Baje sem nekje bral (samo pa zdaj nikjer ne najdem tistega linka), da je Cell tisti, ki požre največ energije. Sicer je res, da je imela Toshiba prste vmes, ampak je IBM namenil za izdelavo tega dele Power5 procesorjev, je pa približno tako okrnjeno, kot je bila serija PowerPC G5 uporabljeno v Mac-ih. Hmm...
Google Nekaj komadov skup zlepljenih PowerPC 601 CPU-jev sestavlja Cell?
The CELL's SIMD processing unit
As you can see, IBM has eliminated the instruction window and its attendant control logic, in favor of adding more storage space and more execution hardware. A Cell SPE doesn't do register renaming or instruction reording, so it needs neither a rename register file or a reorder buffer. The actual architecture of the Cell SPE is a dual-issue, statically scheduled SIMD processor with a large local storage (LS) area. In this respect, the individual SPUs are like very simple, PowerPC 601-era processors.
The main differences between an individual SPE and an early RISC machine are twofold. First, and most obvious, is the fact that the Cell SPE is geared for single-precision SIMD computation. Most of its arithmetic instructions operate on 128-bit vectors of four 32-bit elements. So the execution core is packed with vector ALUs, instead of the traditional fixed-point ALUs. The second difference, and this is perhaps the most important, is that the L1 cache has been replaced by 256K of locally addressable memory. The SPE's ISA, which is not VMX/Altivec-derivative (more on this below), includes instructions for using the DMA controller to move data between main memory and local storage. The end result is that each SPE is like a very small vector computer, with its own "CPU" and RAM.
This RAM functions in the role of the L1 cache, but the fact that it is under the explicit control of the programmer means that it can be simpler than an L1 cache. The burden of managing the cache has been moved into software, with the result that the cache design has been greatly simplified. There is no tag RAM to search on each access, no prefetch, and none of the other overhead that accompanies a normal L1 cache. The SPEs also move the burden of branch prediction and code scheduling into software, much like a VLIW design.
The SPE's very simple front end can take in two instructions at a time, check to see if they can operate in parallel, and then issue them either in parallel or in program order. These two instructions then travel down one of two pipes, "even" or "odd," to be executed. After execution, they're put back in sequence (if necessary) by the very simple commit unit and their results are written back to local memory. The individual SPUs can throw a lot overboard, because they rely on a regular, general-purpose POWERPC processor core to do all the normal kinds of computation that it takes to run regular code. The Cell system features eight of these SPUs all hanging off a central bus, with one 64-bit POWERPC core handling all of the regular computational chores. Thus all of the Cell 's "smarts" can reside either on the PPC core, while the SPUs just do the work that's assigned to them.
To sum up, IBM has sort of reapplied the RISC approach of throwing control logic overboard in exchange for a wider execution core and a larger storage area that's situated closer to the execution core. The difference is that instead of the compiler taking up the slack (as in RISC), a combination of the compiler, the programmer, some very smart scheduling software, and a general-purpose CPU do the kind of scheduling and resource allocation work that the control logic used to do.
