I watched this video on YouTube, and it mentions DDR5 is addressed using 31 bits. 5 bits for the bank, 16 bits for the row, and 10 bits for the column. I feel like a 32-bit address would "fit" more, and than 1 bit is missing. If words for example are 32-bit, and an address is stored in memory, and loaded from a program or by indirect addressing, then 1 bit has to be stripped from it. Was this 1 bit doing something meaningful, or is it simply just thrown away? If the latter, is there any good rationale to explain why 31-bit addresses was superior to 32-bit addresses?
10
-
It maybe that a 32G ram uses 31 bits and 64G ram uses 32 bits.– Rohit GuptaCommented Jan 14, 2023 at 3:12
-
Did you check if there lines are address lines or data lines?– Romeo NinovCommented Jan 14, 2023 at 6:55
-
Segmented addressing is the answer.– harrymcCommented Jan 14, 2023 at 10:39
-
Bank, row and column addressing does not directly translate to address "bits". It is a different method of working. Also if all data transfers are (for example) 32 or 64 bits at a time then you never need to use the least significant bits in an address. You simply transfer an entire row of data that contains the address you need and then at the memory controller you select the correct byte from the data received.– Mokubai ♦Commented Jan 14, 2023 at 11:14
-
en.wikipedia.org/wiki/DDR5_SDRAM#Operation also states that there are 17 row bits, not 16. Your video may not be correctly describing how DDR5 operates, or you are conflating "address bits" with the vastly different physical approach to memory access that is actually used.– Mokubai ♦Commented Jan 14, 2023 at 12:13
|
Show 5 more comments