Memory system and method using stacked memory device dice
View Patent ↗A method and apparatus for organizing memory for a computer system including a plurality of memory devices, connected to a logic device, particularly a memory system having a plurality of stacked memory dice connected to a logic die, with the logic device having capability to analyze and compensate for differing delays to the stacked devices stacking multiple dice divided into partitions serviced by multiple buses connected to a logic die, to increase throughput between the devices and logic device allowing large scale integration of memory with self-healing capability.
1. A device comprising:
a plurality of semiconductor dies stacked and connected together, at least one of the dies comprising a plurality of partitions; and
a plurality of vaults, at least one vault comprising a grouping of the partitions in one of the dies, each vault having an independent connection to a logic die by at least one wide bus;
wherein the independent connection of the logic die to the at least one vault is reallocated from a first wide bus to a second wide bus in the event of a malfunction in the first wide bus.
2. The device as in claim 1 , wherein the number of partitions in each vault is the same as the number of stacked semiconductor dies.
3. The device as in claim 1 , wherein the device is a memory device.
4. The device as in claim 1 , further comprising switch circuits configured to form the vaults.
5. The device as in claim 2 , wherein the switch circuits are located on the dies.
6. The device as in claim 1 , wherein the logic dice is configured to analyze and compensate for differing delays to the stacked semiconductor dies.
7. The device as in claim 1 , wherein each of the dies further comprise switch circuits connected to the logic die.
8. A method for organizing memory, comprising:
stacking a plurality of semiconductor dies, at least one semiconductor die having a plurality of memory cells;
partitioning the at least one die into a plurality of vaults, at least one vault having a plurality of cells;
connecting the vaults to a logic die with wide busses, the connection to vaults being switched with a switching signal, each vault having an independent connection to the logic die; and
reallocating vaults from a first wide bus of the wide busses to a second wide bus of the wide busses in the event of a malfunction in the first wide bus.
9. The method as in claim 8 , wherein the number of partitions in each vault is the same as the number of stacked semiconductor dies.
10. The method as in claim 8 , wherein the switching signal originates from the logic die and each die is provided with a plurality of switches responsive to the switching signal.
11. The method as in claim 8 , wherein the wide busses are Through Silicon Vias (TSVs).
12. The method as in claim 8 wherein the switching signal is configured to provide a data valid period of separate semiconductor dies each have the same timing.
13. The method as in claim 10 , wherein the switching signal is configured to provide a maximum valid data period.
14. The method as in claim 13 , wherein the switching signal is selectively delayed to some dies to provide a maximum valid data period.
15. The method as in claim 8 , wherein the switching signal is a RDQS signal sent to each of semiconductor dies having timing adjustment circuits for each RDQS.
16. The method as in claim 15 , wherein each semiconductor dies output data synchronized to their RDQS.
17. The method as in claim 16 , wherein the RDQS signal adjusts RDQS timing so that data valid period of semiconductor dies have same timing.
18. The method as in claim 17 , wherein the RDQS signal changes when supply voltage and temperature change while semiconductor dies are working to update the timing of data valid period continuously.