Three very interesting things happened over the past couple of weeks here at AnandTech:
Intel’s Spring IDF 2005 turned out to be a multi-core CPU festival, with Intel being even more open than ever before about future plans for their multi-core microprocessor architectures. Intel has over 10 multi-core CPU designs in the works, and they made that very clear at IDF.
At GDC 2005, AGEIA announced that they had developed a Physics Processing Unit (PPU) that could be used to enable extremely realistic physics and artificial intelligence models.
Johan De Gelas went one step further in his quest for more processing power earlier this week to find that there’s quite a lot of potential for multi-core CPUs in the gaming market, at the expense of increasing development times.
So, what do these three things have in common? The aggregate of the three basically summarize what we’ve come to know as the Cell microprocessor - a multi-core CPU, part of which is designed for parallel physics/AI processing for which it will be quite difficult to program.
Cell, at a high level, isn’t too difficult to understand; it’s how the designers got there that is most intriguing. It’s the design decisions and building blocks of Cell that we’ll focus on here in this article, with an end goal of understanding why Cell was designed the way it was.
A joint venture between IBM, Sony and Toshiba, the Cell microprocessor is the heart and soul of Sony’s upcoming Playstation 3. However, this time around, Sony and Toshiba are planning to use Cell (or parts of it) in everything from consumer electronics to servers and workstations. If you don’t already have the impression, publicly, Cell has been given some very high aspirations as a microprocessor, especially a non-x86 microprocessor.