I found the following post on Motley Fool by jasonxsmith and thought it was interesting enough to repost here...
New Benchmarks?
Yes, these benchmarks are correct. And they are confusing.
Try to remember that the i820 chipset is a kludge. It's sort of like a weird science
experiment where you attach the head of a monkey to the body of a duck. It doesn't
work all that well. Ducks are good at flying, and monkeys are good at peeling bananas.
The big "wow" factor is that you got the combined hybrid animal to function at all.
The front-side bus of a Pentium III is designed to work with SDRAM. It can run at
100MHz or 133MHz, and it is wider than the current Rambus channel.
In between the processor and the RDRAM memory is the i820 chipset. On one side, it
looks like SDRAM, on the other, like an RDRAM-capable processor. It fakes out both
the memory and the processor into thinking they can communicate with each other. Since
neither one protests, they do communicate.
Because the width of the RDRAM channel is limited, it must pump 4 16-bit words to fill
the 64-bit word that the processor is expecting. This results in an increased latency over
SDRAM. The i820 must also translate from SDRAM addressing to RDRAM
addressing, and that is no small feat either. Also, the processor is really not designed to
take advantage of some of the strengths of RDRAM. The "wow" factor here is that you
got the two systems to talk to one another at all.
By the way, DDR would have the same problems with a Pentium III. It is limited by its
front-side bus bandwidth and its SDRAM-targeted design.
So, in a nutshell, we won't really know what RDRAM can do until Intel puts out a
processor that can use it. Even then the system may not be perfect, but it should be a lot
better than what we have today. But the current benchmarks are basically worthless -
because they run on a Pentium III.
-JasonX |
