Cell technology: the punch behind PS3

How many processors make light work for Sony's console

TODO alt text

For its new PlayStation 3 console to be as successful as the multi-million selling PlayStation 2 before it, Sony took on an ambitious design challenge - it wanted to offer PlayStation owners more power in every department.

It's arguably why the development of the PlayStation 3 has taken so long and is one of the reasons why the console, priced at a hefty £425 in the UK, costs so much. Compare it to the £299 launch price of the Xbox 360 and the rock-bottom £179 price tag for Nintendo's Wii.

Sony also envisioned the PS3 doing a whole host of other things besides playing Gran Turismo HD. So while the PlayStation 3 could be as simplistic as a games machine, with its Blu-ray disc drive, music and video playback and broadband connectivity it can also act as a home entertainment powerhouse. Of course, such multimedia mastery requires a serious amount of computing power.

So, Sony (who jumped into the R&D bed for a threesome with IBM and Toshiba) dreamt up a new type of processor - the Cell Broadband Engine or simply the ' Cell '. The Cell is an entirely new computing "architecture". Yes, it's a super fast chip in its own right, but it's one that's designed to be part of a distributed computing system that lets you spread out processing tasks among other Cells, even if they're not physically in the same device.

On paper at least, the Cell processor had the potential to win the console for Sony before the console war had even started. In reality, though, there's much more to the PS3 vs. Xbox 360 vs. Nintendo Wii battle than who's got the fastest silicon. The PS3 might be more powerful than its rivals, but by being the last of the next-gen machines to launch, it's playing catch-up with Microsoft and Nintendo.

What is a Cell?

In 2000, Sony teamed up with Toshiba and IBM to build the Cell as a joint project. The concept behind the Cell was the brainchild of Sony legend Ken Kutaragi . He imagined the Cell processor as a "super computer on a chip" and had a vision of distributed computing (i.e. separate systems working together on shared tasks) that he likened to biological cells working together in the human body.

Needless to say, the Cell is quite different from typical PC processors. Conventional CPUs work in a linear way by reading a set of instructions. They try to do as many of these instructions as possible, as fast as possible, and at the same time.

But their effectiveness is limited. When the Cell was first envisaged, CPUs like AMD's Athlon 64 and Intel's Pentium ranges contained a single computing core on a chip. While a single core processor can handle several sets of instructions at the same time, it requires a high clock speed to do so. It's why both AMD and Intel have since rolled out multi-core processors ( Athlon 64 X2 , Core 2 Duo ) to enable computers to process instructions in parallel.

The Cell is quite different from a conventional single-core or dual-core CPU. The Cell architecture used in the PS3 has nine separate processing elements - one conventional, multi-threaded core and another eight supporting units called Synergistic Processing Elements (SPEs).

The conventional core (often dubbed the 'Power Processing Element' or PPE) in the PS3 is a 3.2GHz 64-bit IBM Power Architecture CPU, capable of processing two sets of instructions simultaneously. The SPEs act as co-processors for the PPE and six of them are available for programs to make use of, while the seventh is reserved for the OS. The eighth is disabled, apparently to improve chip yields.

The PPE controls the show inside the PS3, the SPEs do all the hard work.

Related news