To view parent comment, click here.
To read all comments associated with this story, please click here.
You don't make up much at all by using RISC. That's the basic fallacy that people have to get over. You save a little bit of die space, a few stages in the pipeline, and allow some better register allocation. On the other hand, your code density takes a dive, along with your i-cache hit rate.
The performance differences between the Gx and x86 chips aren't due to RISC vs CISC. The G4's FPU performance is bad not because its RISC, but because it's got a single FPU pipeline. The G5's integer performance isn't bad because its RISC, but because its got a long pipeline, only two integer units, and two-cycle latency for simple instructions. Conversely, the G5's floating-point performance isn't great because its RISC, but beceause its got two symmetric FPU pipes each of which are capable of a multiply-add per cycle.
In the end, it comes down to the design of the CPU. PowerPC vs x86 is irrelevent.





Member since:
2006-01-03
The lagging clock speed was the killer; you can only make up for so much by using RISC.