“The Project FLARE Difference”
Before I begin spewing out all the crazy, theoretical in-game applications that something like FLARE could potentially allow for, it’s probably wise to take a look at how it works, and what Square is claiming it will be capable of doing.
In a PDF released on FLARE’s website, there are a few early slides that nicely sum up what the project is about, and why it’s unique. One bold-face line calls FLARE “A technological breakthrough [that] replaces the console with a supercomputer.” Skip ahead one slide, and we’re shown this.
As mentioned, FLARE separates itself from other “cloud” services right off the bat, delegating the term “streaming games” to the OnLives of the world instead. And if you think about it, it makes sense. When I watch a YouTube video, do I call it “cloud video?” What about when I play songs from Spotify – is that “cloud music?” Rarely is the term “cloud” applied to streaming, so Square’s efforts to define what is and isn’t cloud gaming right off the bat, to me, seem justified. You may feel differently.
Either way, FLARE’s underlying technology is undeniably something new, and the the slides that follow actually do a pretty nice job of breaking things down in layman’s terms. Unlike streaming games, FLARE has the ability to harness the power of however many CPUs and GPUs it has access to, and fit them together like puzzle pieces in the most efficient way possible. In other words, it isn’t limited by the constraints of each physical box – each and every CPU and GPU in the FLARE system will essentially be daisy chained together, and be able to be accessed at will.
That’s all well and good, but is it really that much of an advantage? Though FLARE has a sizeable but not mind-blowing edge in the above example (the ability to add CPUs that operate without GPUs and vice versa to the “super computer” is definitely an advantage), the real gains come when even more raw power is needed. Unlike current game-streaming platforms, the boundaries of the “logical unit” in FLARE aren’t boundaries at all – they can, from what I understand, be endlessly expanded. The result is that CPUs and GPUs can be added at will as necessary (though you still need at least one CPU in a physical box full of GPUs), and they’ll all essentially be tacked onto the existing logical unit. The ultimate result is that unlike with streaming games, FLARE can perfectly match CPU and GPU resources, increasing efficiency exponentially as a result.
Beyond that (which is, admittedly, the main technological advantage), FLARE does have one additional trick up its sleeve. When you play a game on OnLive, you’re essentially receiving a video stream. Your input is sent to their servers, processed, causes something to happen in the instance of the game running on their end, and the video of that instance is sent back to you. Of course, it all happens fast enough so that the latency (when it’s working correctly) is negligible, and that’s why it’s so impressive.
Though FLARE also technically runs multiple instances of a game, the fact that it’s not streaming one dedicated instance back to you (and because of the superior underlying structure detailed above) means there’s no need for redundant copies of in-game resources. If what Square says is true, multiple instances of any one game, being played on any given console connected to FLARE, will be able to share resources and essentially use one — or at least far fewer — copies of crucial game files, which will then be available to the many “instances” of the game looking to access them.
