Remote game services (OnLive, Gaikai, etc.)

A colourful game moving at fast speeds like Burnout Paradise (another OnLive demo) is the complete opposite - a total nightmare to encode. 5mbps at 30fps will still look pretty ropey.
What's ropey? On a scale of 1 to 5 where 1 is default YouTube and 5 is DVD, where would you expect results to lie for the more demanding games .
 
Doesn't matter if it's MPEG2, MJPEG or h264, there are some basic truisms about compression. Firstly, if the screen is mostly static or has slow motion, you can get away with absurdly low bandwidth.

It's quite a bit more complicated than what you're thinking, they haven't simple adapted an existing codec, they're written an entirely new codec, and redefined the logical constraints, to deal strictly with the issue of lag/response time. Specifically, they've loosened constraints on errors and failures, and added what sounds like very robust error correction, and they've thrown out the standard GOP method of encoding.

These are the guys who developed Quicktime back in the day, according to him he was the one who actually pushed Apple to develop a video codec at all, so he seems to brign some credibility.

Anyways, what he says is the basically threw out all the restrictions on the codec with regard to looking good when in a still frame, this codec is only dsigned to look in motion, i.e. you will only ever see this feed once: as you're playing. It's specifically designed to tolerate and expect errors, and has a whole bunch of code built in to hide or correct errors on the fly, and an active feedback loop to the server itself.

Then what they do is encode 2 streams of every playsession, the "Live" stream is lower quality, meant to be used only in action, but not for actual playback. It's also based on feedback from the client, and adapted on the fly. The 2nd stream is a highquality static stream which has no feedback loop, and would be used for playback, brag clips, or a multi-broadcast to many users, but not live play.

they had custom chips fabbed, and their sole purpose is to run their codec, this lets them get the per user hardware cost down to something like $30/person (initially the costs was around $10000/person using existing hardware), and literally encode a frame in 1ms.
Sounds pretty cool...also shows alot of faith in this codec, that they would go as far as to fab chips for it.

Again, I'm very skeptical as well, I don't think the quality is going to cut it, it just doesn't seem possible. But it's worth noting there's a whole lot more going on here than you are giving them credit for. They're not just compressing a bunch of h264 video and sending it out over the web, it sounds like they've done a lot of the nuts and bolts engineering that would be needed to really make something like this feasible. It's lightyears beyond MS's 'smooth streaming', though I'm sure they implement something along those lines as well (seems like it would be almost essential in a real world gaming scenario).
 
A colourful game moving at fast speeds like Burnout Paradise (another OnLive demo) is the complete opposite - a total nightmare to encode. 5mbps at 30fps will still look pretty ropey. Crysis is a nightmare to compress too.

Thing is, you don't watch a game like you do a movie.

If there's a ton of motion, you're most likely going really fast, which means you most likely have some pretty strong tunnel vision going on, and won't notice artifacts as much.
If you're not going fast, and everything is slow, then the quality should be nice and high, and bitrate fairly low.

There would be many exceptions to this, but the general rule would probably apply in like 90% of game situations, so I could see how this could work from a human perception point of view.
 
I'm really not getting how this is going to be financially feasible.

They are going to have to have an ultra high end PC with what appears to be a state-of-the-art cutting edge never before seen or rumored to be in existence PCIE encoder card for every single person wanting to play a game. And not only that, but at every single major ISP serving every single metropolitan area the service will be available.

Or do we really think that a cheap PC is going to be able to not only play a game AND encode the video stream at the same time? And if it turns out of to be 2 PC's required to do that, that's going to jack up the cost even more.

Added to that, pubs will probably want some form of license structure that gives them a return on each individual that plays their game.

And that they have experience as the developers of Quicktime doesn't give me much hope on that end either, as Quicktime was never all that good.

I'm having a hard time seeing this as anything but a venture capital money grab.

Regards,
SB
 
The negativity around this just boggles my mind. Why do so many people want it to fail? People I guess just want to tear things down.

BTW, I'm surprised none of you have even discussed or even mentioned the patents that OnLive & Stephen Perlman have received or applied for. The patents started over 7 years ago & there's at least 20 in OnLive's name & way more in just his name. Most of you have discussed Cell's patent to death before it released. Why not do that with OnLive's? I'd prefer that over just saying it's not going to work.

I may be a layman, but what I saw in that latest video has me excited. It definitely doesn't look like vaporware to me.

Tommy McClain
 
The negativity around this just boggles my mind. Why do so many people want it to fail? People I guess just want to tear things down.
Don't confuse lack of confidence with a wish for failure. The negativity you're reading is only reagrds expectations of the service. That actual value of the service is a different topic, and if OnLive can offer effective streamed gaming at an affordable package, it could well be a very good thing. However, this being B3D, we're more interested in how things are done than whether they're worth doing! :p
 
I don't think the negativity is unwarranted until OnLive proves they can deliver everything they claim. Phantom ring a bell?

It is pretty simple: OnLive hasn't shown they can deliver an excellent 720p60 feed of a high end game with a lot of detail and fast movements. Right now OnLive is making promises.

Promises that fly in the face of well established encoding history. Maybe they have solved 1ms encoder, only that it hasn't been done well before.

Promises of lag free gameplay. Adding another 80ms onto gameplay is traditionally a factor. Add in fluctuations and you could have significant gameplay issues. Maybe they can keep latency down and rock solid.

There are a myriad of other issues. What of bandwidth capping by broadband providers? How much does the service cost? Will you own your games? What happens to your "owned" games if they go out of business? What happens on high traffic days (e.g. weekends, holidays, big game release) when a ton of people are on at the same time?

For all we know OnLive could be an expensive service with no ownership and low overheads where you don't always end up with "cutting edge" graphics (instead running on 2 year old machines) and on top of that you get a glitchy image, significant latency with unpredictable fluctuations, no Mods and less control over online play.

Or it could change the market top to bottom.

But right now they don't even have an end product working in realworld situations for us to even give a fair assessment.

Just another start up looking for a piece of the pie and showing off very limited, and controlled, demos and making big promises. I am sure we hope they have the best product ever at the best prices ever. But lets not hold our breath until they start delivering product. It is all about the games afterall.

And a side note: The lack of concrete information about the role of PUBLISHERS and ISPS (think "rights" here) and how this will alter the consumer experience is not something to lightly overlook.
 
Why not do that with OnLive's?
Is there a single innovative algorithm in any of them? (I know you supposedly can't patent algorithms, and yet you can, but system and method patents which don't patent algorithms are not worth reading at all ... so lets not play that disingenuous little game.)

I think engineering wise what they want can be done, and none of the work necessary should be patentable. Although it will require customers with broadband speeds closer to 5 Mb/s than 1.5 (which at over here means another extra 10$ a month over basic subscriptions). Despite PC hardware being cheap it will require very large up front investments, as well as requiring source code access for the games making publishing deals extremely important.

Onlive and similar services have a window of opportunity, and that window is now ... console hardware is weak and expensive, PC hardware is fast and cheap. Console quality rendering, which is all it's subscribers will expect, requires relatively little PC hardware. The moment consoles get close to the next gen their window will be closed again for a while.
 
Last edited by a moderator:
Have online multiplayer possibilities been discussed yet? Something you get with both consoles and a PC of your own is the easy ability and a more or less robust online system, I wonder how far Onlive can get without it. It's almost certainly a big complication, considering how much lag there already is...

And I know that not everyone plays multiplayer ;) being one of such people, but it is still a major feature...
 
They are going to have to have an ultra high end PC with what appears to be a state-of-the-art cutting edge never before seen or rumored to be in existence PCIE encoder card for every single person wanting to play a game. And not only that, but at every single major ISP serving every single metropolitan area the service will be available.

Or do we really think that a cheap PC is going to be able to not only play a game AND encode the video stream at the same time? And if it turns out of to be 2 PC's required to do that, that's going to jack up the cost even more.

Added to that, pubs will probably want some form of license structure that gives them a return on each individual that plays their game.

Most of these questions were answered in the presentation.

The PC's are leased, first of all, not purchased, which is the only way this is feasable. Every 6 months the pc's are upgraded. In addition, many older games do not require a dedicated PC all to themselves, many can be run simply using the cpu, so box's/resources can be shared among users.

The encoding is done on a 20cent chip, dedicated hardware which shouldn't take much resources from the pc's cpu. Sure, quicktime sucks, but that fact shows that these guys have some chops, and just might have the technical knowledge to actually write a new codec that can do what they claim.

The licensing structure is basically set by the publishers, onLive has a base subscription, and takes a cut of any fees, but the rest is left open to the publishers to decide.

Acert said:
I don't think the negativity is unwarranted

Not unwarrented, but definately mis-informed in alot of cases.
 
Most of these questions were answered in the presentation.

The PC's are leased, first of all, not purchased, which is the only way this is feasable. Every 6 months the pc's are upgraded. In addition, many older games do not require a dedicated PC all to themselves, many can be run simply using the cpu, so box's/resources can be shared among users.

The encoding is done on a 20cent chip, dedicated hardware which shouldn't take much resources from the pc's cpu. Sure, quicktime sucks, but that fact shows that these guys have some chops, and just might have the technical knowledge to actually write a new codec that can do what they claim.

The licensing structure is basically set by the publishers, onLive has a base subscription, and takes a cut of any fees, but the rest is left open to the publishers to decide.

So far everything you have described is similar to all the hubbub, marketing and venture capital grabbing that Phantom Entertainment (previously Infinium Labs) did for The Phantom console.

I have yet to see anything that gives me more hope that this is more of an actual product than the Phantom.

A 20 cent chip never heard of before performing orders of magnitude better than 5000 USD encoding systems? They could make a mint just on that alone, yet they're going to sink it into a "cheap" console?

Leasing of PCs for 6 months at a time? Who's going to do the leasing when hardware devalues so much in 6 months time, making it a money losing proposition for anyone leasing the hardware.

You also can't go with all cheap hardware. What if all your customers decide they REALLY want to play the latest and greatest graphics powerhouse game? You'll still need a high end PC + high end graphics card for every single sytem that is meant to serve a customer just to cover that eventuality.

So far, I'm seeing a lot of "it's too good to be true."

Would I love it if it all comes out as promised? Damned straight I would. Just like I would have loved it if The Phantom had delivered as promised.

Regards,
SB
 
The encoding is done on a 20cent chip
Unless they made their own ASIC it's not going to be 20 cent (and if they did it's going to take a good long time to earn back the non recurring costs before you can honestly say it's 20 cents). FPGAs are incredibly ovepriced for their capabilities, DSPs likewise ... really it's hard to beat good old x86 for encoders in initial cost (power consumption sucks though).
 
However, this being B3D, we're more interested in how things are done than whether they're worth doing! :p

Yet none of you have looked at the patents. And some of you have blindly disregarded the facts presented in the recent video. bleh

Tommy McClain
 
Is there a single innovative algorithm in any of them? (I know you supposedly can't patent algorithms, and yet you can, but system and method patents which don't patent algorithms are not worth reading at all ... so lets not play that disingenuous little game.)

I don't know, have you even attempted to look at them? I know I have, but I'm not technically competent to understand them. If you're capable of understanding them, whats stopping you from reading them?

I did see that in one patent they're not explicitly patenting just a simple video codec, but instead a system and method for "compressing streaming interactive video" - PA #20090119736. They at least have 11 patent applications dealing with streaming interactive video. Somewhere in there I think some of the technical questions can be answered.

Tommy McClain
 
Unless they made their own ASIC it's not going to be 20 cent (and if they did it's going to take a good long time to earn back the non recurring costs before you can honestly say it's 20 cents). FPGAs are incredibly ovepriced for their capabilities, DSPs likewise ... really it's hard to beat good old x86 for encoders in initial cost (power consumption sucks though).

In the video it's a $10 chip running at 2w, see video at 40:40. Said they can put two of them on a board & cost about $25-$30 per user. Two are necessary for both the live stream & the media stream, but there can be multiple compressor boards per system. Before they developed the algorithm to run on TWO dual quad-core Xenon machines(one for each stream) costing $10,000 per user.

Tommy McClain
 
Even that is a very strange price ... if they used an off the shelf SOC meant for mobile use with build in x264 I'd believe it, but something programmable with enough grunt to handle a 720P60 stream at 10$? Also they are supposedly only using those chips at 1/16th capacity (1 msec encoding time, so not much to do for the other 15 msec till the next frame arrives). I'm just not buying it.
 
Watch the video will you please? They're not doing traditional video encoding(groups of pictures).

Tommy McClain
 
Being untraditional is no reason to not obey the laws of physics ... if it adds 1 msec of latency then it has to encode the entire frame in 1 msec (and wait for the next one for 15 msec).

GOPs are not really relevant any way, they didn't determine the latency in the past and they don't determine the latency now ... only b-frames put a lower limit on encoding latency. You can do low latency encoding with just about every video coding standard when using only I&P frames.
 
I watched the video last week, but I'm not fully convinced it will work in the extreme case (beta test may be fine, but peak hours may be expensive/overloaded). The explanation he gave is not rigorous enough to dispell doubters. It may be possible to target specific regions/segments though.
 
Back
Top