Bus trade-offs
Thinking about the differences between a RIMM and DIMM physical bus, IMO the key differences are:
A RIMM bus is narrow, physically long and requires a high speed clock for data bandwidth.
A DIMM bus is wide, physically short and requires a lower speed clock for data bandwidth.
RIMM's
The bus is highly engineered. Because the paths are relatively long, (they wind through the RIMM's), and at very high frequency's, they are designed to use strip or micro strip line to achieve transmission line characteristics. The line lengths must be tightly controlled to equal physical length to minimize "jitter", including compensation for "bends" in the lines. And to achieve transmission line characteristics, every active signal requires adjacent Via's to the ground plane to provide a path for the image current. (The image current is analogous to a two wire transmission line where the second wire is the image in the ground plane). The propagation delays are matched against control signals through innovative designs. Even so, the data "eye" is relatively narrow due to the high frequencies. (800 MHz on PC800). But with proper controls in the manufacture of the PCB's this is acceptable. IMO, this system will always be more expensive.
DIMM's
The data bus is four times as wide and consequently equivalent data bandwidth is achieved at much lower clock rates. Unlike RIMM's where the bus winds through the RIMM's, a DIMM bus merely extends to the adjacent connector. In the case of DDR266, the clock frequency is 1/3 of a PC800, and consequently the data "eye" is much wider. Since the bus doesn't appear to be designed as stringently as a RIMM, I suspect the system will be noisier. The wider data "eye" compensates for this. I suspect these systems will be cost competitive with existing PC133 designs.
Both systems have their advantages, and I suspect the memory chip makers and the market will be the final decision maker in their relative success.
These are just my opinions, feel free to take shots. |
