In Theory: Can CPU co-processor boost PS3 power?

DF dissects Sony's patent application.

The PS3 blogosphere is alive with reports that PlayStation 3 is set to be beefed up with the release of a new Cell co-processor, based on a recently unearthed patent application located by FreePatentsOnline. However, such reports appear to be a combination of wishful thinking and a fanciful interpretation of the paperwork released by SCE.

The patent itself is pretty much self-explanatory: Sony is looking to protect the means by which it plans to interface multiple "mini-Cell" processors for potential workstations, home PCs - and yes - games consoles. In the example application given in the patent documentation, the CPU itself consists of a single PPU main processor, just like the PS3, but only has four SPUs - one of which is dedicated to encrypting and decrypting the traffic across the bus to the other "mini-Cells".

The notion of multiple Cells working together in a distributed computing network is nothing new. In fact, it was one of the core philosophies of PlayStation architect Ken Kutaragi, and he discussed it at length during the run-up to the PS3's launch. He envisaged a world where multiple Cell-based devices could talk to each other and create a giant super-computer network. This "new" patent simply seems to be a technical explanation of how, potentially, Cell-based systems could communicate with one another.

Traditionally, Sony doesn't exactly go out of its way to obscure patents related to the PS3, as the recent PS2 upgrade module patent demonstrated - the invention being discussed there was clearly and obviously talking about Sony consoles past and present. Tellingly there is no real mention of the PS3 specifically in this new submission. Combine this with the technological and commercial realities and the notion that this patent will result in a direct PS3 upgrade is tenuous to say the least.

However, this is not to say that the PS3 is not capable of distributed computing along the general lines laid out in the patent. Indeed, enough of Ken Kutaragi's vision remains in the final PS3 design to make the concept a reality, and there have already been at least two public demos showing the idea at work.

Back in 2005, Kutaragi's notion for distributed computing relied on the Cell units being able to transmit and receive data to one another quickly and easily. This manifested in the final PlayStation 3 in the form of the gigabit LAN port on the rear of the unit. This was seemingly a hugely over-engineered solution to a very basic problem: in this case, hooking up a PS3 to a router. It's safe to say that the real reason for its inclusion was to connect the console to other computers for the rapid transfer of data.

The ultra-fast (but not very good) CodecSys h.264 video encoder is probably the only major use found for the theoretic maximum 128MB per second bandwidth the gigabit LAN port offers, outside of PS3 game development. However, the recently revealed PS2 "upgrade" patent - which essentially sees a hypothetical PS2 chipset attached to the PS3 - also uses the superior gigabit bandwidth to transfer across the video output of the adaptor.

However, we have seen two viable, working tech demos from none other than Polyphony Digital which suggest that Kutaragi's distributed computing dream could manifest in some form in PS3 games. Back in October 2008, prototypes of Gran Turismo 5 surfaced showing the game running at 3840x2160 on a massive, so-called "4K" display. Another demo showed GT5 running at the conventional 1080p resolution, but this time operating at a staggering 240 frames per second. The secret behind this achievement? Distributed computing: in this case, GT5 was running across four PS3s, synchronised and talking to each other using the gigabit LAN port.

In the case of the 4K resolution demo, each console updated a quarter of the screen at 1080p and outputted it at 60Hz. With the 240FPS demo, the consoles again ran at 1080p60 but on a 4.2ms delay cascade from one console to the next. The only thing stopping this demo working at home - assuming you have a 4K or 240Hz projector/screen - is the use of an additional piece of kit that merged the four HDMI outputs before presenting them to the bespoke display.

But to invoke the title of this piece, in theory multiple PS3s could indeed be networked together to achieve a superior gaming experience. The same basic principle Polyphony used could be employed - the same code on each console could be run simultaneously, with each PS3 producing a different video output based on the same scene. The "slave" PS3 would beam across the frame it has created to the "master" console via LAN, which would make use of both rendered frames and output the result over HDMI. A 720p30 game could then run at 720p60, or a stereoscopic 3D game could have each individual viewpoint rendered by each console, meaning 3D without the kind of compromises we've seen on many of the titles released to date.

The only limitation here is the amount of bandwidth that can be squeezed out of the gigabit LAN port. Its 128MB per second throughput is the theoretical maximum, and 720p60 at the native 24-bit RGB of the HDMI output is around 150MB/s. However, 720p30 should be a comfortable fit. 1080p could be achieved with compression - though this, and indeed the process of transmitting the data across the port to begin with, could add latency.

Whether we'll ever actually see anything like this is unlikely. Running the same game across two PS3s for an enhanced experience would have ultra-limited appeal and would represent a marketing nightmare - though conceivably, more people have two PS3s than own a 3DTV. It's just as wacky as the notion of upgrading the PS3 via bolt-on CPUs. A far more effective boost would be achieved via a RAM and/or GPU expansion - though the lack of expansion ports would be something of an insurmountable problem here.

If there's one area of the PS3 where there's a surfeit of power, it's in the existing CPU: four years on from release, the PS3's Cell remains a processing powerhouse and there is no real need for this part of the console to be enhanced at all for current-generation games.

There's also a more fundamental question: where would such an adaptor plug in, exactly? Cell's internal communications bandwidth is in the region of 50GB/s. Communicating with an external secondary processor with a miserly 128MB/s via gigabit LAN isn't going to cut it. Over and above that, the advantages such an expansion would offer would not dramatically improve the games we're playing today - innovations such as Kinect and Move fundamentally affect the gameplay experience in a way extra processing power on its own is unable to match.

With Ken Kutaragi having departed Sony and Cell development having slowed down somewhat in recent years, it is difficult to imagine what part the innovative architecture has to play in any future console, aside from improving the chances of PS3 backwards compatibility. However, at the very least, the existence of this recent patent application suggests that SCE's engineers are, at least, still investigating the latent potential of the Cell tech.

Comments (59)

Comments for this article are now closed, but please feel free to continue chatting on the forum!