Switched interface stacked-die memory architecture
Systems and methods disclosed herein include those that may receive a memory request including a requested memory address and may send the memory request directly to an address decoder associated with a stacked-die memory vault without knowing whether a repair address is required. If a subsequent analysis of the memory request shows that a repair address is required, an in-process decode of the requested memory address can be halted and decoding of the repair address initiated.
1. An apparatus, comprising:
a memory vault comprising a stacked plurality of memory arrays, each memory array located on one of a plurality of stacked memory dies;
a memory vault controller (MVC) located on a logic die stacked with the stacked memory dies and communicatively coupled to the memory vault;
a memory vault repair logic (MVRL) component of the MVC, the MVRL to receive a memory request including a requested memory address, the MVRL including at least one lookup table to translate the requested memory address into a repair address, and the MVRL to send at least one of the memory request or a modified memory request to a selected vault, the modified memory request to include the repair address.
2. The apparatus of claim 1 , wherein the MVC is configured to provide at least one of memory sequencing operations, data buffering operations, damaged memory array repair operations, or damaged through-wafer interconnect (TWI) repair operations.
3. The apparatus of claim 1 , further comprising:
a feedback loop operable to determine whether the requested memory address references a non-defective memory location.
4. The apparatus of claim 1 , further comprising:
a spare memory die component of the memory vault to use as a repair array.
5. The apparatus of claim 1 , further comprising:
at least one spare memory array fabricated on the logic die common with the MVRL to use as a repair array.
6. The apparatus of claim 5 , wherein the spare memory array comprises at least one of a static random-access memory (SRAM) or a dynamic random-access memory (DRAM).
7. An apparatus, comprising:
a memory vault comprising a stacked plurality of memory arrays, each memory array located on one of a plurality of stacked memory dies;
a memory vault controller (MVC) located on a logic die stacked with the stacked memory dies and communicatively coupled to the memory vault;
a memory vault repair logic (MVRL) component of the MVC, the MVRL to receive a memory request including a requested memory address and to send at least one of the memory request or a modified memory request to a selected vault, the modified memory request to include a repair address;
a partial address decoder portion of the MVRL, the partial address decoder to partially decode the requested memory address;
at least one repair address lookup table communicatively coupled to the partial address decoder, the at least one repair address lookup table to translate the requested memory address to the repair address; and
bad block logic coupled to the partial address decoder, the bad block logic to select the at least one repair address lookup table from a plurality of repair address lookup tables.
8. The apparatus of claim 7 , wherein the plurality of repair address lookup tables comprises at least one of a direct-mapped table, a fully associative tag random-access memory (RAM), or a set associative tag RAM.
9. The apparatus of claim 7 , wherein the bad block logic is configured to select the at least one repair address lookup table based upon at least one of a number of bad blocks or a number of words in a bad block as determined by the partial address decoder.
10. The apparatus of claim 7 , further comprising:
a repair decode assessment module coupled to the partial address decoder, the repair decode assessment module to estimate a latency associated with translating the requested memory address to the repair address; and
a variable latency decision module (VLDM) coupled to the repair decode assessment module, the VLDM to cause the memory request including the requested memory address to be passed to a memory address decoder associated with the selected vault if the estimated latency is greater than a selected amount.
11. The apparatus of claim 7 , further comprising:
an arithmetic/logic unit (ALU) coupled to the repair address lookup table to calculate the repair address using an address offset stored in the repair address lookup table as an entry associated with the requested memory address.
12. The apparatus of claim 7 , further comprising:
a memory address decoder coupled to the memory vault to decode a memory address into at least one of a memory die identifier, a memory bank identifier, a row address, or a column address; and
address bus gating logic coupled to the memory address decoder to pass the requested memory address to at least one of the memory address decoder or the partial address decoder.
13. The apparatus of claim 12 , further comprising:
an address selector component of the memory address decoder to reject a partially decoded requested memory address and to initiate decoding of the repair address if the requested memory address is determined to reference at least one defective memory cell.
14. An apparatus, comprising:
a memory vault comprising a stacked plurality of memory arrays, each memory array located on one of a plurality of stacked memory dies;
a memory vault controller (MVC) located on a logic die stacked with the stacked memory dies and communicatively coupled to the memory vault;
a memory vault repair logic (MVRL) component of the MVC, the MVRL to receive a memory request including a requested memory address and to send at least one of the memory request or a modified memory request to a selected vault, the modified memory request to include a repair address;
a partial address decoder portion of the MVRL, the partial address decoder to partially decode the requested memory address;
at least one repair address lookup table communicatively coupled to the partial address decoder, the at least one repair address lookup table to translate the requested memory address to the repair address;
bad block logic coupled to the partial address decoder, the bad block logic to select the at least one repair address lookup table from a plurality of repair address lookup tables; and
a feedback loop operable to determine whether the requested memory address references a non-defective memory location.