DDR2 Memory Controller User's Guide

www.ti.com
Col. 0 Col. 1 Col. 2 Col. 3 Col. 4 Col. M−1 Col. M
Row 0, bank 0
Row 0, bank 1
Row 0, bank 2
Row 0, bank P
Row 1, bank 1
Row 1, bank 0
Row 1, bank 2
Row 1, bank P
Row N, bank 2
Row N, bank 1
Row N, bank 0
Row N, bank P
Peripheral Architecture
Figure 11 shows how the DSP memory map is partitioned into columns, rows, and banks. Note that during
a linear access, the DDR2 memory controller increments the column address as the logical address
increments. When the DDR2 memory controller reaches a page/row boundary, it moves onto the same
page/row in the next bank. This movement continues until the same page has been accessed in all banks.
To the DDR2 SDRAM, this process looks as shown on Figure 14 .
By traversing across banks while remaining on the same row/page, the DDR2 memory controller
maximizes the number of activated banks for a linear access. This results in the maximum number of
open pages when performing a linear access being equal to the number of banks. Note that the DDR2
memory controller never opens more than one page per bank.
Ending the current access is not a condition that forces the active DDR2 SDRAM row/page to be closed.
The DDR2 memory controller leaves the active row open until it becomes necessary to close it. This
decreases the deactivate-reactivate overhead.
Figure 13. Logical Address-to-DDR2 SDRAM Address Map
A M is number of columns (as determined by PAGESIZE) minus 1, P is number of banks (as determined by IBANK)
minus 1, and N is number of rows (as determined by both PAGESIZE and IBANK) minus 1.
SPRUEK5A – October 2007 DSP DDR2 Memory Controller 21
Submit Documentation Feedback