Why Project Flare Might Just End the Console War

An anonymous reader writes "Project Flare, the new server side gaming technology from Square Enix, turned heads when it was announced last week. The first tech demos do little more than show the vast number of calculations it can handle with hundreds of boxes tumbling down in Deus Ex, but the potential is there to do much more than just picture-in-picture feeds in MMOs. As a new article points out, what's most interesting is the potential to use the technology for games that use more than one system — OnLive may have used this tech before, but only to play games you can buy on discs in the shops anyway, but the future is in games that need the equivalent of dozens of PS4s or Xbox Ones to power them. Ubisoft has already partnered with Square on the project."

Comments

IMO, it is not going to work

[faragon]

Why to spend power in datacenters when people can use it at home? Other than vendor-lock, is non-sense. Another thing is how scalabe the thing is, etc.

Re:IMO, it is not going to work

[Anonymous Coward]

I'd imagine there'd be some scalability advantages for specific use-cases (re-use of assets, models, animations and the game world across multiple instances of the game), so MMOs could generally benefit from this approach because many users share the same content at the same time, while it would be close to useless for single-player games where basically every player has different content on-screen and in-game than every other player.

Re:IMO, it is not going to work

[drinkypoo]

Why to spend power in datacenters when people can use it at home? Other than vendor-lock, is non-sense. Another thing is how scalabe the thing is, etc.

One example is to provide functionality which cannot be provided by the console machines themselves. For example, games for the Xbox 180 are going to have the option to use Azure to run game servers. One of the major frustrations of console gaming today is that one of the game consoles has to play server.

From the summary, though, the idea is to provide games more powerful than what your console can actually run. With a large enough playerbase it might actually be feasible. It costs a lot of CPU to perform a lot of physics calculations, but if you only have to perform them once for a whole bunch of players' updates because they're all looking at the same thing, then you're going to save some cycles there.

Re:IMO, it is not going to work

[mwvdlee]

It's basically a way to keep the price of consoles at a point where people will still buy them, while being able to offer the level of processing power that would be too expensive. I'm sure the monthly subscriptions will cover cost of datacenters and power consumption.
I don't know if it's vendor lock-in, but atleast this is a way to offer paying customers a better experience than pirates instead of the other way around with current DRM.

Re:IMO, it is not going to work

[abies]

Are reallly contemporary FPS games calculating physics separately on each machine? Given that physic calcs tend to be non exact and a bit of chaos theory you could end up in really different worlds very soon.
If we are just talking about moving 'server' from one of player consoles to the dedicated datacenter (like most 'normal' games do it), then it hardly looks to be exciting?

Re:IMO, it is not going to work

[drinkypoo]

Are reallly contemporary FPS games calculating physics separately on each machine?

The answer is sort of and also sometimes. I can't actually speak to FPS specifically, but in GTAV the game clearly does independent physics calculations when it thinks you're far away from other players. You can tell because things get squirrely when you catch up to a lagged player, or when a lagged player catches up to you. The game doesn't bother to synchronize events which it decides can't affect other players.

Re:IMO, it is not going to work

[Moryath]

Works great till you realize the USA is currently worse than a third-world country in terms of broadband penetration and up/down speeds...

Re:IMO, it is not going to work

[WaffleMonster]

It's basically a way to keep the price of consoles at a point where people will still buy them, while being able to offer the level of processing power that would be too expensive.

Current consoles are well beyond the point of diminishing returns with regards to graphics power while cost of replicating existing capabilities keep getting cheaper year after year.

I don't know if it's vendor lock-in, but atleast this is a way to offer paying customers a better experience than pirates instead of the other way around with current DRM.

On what planet does high latency translate into a better experience?

Re:IMO, it is not going to work

[epyT-R]

The physics stuff is done on the server. The non interactive animations are done clientside. However, the clients have code to anticipate the server's next state, 'correcting' itself when the prediction fails to match. This results in smoother animation and reduced apparent lag.

Re:IMO, it is not going to work

[SuricouRaven]

Write a mod for ut2k4 and you'll soon start seeing how it works. The local game runs a simulation, but subject to correction by the server. Extrapolating events until the packets catch up. Really latency-sensitive things like sniping are handled locally. This can lead to some very strange things happening at times:

1. Run past a window.
2. Clear the window.
3. Your movement is passed to the server, and then to another player.
4. Other player snipes you.
5. Snipe victory is reported back to the server, then to you.
6. Half a second after passing the window, you drop dead. Headshot. Even though from your perspective, you were in a place you should have been out of sight. Serves you right for running past a sniper-visible window.

Generally the game is good enough that almost all of this is transparent though. Only the exceptionally observent notice it. Still rather strange to code for, as everything you write is actually being executed three times in parallel (On the server, on the client, and on everyone else's client), and you need to make sure that all three executions eventually give the same result, even if not at quite the same moment.

Re:IMO, it is not going to work

[WaffleMonster]

Its very simple - power savings, and cheaper thin consoles for end users.

What power savings? Power is being consumed somewhere else where as a customer YOU are paying for that too. Lets not forget about additional power requirements required to push insane number of real-time bits for trivial reasons over the Internet.

This would not work too well for multiplayer because of the latency between user-server-user, but would be great for single player.

Since everyone would experience input latency and there is no network latency for the multi-player link latency would be about the same persistent problem whether it were single or multiplayer game. The only lag assuming lack of operator incompetence would be in the form of input delay with very limited opportunities to compensate with prediction algorithms. Nobody who plays on anything approaching a competitive basis would touch this thing.

Re:IMO, it is not going to work

[The MAZZTer]

AFAIK with server<-&grclient you have all clients doing physics calculations BUT they are only used for prediction; the server updates the final position and orientation using its own physical calculations (but since the clients are predicting it can do it with a low-resolution in terms of time).

Re:IMO, it is not going to work

[rsilvergun]

Don't forget the ISPs have more or less killed net neutrality and will soon be charging $$$ for bandwidth on the lines run with your tax dollars.

Re:IMO, it is not going to work

[aaaaaaargh!]

It's basically a way to keep the price of consoles at a point where people will still lease them, while being able to offer the level of processing power that would be too expensive.

Fixed that for you. This has nothing to do with buying anything, it's a temporary lease and the servers will be shut off sooner than later.

Re:IMO, it is not going to work

[clockwise_music]

Works great till you realize the USA is currently worse than a third-world country in terms of broadband penetration and up/down speeds...

Sorry, have you been to any third world countries? In many you're lucky if you can get dial-up speeds, yet alone a constant connection.

Re:IMO, it is not going to work

[khchung]

Why to spend power in datacenters when people can use it at home? Other than vendor-lock, is non-sense. Another thing is how scalabe the thing is, etc.

How about 100% cheat prevention? When all the computing is done centrally, how could you possibly cheat in the game anymore?

Plus, it totally eliminated the lag factor in FPS, as only the central server do the processing and rendering. Rubberbanding and blinking/shifting enemies will be eliminated.

The only lag now comes between your end to the server, which, while non-zero, is at least consistent from game to game.

With only 1 copy of the world, then the number of players will only be limited by the number of CPU doing rendering from the POV of each player, and that is probably easier to scale as the rendering process is read-only. So you can have MASSIVE number of players in the same game, imagine hundreds of player all in the same battlefield, and that limit can be increased by a server upgrade instead of waiting 5 years for another console generation.

Aren't these reasons good ones?

Re:IMO, it is not going to work

[wwalker]

Sniper rifle should be the *least* latency-sensitive weapon. In real life, no sniper can hit a running target at any reasonable distance (unless they are running directly towards, or away). More so if the target is passing by a window and is only visible for a fraction of a second, which makes any sort of leading practically impossible.

Re:IMO, it is not going to work

[Smauler]

Why spend $500 on PS4 at home when you can get a cheap client for under $100 and a $10/month subscription to such a cloud service

Because if those were the two options when I bought my PS3, the first would have cost me $500 by now, and the second would have cost me $940 by now.

Re:IMO, it is not going to work

[Dodgy G33za]

Unless they have an aimbot

Re:IMO, it is not going to work

[denmarkw00t]

Sure, but you also get a lot of good hits out of knowing where doorways and passages behind walls are: I've sniped a number of people behind walls because I knew, at spawn, there would be a high chance of *somebody* being behind it - given the narrow passageway and the odd curiosity of players going "should I step out there...," the chances for a hit on that wall are insanely high. If you train on a frequented path in an FPS, other than Battlefield which is HUGE, and shoot one round every 5 seconds after spawn, you'll probably hit something. Probably. IRL? You generally don't "know the map" IRL, so the odds drop a lot.

Re:IMO, it is not going to work

[Anonymous Coward]

>How about 100% cheat prevention? When all the computing is done centrally, how could you possibly cheat in the game anymore?
Most games have all, or nearly all, of the processing happening on the server as it is. Doesn't stop cheaters. Sure, they can't exactly just memory edit the amount of gold/points/health/whatever they have anymore, but there are an infinite number of other ways to cheat. Think about botters; that doesn't rely on client-side processing. Aimbots in FPS games do not need to rely on client-side processing (from the game, anyways) either; they will detect enemies and simulate mouse movement to auto-aim.

>Plus, it totally eliminated the lag factor in FPS, as only the central server do the processing and rendering. Rubberbanding and blinking/shifting enemies will be eliminated.
No it doesn't. In fact, it's going to make it worse. It might *look* different, but the actual effect will be more detrimental to your gameplay. If you have issues with lag with a very, very low amount of information being transferred (ie. position updates), then what the fuck makes you think upping it to uncompressed 1080p video streaming is going to improve it? Instead, it'll be like trying to watch a Youtube video that is constantly trying to buffer. Even if, and I stress "if," it were to stream halfway decently for you, you are still going to be feeling the effects of the lag. Everything will feel sluggish and the controls will always seem to trigger actions that happen much later than when you pressed the button.

>With only 1 copy of the world, then the number of players will only be limited by the number of CPU doing rendering from the POV of each player, and that is probably easier to scale as the rendering process is read-only. So you can have MASSIVE number of players in the same game, imagine hundreds of player all in the same battlefield, and that limit can be increased by a server upgrade instead of waiting 5 years for another console generation.
This can already be done and it does not need that the video output be rendered on the server which requires even more massive servers.

Re: IMO, it is not going to work

[loufoque]

South Africa isn't a third world country.

Re:IMO, it is not going to work

[khchung]

>How about 100% cheat prevention? When all the computing is done centrally, how could you possibly cheat in the game anymore?
Most games have all, or nearly all, of the processing happening on the server as it is. Doesn't stop cheaters. Sure, they can't exactly just memory edit the amount of gold/points/health/whatever they have anymore, but there are an infinite number of other ways to cheat. Think about botters; that doesn't rely on client-side processing. Aimbots in FPS games do not need to rely on client-side processing (from the game, anyways) either; they will detect enemies and simulate mouse movement to auto-aim.

Think again. When all processing is done in the server and only the screen is sent to the client, wall hack becomes impossible. Aimbot? Your aimbot better be able to identify an opponent's head from the displayed graphics. Not the say that's as difficult as in doing so in real video footage, but it raised the bar so high, that anyone able to pull that off would be quite an expert in pattern and facial recognition, and not just a matter of finding the coordinate for the opponent's location in the data stream.

>Plus, it totally eliminated the lag factor in FPS, as only the central server do the processing and rendering. Rubberbanding and blinking/shifting enemies will be eliminated.
No it doesn't. In fact, it's going to make it worse. It might *look* different, but the actual effect will be more detrimental to your gameplay. If you have issues with lag with a very, very low amount of information being transferred (ie. position updates), then what the fuck makes you think upping it to uncompressed 1080p video streaming is going to improve it? Instead, it'll be like trying to watch a Youtube video that is constantly trying to buffer. Even if, and I stress "if," it were to stream halfway decently for you, you are still going to be feeling the effects of the lag. Everything will feel sluggish and the controls will always seem to trigger actions that happen much later than when you pressed the button.

You assumed current technology for streaming video is the only solution. But given that the server had all the original 3D information to generate the video frame to begin with, many other techniques becomes available. E.g. Regardless of resolution, the server can just send the polygons need to be drawn to the client side for a frame, and then the changes for the next frame, etc. Which would reduce the data size tremendously. Instead of 1920x1080x24 bits = ~6MB a frame, the position of 10 thousand vertices will only take ~40KB assuming 4 bytes each. Textures can be stored locally, then the 100KB from the server could give enough information to the client to render the image.

As for the control latency, it is there in current FPS anyway, it's just a matter of how well the game masks it. Actually, it is worse in the current FPS, as my movement really have to make it, not only the first hop to the server, but make one more hop to my opponent's machine for me to avoid getting hit from his POV.

If you played FPS then you will know about turning a corner to hide, only 2 seconds later to have the game tell you that you actually got hit and killed before you turned that corner, because your opponent lagged so badly he didn't get the data of your movement until after he already shot you.

>With only 1 copy of the world, then the number of players will only be limited by the number of CPU doing rendering from the POV of each player, and that is probably easier to scale as the rendering process is read-only. So you can have MASSIVE number of players in the same game, imagine hundreds of player all in the same battlefield, and that limit can be increased by a server upgrade instead of waiting 5 years for another console generation.
This can already be done and it does not need that the video output be rendered on the server which requires even more massive servers.

This can already be done?? Which FPS running on console give you games with 100+ players in it?

Re:IMO, it is not going to work

[Anubis IV]

I've heard your line repeated before around here, and it's time it was finally put to rest, since it grossly overstates the problem and muddles any attempts at rational discourse. Going by your own metrics for judging success, how many actual third-world countries do you see above the US in these lists?

Table of broadband penetration rates by country (sort the table by %)
Mean upload speed rankings by country
Mean download speed rankings by country

By my count, there are none that are ahead across those rankings.

Unless you're sticking to the original definition of "third-world country" (i.e. one that was neither associated with NATO and its allies nor the Soviet bloc and its allies), rather than the modern meaning of the term, I don't think it's fair to generalize in such a way as you have. Admittedly, there are a few developing nations ahead of the US on some of those lists, but they appear to have done so by sacrificing one metric for the benefit of the others (e.g. Senegal is ranked #10 globally for upload speed and #14 for download speed, putting it ahead of nearly all of the developed Western world, but its low single-digit penetration rate ranks in the 100s and indicates that only the elite of society are benefitting from those speeds). At least in my glancing over the lists, no developing nations stood out to me as being ahead in terms of both speed and penetration rate, though I'll admit I may have missed them.

Could the US use improvements in this area? Absolutely, and I want to be clear on that point. But what I consistently notice is that people, particularly those who have either never left the US to experience other parts of the world, or else those from smaller countries who have never traveled across a single country as large as the US, have no appreciation for just how difficult of a problem the US faces as compared to many other developed nations, simply due to its massive size. When you compare the US against countries like Australia, Brazil, Canada, China, and Russia, which face similar (though clearly not identical) issues of scale, you find that the US is generally either comparable or ahead for most of those metrics (insert the requisite asterisk here with a note about how Macau, Hong Kong, etc. are not indicative of mainland China as a whole).

Again, improvements are still needed in these areas, but it's hard to have a rational discussion when you have one side overstating the situation to such an extreme. The fact is, the US' system is roughly comparable or ahead of countries that face comparable issues of scale. Which isn't to say that we should pat each other on the back, congratulate ourselves for a job well done, and think that everything's rosy. Rather, it's merely to say that things are probably about where they should be expected to be and that they're actually pretty decent when compared to most of the world, but that we clearly need to keep working at making it better, since there's lots of room for improvement.

Console DRM?

[Hsien-Ko]

Imagine your game not starting because the 'physics' servers are down or you can't connect to them.....

Re:Console DRM?

[Baloroth]

In theory, a properly designed system would have a local physics engine that takes over when the remote engine is unavailable, albeit at reduced fidelity.

Of course, that's not how it will end up working, especially with Ubisoft involved. The real goal here is more control disguised as improvements.

Re:Console DRM?

[fuzzyfuzzyfungus]

Not much of a problem when you're dealing with an MMO-type game that requires you to be online anyway. Not more of a problem than the game-server being down, anyway.

The bigger problem will be getting game developers to take on the challenge (and risk, if it means that they can only sell their game to people who are paying extra for Xbox Live Platinum Cloud Physics Foundation Edition Home Ultimate) of creating a type of game that, like the MMO, is only possible in the context of this remote capability, rather than having the remote capability be an obvious cash grab/lockin attempt. Think of how much people liked the 'online' features of SimCity's recent reboot... It was obvious to all that it's value was close to zero to them, and its costs were substantially greater, and lo, great was their wrath.

In the hypothetical case of the game that could only be realized with 'Flare', people will grumble; but put up with it. If it's transparent to all that you need a monthly subscription and a fast broadband connection just so that the "generic glass fragments flying outward" effect that the art guy spent all day tweaking to look exactly like it does in the movies can be removed and replaced by a realtime finite element analysis of the window and computational fluid dynamics simulations of the trajectories taken by each glass fragment (totally different on the maps with higher atmospheric pressure, worth it!) in real time? That's a tech demo, not a feature.

Re:Console DRM?

[bondsbw]

Not everything that requires the Internet is DRM.

And frankly, even if DRM was the primary intent, it's one of the less annoying DRMs since 1) online multiplayer, by definition, requires the Internet, and 2) economics:

Let's take the Xbox One. Microsoft claims to be adding 3 more Xbox One units for each one sold in the cloud, but really that's based on an average or maximum use situation. In reality they will have 300,000 cloud machines powering the number of users that might be gaming at any one time. Compare that to the cost of quadrupling the power of the Xbox One in the 80M consumer machines that will be sold over the next several years. And that doesn't even touch on the cost efficiency of heavily optimized power and cooling conditions in a server farm.

Of course, the cloud will be working on the latency-agnostic calculations, things like AI and up-front scene lighting calculations and of course simply serving multiplayer games and anything that needs to be calculated the same for all players. But that frees up your local system resources for those things that can't be offloaded to the cloud.

I'll take a bit of DRM for that.

Re:Console DRM?

[Anonymous Coward]

Imagine your house is on fire. It would be hard to play games in that situation as well.

I see they hired the mafia to handle difficult sales.

Right...

[Pinhedd]

OnLive was such a bastion of success wasn't it?

Re:Right...

[Zero__Kelvin]

Yes. I thought everybody knew that OnLive was an angular structure projecting outward from the curtain wall of an artillery fortification of success.

Onlive - take two

[WaffleMonster]

This reminds me of a certain coffee stand I drive by each day on my way to work.

Every few months it closes down, sold to a new owner who improves it and re-opens only to close down a few months later.

In the last few years I can count on one hand times drive thru was something other than completely empty.

Sometimes people just can't take a hint.

Re:Onlive - take two

[Zero__Kelvin]

That's weird. You would like that things would have changed with all the advances in coffee since 2005!

Great, Square ...

[Anonymous Coward]

The company that last released a good game 16 years ago. I can barely contain my excitement.

No thanks

[epyT-R]

There's a reason I've cut cable tv from my life. Being remote controlled and the only game in town, it's become overpriced, ad-laden, and content thin. If that's where gaming is going I will have to cut that too. The prospect of overpaying to 'stream' a laggy, ad-filled game experience with overly-constrained lossy-compressed AV doesn't sound inviting either. I LIKE the idea of having power under the hood locally, so to speak, just like I want server binaries for games to run my own servers and mod tools to make my own mods/maps. This way the game stays alive as long as there are interested players and doesn't die the moment it stops making money for its creators. To top it off, the current 'cloud' model for a lot of software now charges the 'owner-controlled boxed software' prices of the 90s for what amounts to a rent-a-go arcade level of service. What a rip-off.

The more computing looks like ibm's wet dream of 'service', the less interesting and more oppressive it gets. No thanks.

Obvious: latency

[De Lemming]

Even with modern broadband, latency is still an issue for these kinds of applications. In the article are some examples of currently used server side gaming enhancements, like "Forza 5 will even use cloud computing to monitor the way you drive, and alter virtual drivers’ AI (artificial intelligence) accordingly." That has no need for low latency. But if you want the environment to immediately react to players actions, there need to be low latency. And you can't remove the distance (and related network infrastructure) between the player and the data center.

Re:Obvious: latency

[Coppit]

Full disclosure: I work for NVIDIA on cloud gaming.

I was as skeptical as you about the latency. In this interview. Phil Eisler talks about 200ms of XBox + TV latency that people live with every day. (See page 2) If that's our target, then that's pretty doable, since with strategically located data centers you can get the network latency down to 20-30 ms.

In the work we're doing, we're actually focusing more on hitching in the game than latency, since the latency isn't as big a deal if you're say in the Bay Area where one of our test clouds is. Heck, I played Trine2 from the east coast and it was very playable. I wouldn't play BF4 across the country, of course. :)

There are plenty of other risks with the idea, but I wouldn't put latency at the top of the list.

Re:Obvious: latency

[viperidaenz]

20-30ms.... on a home broadband connection, anywhere in the world?
Light only travels 185miles per millisecond.

Microsoft just debunked this for FPS games

[PhrostyMcByte]

Some people kept saying "It's not that bad right now, it'll work eventually!", but Microsoft just (accidentally) tested OnLive's idea for low-latency games by introducing some small input lag into Windows 8.1. Guess what? FPS gamers noticed.

Other game types which don't need super low latency, I'm sure, will eventually get here if only because game companies are still annoyingly DRM-focused and this will make piracy impossible.

Turn-based

[tepples]

On what planet does high latency translate into a better experience?

On fictional planets in essentially turn-based games, like what Square Enix has been putting out since Dragon Quest/Warrior and Final Fantasy in the NES days. The latency doesn't have to be any better than, for example, the ATB recharge time in FFVII.

Engineer: We need to optimize the physics engine.

[technosaurus]

Management: But you said it was working.
Engineer: It is ... at 3 FPS on a standard PC.
Management: Perfect, we can sell it by the hour until then.
Engineer: You're kidding right?
Management: No, seriously, BTW your new project starts tomorrow.
Engineer: but....

Size is an irrelevant argument for many of us.

[zooblethorpe]

Could the US use improvements in this area? Absolutely, and I want to be clear on that point. But what I consistently notice is that people, particularly those who have either never left the US to experience other parts of the world, or else those from smaller countries who have never traveled across a single country as large as the US, have no appreciation for just how difficult of a problem the US faces as compared to many other developed nations, simply due to its massive size.

It would be one thing if the argument were solely that people living in Story, Indiana or Nothing, Arizona couldn't get broadband speeds.

While that is an issue, it's not what causes much of the complaint about the state of internet services in the US.

I live in Seattle, within the city limits. I can't get better than 4.5mbps down on a good day, and certainly not in the evening when everyone's watching Netflix, short of ponying up for a business line to the tune of substantially more expense. Five years ago, I lived within spitting distance of the Google campus, and couldn't get better than 1.5mbps down. These are major cities, densely populated, with all the infrastructure right there.

By comparison, when I left Japan in 2005, my bare-bones residential service -- the cheapest, slowest, least-of-everything-and-still-be-online package gave me 18 mbps for around $30 a month. And it was scheduled for an upgrade, at no cost to the subscriber, to 24 mbps two months later.

The key difference? Competition. For all the malarkey about free markets and rainbows, the US market sucks for internet services. A handful of companies have effective geographic monopolies (or at least very small cartels), giving them leverage to jack prices and keep services at the bare minimum. In Japan, the kind of lockdowns that are the status quo in the US aren't possible due to an effective regulatory regime, necessitating that companies actually compete for consumers' business on the basis of service and price. The differences are amazing. Or depressing, depending on where you live.