risk. As has been noted, it sounds like OnPlay is using pre-existing game code, so there's no cost for the developer to make a port. I would asume OnPlay pays the companies, then recoups the costs through the user's fees -- it doesn't make much sense for software developers to pay OnPlay to make money renting their game out. Why wouldn't a developer say "you get it to work, you can pay us to use our games"? If the system is vaporware, then there's no loss to the developer.

Maybe it does work -- but just because a PR dude says it does doesn't make it true.

@Freeload
Thats fair enough, they've managed to keep it a secret for a number of years but I'm not buying into that either, this is the era of blogs and big mouths. The more I think about it the more sceptical I get, and nothing you have said is changing my mind, unfortunately.

@PearOfAnguish

Have you ever tried to do anything like they are doing?

How the hell would you know a single thing about how they are handling this.

You are using examples of stuff you know on something you know nothing about.

I once inserted a videotape into a slot to watch a film. Is that how I run a 3D movie in and Imax theatre?

I don't have any clue about Imax theaters but I once inserted a videotape into a slot so it's probably the same thing right?

I mean I'm probably and authority on Imax 3D cinema now.

Did you hear anything about the Wii MotionPlus before it was shown?

Did you see or hear anything about the new Zelda DS game before Nintendo showed it?

Did you know that EA, Take Two, Ubisoft etc were working with these guys in any way shape or form before they announced it the other day?

Then it simply cannot be real.

Someone who works at Rockstar North in a high up position didn't know there was a DS version of GTA in the works until they brought in a copy from Rockstar Leeds for the Edinburgh Qa team to test. Did you know that?

These things can and do happen and you are as likely to know as any other joe nobody.

You're talking SHIT mate.

James Bond flew in a hydrogen powered jetpack several decades ago. The waste product was WATER.

This is real technology - but it isn't practically or commercially VIABLE to be placed into mass production yet.

Hydrogen cars need Hydrogen. Where does this come from? Why, WATER - but how do you separate it? Using ENERGY, from somewhere else. So - it just moves the pollution to somewhere else, at the moment.

Just like this streaming idea just moves the expense to the server farm. It's still hideously expensive to run, and doesn't work too well considering our current broadband infrastructure. Just like hydrogen cars and how they're impractical given our current energy/fuel infrastructure.

These are very interesting technologies, and they WILL gain momentum - but we already KNOW about them, and we know their shortcomings.

Just because you've just got Sky Digital - stop telling me about shows that have been on repeat for the past 3 fucking years.

You 'tard.

No.

James Bond was suspended on wires and that thing he was attached to was a prop.

Turnip!

http://www.youtube.com/watch?v=TOfzbaglUOo

Sooner than you think...


The technology has been ready for years. Just because there are people like you who are not ready to take the leap that doesn't make the technology any less real or viable .

Just because not everyone in the world has high speed broadband that doesn't mean their system doesn't work. That's just logistics.

They never claimed such things as everyone can use it perfectly etc. They claimed that if you meet certain criteria in terms of the connection you have and the hardware and you download their application and so on, then it will work.

There's nothing yet that anyone has said that has convinced me that much is not accurate.

They even stated you would have to be within a certain distance of the servers due to the time, speed of light etc, and that is why they used the 1000 miles statement and not that you would have perfect lag free gaming across the known universe and so on.

From what I have read and watched I'm going to choose to believe most of what they are saying, give or take a few details, and believe that the service exists in a form pretty much as they say it does and that it does pretty much what they say. I'm going to believe that at worst the service will offer SD gaming on demand with none of these issues that everyone else is bringing up, based on what I have seen, and for me that's all good.

It may not be 1080p 120fps motion controlled virtual reality but it seems to me that I should basically expect the equivalent of what I got last gen on say my GC, in terms of res and framerate etc, running on my TV directly and streaming from online somewhere, with no need to upgrade hardware or download patches, with cross compatibility etc etc.

Sounds pretty sweet and just as groundbreaking as they claim.

^ What a fucking 'tard we have in our midst:

http://en.wikipedia.org/wiki/Bell_Rocket...

http://en.wikipedia.org/wiki/Thunderball...

1965! Amazing, eh?!

Cock.

All I can say is put any spare cash you have into Intel, AMD and Nvidia if you believe this nonsense. Because they will be selling millions of CPU's and video cards over the coming months, or will they? I'm sure these companies wont give a sh*t whether they are selling to the public or to server clusters. If this OnLive service is true or bullsh*t you probably find out with the said companies share prices over the coming months.

@PearOfAnguish

Have you ever tried to do anything like they are doing?

How the hell would you know a single thing about how they are handling this.

You are using examples of stuff you know on something you know nothing about.

You do know I was replying to someone's speculation about system specs, and that I did state that such speculation was pointless until we knew more about the technical details?

You need to calm down, pal, you're starting to sound like a proper foaming-at-the-mouth lunatic.

I honestly can't see our UK ISPS willing to commit their already peaking bandwidth to this. Especially when they're still in the process of trying to keep up with the current demand.

If I was to make my own onlive service here's how I would do it:
1) License this software: http://www.streammygame.com
2) Build 1000 half-decent PCs at a cost of £500 each (total = £0.5m). With a contention ratio of 10:1 that allows me to have 10,000 customers.
3) Sell the service at £30/month
4) Assuming I get all 10,000 customers, after 1 year I've made £3.2m, enough to cover my initial purchase of PCs & other costs.

Clearly I'm greatly simplifying here, but I think it demonstrates the point that this service is feasible.

@PearOfAnguish

You sound a bit like a rapist but who am I to say what you are capable of.

^ I watched that clip when it was first broadcast - you cock. You're not teaching anyone anything, you 'tard.

@hiddenranbir
Have you ever watched BBC iplayer? Its video is just about watchable even though I get enormous amounts of blocking on it with a 10meg connection. the video is already encoded with no alterations needed. Now imagine a game were the video needs to be encoded 60 times per second not to mention the processing required from your inputs and the uploading of that! Jesus, give me strength. These guys might as well have said they've invented interstellar travel.

Shush now, Freeload, it's quiet time. Go have a sit down and a nice cup of tea.

You can have non-polluting power plants. Windmills, dams, ocean-thingies and probably more. Those in turn can help create other non-polluting sources. However, in any kind of production process there is always waste.

As for On-Live. This is a nice idea. Humans have gotten things to Mars so why should this be a problem? ;)
Also, the video feed itself just goes on the same basis as IPTV and VOD through IPTV. VOD has stop, pause, fastforward, etc and those controls have immidiate effect. I see no reason why this shouldn't work as well. Besides the obvious problem of the machines running the games.

@Ryze

So I was wrong about a movie prop...but somehow I feel it just kind of validates my point about the water based car all the more...

I'll take that.


The time it's taken for this to become commercial, and it still isn't really, doesn't change the fact that those guys were pretty much right all along.

The difference with OnLive of course is that they were telling everyone behind the scenes about the basic idea 7 years ago and in a few months we will be beta testing it and then hopefully shortly after that it will be out in the real world.

That's what they are claiming and so far I still don't see a single thing that convincingly shows me otherwise.

@PearOfAnguish

And you're possibly controlling and violent too.

I mean you don't know me and you're telling me not to speak and what to do.

That's quite frightening.

I hate to think what you might do to someone in the same room as you if you didn't like what they were saying.

He's just got bad DNA. He can't help it.

/gives up

Have the meds not taken effect yet, Freeload? You should take the whole bottle just to be certain.

"VOD has stop, pause, fastforward, etc and those controls have immidiate effect. I see no reason why this shouldn't work as well."

The difference there is that you're just pressing a single button every so often, whereas in a game you might be hammering buttons and pressing analogue sticks and all the rest of it. That's a lot more input to handle, and reaction time is massively important in fast-moving games, even the slightest lag would be enough to deter a lot of gamers.

@speakingur
Because going to Mars is feasible if you know what force you need to get out of Earths gravity which was known in the 18th Century. Yet we still didn't get to Mars until the middle of the 20th Century, do you see a theme here with stages of technology?
Now, run a fibre-optic cable from my house to every other house on the planet and what OnLive is proposing might just work because it bypasses all the bottlenecks in the system. Either that or someone at OnLive is of Einstein genuiss who just happened to have kept it quiet for the last seven years. To be honest, I don't even think Einstein could make this work.

@polaris70

Look, I can't say about anything else but they had it running live on the show floor at GDC, supposedly from their server 50 miles away, and no one so far has said it didn't work. They have the games running across multiple platforms, including Mac, and Crysis isn't even on Macs so I've been told. They had the conference with all the interface being demoed. They officially announced that major publishers are already supporting them, so something must have convinced them of the service for them to jump on board. I mean these publishers are not completely retarded you know. they are veterans in the industry and have track records of successful products and services. They are letting people sign up for the Beta now. They have given solid dates (although whether they will stick to those is a different matter, but that doesn't mean it's not real, it's just they might do a Nintendo on us). They took the time to answer any concern anyone came up with with.

They can't do much more to convince us that that, until it's actually out and by that point they don't have to convince anyone because it's out.

I say they have provided enough proof for the stage they are at and all you people who are still doubters are clearly the type of people who just want to doubt everything because you are negative about everything unless you get to experience it too.

@PearOfAnguish

And apparently you possibly also like to prey on those who are weak and sick and need medication, simply for your own pleasure and amusement.

You are probably racist too and sexist and discriminate without thought.

It just gets worse and worse.

I'm off to bed :)

Assuming it all worked as promised, would you want to give up your console and/or computer? I like having my Xbox there for video and music as well as games, and when something goes wrong I can take the disc out and hit it. The PS3 does all kinds of cool media centre things. And I like having a collection of games on the shelf. Do you want to lose all that for a system that relies on your internet connection being 100% reliable?

Well I can't deny that, Freeload, I am the sexiest.

@Freeload
They have provided absolutely no proof accept from a controlled environment and taking them on their word. I hope this works, I really do but I will wait until the end of the year when we will have at least some proof with a public beta test. I hope they succeed but until they do I'll file it under 'cold fusion', not because I'm heartless but because I'm sceptical of guys standing up on a stage introducing technology that doesn't exist or is 50 years in the future. I stand to be corrected, I hope they do it.

VOD also uses caching and hard drives to manipulate linear media. It wouldn't matter if there was a delay, but there's not because pressing pause just pauses the video as it's being displayed on your local machine. rewinding just backs up to the cached video on the HDD or in RAM.

Your inputs on a gaming system need to be picked up by the server and interpreted by the game, with the results being encoded, then forwarded to you as a compressed moving image.

Pausing that image, or even recording it would be very easy, if required for any reason. Different concept.

I'll enjoy watching this develop.

It has to beat the Wii and 360 (in terms of quality / cost / reliability / convenience), to be viable. Integration with other living room devices such as TVs and digital boxes will probably be key.

IPTV is available in many places outside the US, including the UK in the form of BT Vision.

Edit: Man, that makes me sound like some viral marketer... but yeah.

Lot of tech heads here think this wont work maybe they are right,but Ive look at this from a simple point of view that nobody else has mentioned.
Would all those big name publishers have backed this if they wernt confident in it?
they must have been very impressed with what they have seen to have jumped onboard.

Ok so if they have really got this tech working why in gods name would the have kept it to this type of service, every media distribution company on the planet would want to license it, they would be rich from that alone. Is all just to good to be true.

They want a cheque - that's all. They only need to licence their PC games and collect a cheque. They don't care.

I wouldn't!

This is now a tardmagnet, so I'll leave it.

So ea would back any crap?
Publishers would not sign up to this unless they were confident it would work.
If the military can fly remote drones from thousands of miles away using data streams this can be pulled off.
Its going to take military technology to get this working.

@niteninja
Maybe thats why so many wedding parties have been bombed. They thought they were over the hill in the caves, but they didn't factor in the lag, lmao

Seven years of secret development?
Did the department of defence allow onlive access to military computers to build this thing?
Public tech is 10 years behind.

I think you've been reading too much Area 51 pal.

The game companies don't invest anything, they just gave them the right to use their software at the moment, surly because they don't want to be left out whenever those people can come up with the real thing.

The investors and Perlman will make money with stock, patents and selling assets. Nobody will have to run. They will found another company and when they come up with another thing a lot of people will say admiringly they were the ones coming up with server-side gaming.
Stupid small-time investors easily impressed will be the primary losers. But nobody cares about them anyway.

Well ... time will tell, but I do think this article was retarded exactly by the same reasons my poor friend Freeload pointed out.

Time will tell. My bet is that the earth is pear-shaped :)

I personally think this article is on the money, but I think it'll be interesting to see what happens. I spoke to someone involved in the project today at GDC, who said:

1. It'll be a US-only launch [so this is all a bit of a moot topic for most of us!]
2. They have already done deals with ISPs to put servers directly into their facilities, so theyll be close to the users
3. He's used it on a 3mbit connection and it's worked fine
4. They expect to be the 2nd biggest consumer of bandwidth in the world within 12 months (presumably Google is #1?)
5. They have created their own custom blade servers which can run multiple instances of the game (1000 game sessions in a single server rack)
6. They have $100m in funding
7. It takes one week to take a game from a publisher and port it to their server platform
8. They are well aware of the other uses for this technology but gaming is the first interesting app.

Dear lord, some people will believe anything. Even their website screams "VAPOURWARE!"... look at the positions that they are hiring for...

Operations:
Network Engineer

Engineering:
Embedded multimedia engineer
Games SDK engineer
Mechanical/Thermal engineer

Those "open" positions are pretty much the types of people that would tell you why it cannot be done with current technological limits (the-last-mile, heat generation and dissipation dynamics, cluster failover, etc). I wonder what the patents describe, and whether they have the video compression working in anything other than highly controlled environments...

The first 10 minutes of the presentation (all I've watched so far, but have read a lot of the articles floating around about it already, and am an online engineer myself) list multiple technical challenges that simply cannot scale with today's, or even tomorrow's, technology. Virtualizing that many game instances in parallel whilst maintaining a level of interaction that is acceptable is not a linear scale-up, as anyone with parallelization experience would be able to tell you. They may well have gotten real-time compression of a non-linear video output down to 1ms, but that will not hide the problem of input lag to the server, which will be around 50ms at best (3-4 frames at 60fps), game simulation of 17ms (1 frame at 60fps), video compression time of 1ms (which I am assuming to be best case with no other instances vying for contention time), relay to the client (50ms, again at best) and decompression time, which you can BET is not going to be 1ms - let's be generous and assume 10ms... it is, after all, hardware decompression :)

So all in, you're looking at:
50ms + 17ms + 1ms + 50ms + 10ms == 128ms

Best case.

Not too bad. Most online game now rely on that kind of delay in order to provide synchronization between peers. But...

...that's best case. And a single instance. Add virtualization costs, context switching multiple instances of virtualized sessions, maintaining said instances in huge banks of memory (you cannot cache them - fetching them from virtual memory would be a huge performance hit), compression of multiple non-contiguous video streams, transmission of multiple compressed video streams, storage and restoration of "paused" sessions, etc, etc, and you have now got to face and handle a multitude of technical challenges that a lot of very smart (and specialized) people have been working for many years to overcome. 7 years and this relatively small company have got the video compression problem licked? That's the least of their problems, and one that would have solved itself gradually over the years with increased computational capabilities and exponentially decreasing silicon manufacturing costs. The scaling problems are far more challenging and require several technological paradigm shifts before they can be realized, but must be overcome before the system is viable for the mass-market.

The only question now is how many VCs will they fool into parting with their money before the impossibilities of what they are proposing are accepted...

Rauper,

in responce to your points, i offer the following from the perspective of a Senior Windows and VMware Engineer. I specialize in blade servers and virtualization

1. It'll be a US-only launch [so this is all a bit of a moot topic for most of us!]

No arguement from me here

2. They have already done deals with ISPs to put servers directly into their facilities, so theyll be close to the users

Alright, plausible, but i have my doubts, I don't see ATT, Verizon, Comcast, COX, Cablevision being terribly receptive to oodles of bladeservers shoehorned into their datacenters

3. He's used it on a 3mbit connection and it's worked fine

Haven't used the service, can't comment.

4. They expect to be the 2nd biggest consumer of bandwidth in the world within 12 months (presumably Google is #1?)

I severely doubt this, the sheer costs involved with building a server infrastructure that large is insane.

5. They have created their own custom blade servers which can run multiple instances of the game (1000 game sessions in a single server rack)

I find this particular point HIGHLY unlikely. 1000 game sessions in a single server rack? A standard rack is 42U of rackspace.

The most space-efficient blade design i know of its the HP Bladecenter C-class, it fits 16 server blades into a rack space of 10u. Thus, you could fit 64 blades in a single rack, this is NOT counting any SAN, which you would have to have.

The most powerful blade available is the HP ProLiant BL680c G5 server blade, which supports 4x 6-core Xeon e-7450 2.4ghz CPUs with a max of 128gb of ram.

Now, keep in mind, although this system has a total of 24 processing cores with 128 gb of ram, it has NO VIDEO CARDS to speak of.

There is absolutely no way in hell you could fit a single, let alone multiple Nvidia 9xxx or 2xx series or ATI 3xxx or 4xxx series cards into a blade. There isn't physical space for one, there isn't thermal tolerance for another.

Those Intel Xeon E-7450s have a 90watt TDP, so a total of 360 watts of thermal dispersion is needed for all four.

For comparison, the Nvidia GTX 280 has approximately a 300 watt TDP. Yes, THREE HUNDRED watt TDP. For one card. Imagine now the cooling issues with a system with 2 GPUs, or 4 or god forbid 8. Quad-SLI is a strain on the cooling system of a fullsize tower, its impossible in a blade.

Even if i give them benefit of the doubt and say they're using Nvidia's TESLA external GPU concept, that opens another can of worms. Those are horrendously expensive. $1300 per card, which is basically an Nvidia 280gtx with 4gb vram.

The only one i can find for sale is this.

http://www.tycrid.com/?page_id=33
$9k for 4 GPUs which aren't even designed for gaming in the first place! They're designed for CUDA!

The final problem is how are they going to run multiple instances?

If VMware and any other virtualization platform that i know of can't do true hardware accelerated 3d graphics, how did these guys do it?

A startup doing virtualization better than a company(VMware) who does 1.5 billion in sales per year and has a market cap of 20 billion dollars? I don't think so.

I hope they prove me wrong, but i smell snake oil. Lots of it.

6. They have $100m in funding

Barely enough to set up for 20,000 users even going by THEIR highly unlikely estimates of what will run in a single rack.

Going by the above numbers for blades. The blades alone cost $20k apiece, 64 in a rack. That's 1.28 million for just the blades. It'd be another $54000 per blade for the GPUs if you use the Tesla solution. 1 GPU per core. 6 tesla units at $9k apiece for a total of 24 GPUs.

$54,000 x 64 blades is another 3.45 million dollars. So, for a "mere" 1000 users, we're already at 4.8 million dollars. Just for the blades, Tesla setup. We haven't yet talked about Blade Enclosures. Data center rental or owning costs. Power. Networking. SAN. Fibre. BANDWITH. Not to mention the costs incurred installing and supporting this beast with Virtualization experts and server experts.

In hardware alone, you've blown through your 100 million dollars after allowing 20,000 users concurrent. Before any of the other ESSENTIALS that i mentioned above.

This setup isn't implausible, its IMPOSSIBLE.

7. It takes one week to take a game from a publisher and port it to their server platform

Ok, i won't argue.

8. They are well aware of the other uses for this technology but gaming is the first interesting app.

*shrug* alright.