I'm working on a design that has a 16 bit DDR4 memory controller and also has a reference schematic. The reference design has two separate x8 DDR4 memory devices connected to the controller DQ[0:7] and DQ[8:15], respectively, and both memory devices share the same address lines. What would be the practical difference between using two x8 DDR4 memory devices as in the reference schematic instead of one x16 DDR4 device that would be cheaper, considering that the final amount of memory is the same?
Add a comment
|
1 Answer
\$\begingroup\$
\$\endgroup\$
Performance is very much dependent on the SW application and what type of transaction are used more often. in general the DDR bandwidth can be reduced going from x8 to x16, simply because x8 has 16 available open pages vs. x16 has 8 available open pages. the specific SW needs to be tested on the system to give you specific concrete results to determine the performance differences.
Source: https://community.nxp.com/t5/Layerscape/LS1043-DDR4-memory-organization-x8-vs-x16/td-p/1686200
So it comes down to benchmarks and application and here is an example of that with benchmarks.
If you know your application and it can do better with more pages then use x8
