Patent Application
20170277463
This application claims priority under 35 U.S.C. §119 to Korean Patent Application No. 10-2016-0036643 filed on Mar. 28, 2016, the disclosure of which is herein incorporated by reference in its entirety.
Exemplary embodiments relate to a semiconductor memory technology, and more particularly, to a nonvolatile dual in-line memory module capable of independently accessing volatile memory devices therein with reduced number of signal lines, and a method for operating the same.
In most cases, a single controller is coupled to and controls two or more memory devices.
As shown in
As shown in
As the number of memory devices coupled to the single controller increases, the number of required signal lines increases, which increases difficulty in system design and Increases the fabrication cost.
Various embodiments are directed to a nonvolatile dual in-line memory module capable of independently accessing volatile memory devices therein with a reduced number of signal lines, and capable of performing a backup operation to data of a valid area against power failure by a host.
In an embodiment, a nonvolatile memory module may include: a plurality of volatile memory devices sharing a data bus through which data is transmitted and a control bus through which a command and an address are transmitted; at least one nonvolatile memory device; and a controller suitable for backing up data stored in the plurality of volatile memory devices into the nonvolatile memory device at a power failure of a host, and restoring data backed up in the nonvolatile memory device to the plurality of volatile memory devices at recovery of the power failure, the controller including: a command/address snooping logic suitable for snooping on a command and an address inputted from a memory controller of the host, and analyzing valid areas of data stored in the respective volatile memory devices; and a command/address control logic suitable for selecting the volatile memory device having the valid area of data based on analysis results of the command/address snooping logic, and backing up selected volatile memory into the nonvolatile memory device.
The command/address control logic may set a command address latency (CAL) for recognizing the volatile memory device having the valid area of data to a first value, and sets a command address latency of remaining volatile memory devices to a second value different from the first value.
The second value may be greater than the first value, and a difference between the second value and the first value may be equal to or greater than a row address to column address delay time (tRCD: RAS to CAS delay).
The difference between the second value and the first value may be less than a row precharge time (tRP).
The command/address control logic may include a logic which performs a distributed refresh operation for uniformly distributing a refresh cycle over the plurality of volatile memory devices while programming a memory page of the nonvolatile memory device; a logic which operates the plurality of volatile memory devices under a low power mode, in which the plurality of volatile memory devices use a lower power than in a normal power mode, while a new memory page of the nonvolatile memory device is prepared and written; and a logic suitable for recovering the plurality of volatile memory devices to the normal power mode after the new memory page of the nonvolatile memory device is written.
In an embodiment, a method for operating a nonvolatile memory module including a plurality of volatile memory devices which share a data bus through which data is transmitted and a control bus through which a command and an address are transmitted, a nonvolatile memory device, and a controller which backs up data stored in the plurality of volatile memory devices in the nonvolatile memory device or restores data backed up in the nonvolatile memory device to the plurality of volatile memory devices, according to a fail/recovery of power of a host, may include: snooping, by the controller, on a command and an address inputted to the plurality of volatile memory devices from a memory controller of the host; analyzing the command and the address, and analyzing valid areas of data stored in the respective volatile memory devices; and selecting the volatile memory device having the valid area of data based on results of the analyzing, and backing up selected volatile memory into the nonvolatile memory device, when a fail in the power of the host is detected or backup is instructed from the memory controller of the host.
The controller may set a command address latency (CAL) for recognizing the volatile memory device having the valid area of data to a first value, and may set a command address latency of remaining volatile memory devices, to a second value different from the first value.
The second value may be greater than the first value, and a difference between the second value and the first value may be equal to or greater than a row address to column address delay time (tRCD: RAS to CAS delay).
The difference between the second value and the first value may be less than a row precharge time (tRP).
The backing up selected volatile memory may include performing a distributed refresh operation for uniformly distributing a refresh cycle over the plurality of volatile memory devices while programming a memory page of the nonvolatile memory device; operating the plurality of volatile memory devices under a low power mode, in which the plurality of volatile memory devices use a lower power than in a normal power mode, while a new memory page of the nonvolatile memory device is prepared and written; and recovering the plurality of volatile memory devices to the normal power mode after the new memory page of the nonvolatile memory device is written.
A nonvolatile memory module may include: volatile memory devices suitable for storing data provided from a host through a common data bus, a nonvolatile memory device suitable for backup of data stored in the volatile memory devices, and a controller suitable for analyzing valid areas of data stored in the respective volatile memory devices by snooping on a command and an address provided to the respective volatile memory devices from the host through a common control bus, selecting one or more volatile memory devices having the valid areas of data among the volatile memory devices based on a result of the analysis, and backing up data of the selected volatile memory devices in the nonvolatile memory device, upon a power failure of the host.
According to embodiments of the present invention, it is possible to independently access volatile memory devices with a reduced number of signal lines of a data bus in the nonvolatile dual in-line memory module, and it is possible to perform a backup operation to data of a valid area when a power failure of a host occurs.
Various embodiments will be described below in more detail with reference to the accompanying drawings. The present invention may, however, be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present invention to those skilled in the art. Throughout the disclosure, like reference numerals refer to like parts throughout the various figures and embodiments of the present invention.
The present disclosure relates to a nonvolatile dual in-line memory module in which a controller may independently access volatile memory devices sharing a data bus and a control bus with a reduced number of signal lines. Hereafter, to facilitate understanding of a nonvolatile dual in-line memory module in accordance with an embodiment, descriptions will be made sequentially from detailed configurations of the entire system.
Per-Dram Addressability (PDA) Mode of Volatile Memory Device
First, descriptions will be made for the PDA mode and the command address latency (CAL) of a volatile memory device.
In the PDA mode, an independent mode register set operation is performed for each volatile memory device. When the PDA mode is set, validity of all mode register set commands may be determined according to the signal level of a zeroth data pad DQ0. If the signal level of the zeroth data pad DQ0 is ‘0’ after a write latency (WL=AL+CWL where WL denotes write latency, AL denotes additive latency and CWL denotes CAS write latency), all mode register set commands applied may be determined as valid, and, if the signal level of the zeroth data pad DQ0 is ‘1,’ all mode register set commands applied may be determined as invalid and may be neglected.
Referring to
If the signal level of the zeroth data pad DQ0 is continuously retained as ‘1’ at the point of time 202, the mode register set command MRS applied at the point of time 201 is determined as invalid and is thus neglected. That is, the setting operation of the volatile memory device is not performed.
Command Address Latency (CAL) of Volatile Memory Device
A CAL indicates the timing difference between a chip select signal CS and the remaining signals among control signals transferred through a control bus (CMD/ADDR_BUS). If the CAL is set, a volatile memory device determines as valid only the control signals inputted after a time corresponding to the CAL passing from the enable time of the chip select signal CS. The value of the CAL may be set by a mode register set (MRS).
Since the command CMD and the address ADDR are also applied at points of time 304 and 306 when a time corresponding to the CAL (3 clocks) passes from points of time 303 and 305 and the chip select signal CS is enabled, the command CMD and the address ADDR applied at the points of time 304 and 306 may be recognized as valid by the volatile memory device.
Basic Configuration of Dual in-Line Memory Module (DIMM)
Referring to
Control signals are transferred from the controller 400 to the volatile memory devices 410_0 and 410_1 through the control bus CMD/ADDR_BUS. The control signals may include a command CMD, an address ADDR and a clock CK. The command CMD may include a plurality of signals. For example, the command CMD may include an active signal (ACT), a row address strobe signal (RAS), a column address strobe signal (CAS) and a chip select signal (CS). Although the chip select signal CS is a signal which is included in the command CMD, the chip select signal CS is separately shown in the drawing to represent that the volatile memory devices 410_0 and 410_1 share the same chip select signal CS. The address ADDR may include a plurality of signals. For example, the address ADDR may include a multi-bit bank group address, a multi-bit bank address and a multi-bit normal address. The clock CK may be transferred from the controller 400 to the volatile memory devices 410_0 and 410_1 for synchronized operations of the volatile memory devices 410_0 and 410_1. The clock CK may be transferred in a differential scheme including a clock (CK_t) and a clock bar (CK_c) acquired by inverting the clock (CK_t).
The data bus DATA_BUS may transfer multi-bit data DATA0 to DATA3 between the controller 400 and the volatile memory devices 410_0 and 410_1. The respective volatile memory devices 410_0 and 410_1 are provided with data pads DQ0 to DQ3 to be respectively coupled with data lines DATA0 to DATA3 of the data bus DATA_BUS. A particular data pad such as, data pad DQ0 of the respective volatile memory devices 410_0 and 410_1 may be coupled to different data lines DATA0 and DATA1. The specified data pads DQ0 may be used to set latencies for recognizing the control signals on the control bus CMD/ADDR_BUS.
The controller 400 may control the volatile memory devices 410_0 and 410_1 through the control bus CMD/ADDR_BUS, and may exchange data with the volatile memory devices 410_0 and 410_1 through the data bus DATA_BUS. The controller 400 may be provided in the DIMM, may set latencies for allowing the volatile memory devices 410_0 and 410_1 to recognize signals on the control bus CMD/ADDR_BUS, to different values, and may access a volatile memory device desired between the volatile memory devices 410_0 and 410_1, by using the latencies. This will be described below in detail with reference to
The first volatile memory device 410_0 and the second volatile memory device 410_1 may share the control bus CMD/ADDR_BUS and the data bus DATA_BUS. The first volatile memory device 410_0 and the second volatile memory device 410_1 may also share the chip select signal CS. The first volatile memory device 410_0 and the second volatile memory device 410_1 may be set with different latencies for control signals to be transmitted through the control bus CMD/ADDR_BUS. A latency may mean the timing difference between a reference signal for example, the chip select signal CS and the remaining signals CMD and ADDR among signals on the control bus CMD/ADDR_BUS. Due to the fact that the first volatile memory device 410_0 and the second volatile memory device 410_1 are set with different latencies with respect to the control bus CMD/ADDR_BUS, the first volatile memory device 410_0 and the second volatile memory device 410_1 may be independently accessed by the controller 400, which will be described below in detail with reference to
As may be seen from
Basic CAL Setting Operation of DIMM
Referring to
At step 511, the controller 400 may control the first volatile memory device 410_0 and the second volatile memory device 410_1 to enter a PDA mode. This may be implemented by applying the command CMD corresponding to a mode register set command (MRS) and applying the address ADDR as a combination corresponding to entry to the PDA mode.
At step 512, the command address latency CAL of the first volatile memory device 410_0 may be set to ‘0’. This may be implemented by applying the command CMD as the combination corresponding to the mode register set command (MRS), and applying the address ADDR as a combination corresponding to setting of the CAL to ‘0’ and applying the signal level of ‘0’ to the zeroth data line DATA0 corresponding to the zeroth data pad DQ0 of the first volatile memory device 410_0 after a write latency WL (WL=AL+CWL) passes from the application time of the command CMD. Referring to
At step 513, the command address latency CAL of the second volatile memory device 410_1 may be set to ‘3’. This may be implemented by applying the command CMD as the combination corresponding to the mode register set command (MRS), and applying the address ADDR as a combination corresponding to setting of the CAL to ‘3’ and applying the signal level of ‘0’ to the first data line DATA1 corresponding to the zeroth data pad DQ0 of the second volatile memory device 410_1 after the write latency WL (WL=AL+CWL) passes from the application time of the command CMD. Referring to
Since the command address latencies CAL of the first volatile memory device 410_0 and the second volatile memory device 410_1 are set differently from each other, the controller 400 may access the first volatile memory device 410_0 by applying the command/address CMD/ADDR at the enable time of the chip select signal CS at step 521 or may access the second volatile memory device 410_1 by applying the command/address CMD/ADDR after 3 clocks from the enable time of the chip select signal CS at step 522.
In the operations at the points of time 701, 702, 703, 704, 707, 708, 709 and 710, it is possible to access only one volatile memory device of the first volatile memory device 410_0 and the second volatile memory device 410_1. Further, in the operations at the points of time 705, 706, 711 and 712, it may be possible to access both the first volatile memory device 410_0 and the second volatile memory device 410_1 by applying the valid command CMD at the enable times of the chip select signal CS as well as applying the valid command CMD after 3 clocks from the enable times of the chip select signal CS.
According to the embodiment described above with reference to
While it was exemplified in the above embodiment that the volatile memory devices 410_0 and 410_1 are set by the controller 400 to have different latencies with respect to the control bus CMD/ADDR_BUS, this is for an illustration purpose only and it is to be noted that the volatile memory devices 410_0 and 410_1 may be programmed to have permanently different latencies. For example, the latencies of the volatile memory devices 410_0 and 410_1 with respect to the control bus CMD/ADDR_BUS may be fixed when fabricating the volatile memory devices 410_0 and 410_1, or the latencies of the volatile memory devices 410_0 and 410_1 with respect to the control bus CMD/ADDR_BUS may be fixed through permanent setting for example, a setting using a fuse circuit, after fabrication of the volatile memory devices 410_0 and 410_1.
Furthermore, the difference in command address latency CAL between the volatile memory devices 410_0 and 410_1 may be equal to or greater than a delay time tRCD, that is, RAS to CAS delay, of a column address from a row address. Additionally, the difference in the values of command address latencies CAL between the volatile memory devices 410_0 and 410_1 may be less than a row precharge time tRP. That is, dCAL (CAL difference)≧tRCD, and dCAL<tRP.
Referring to
At a point of time 802, the chip select signal CS may be enabled, and a read operation RD may be instructed by the command/address CMD/ADDR. Then, the first volatile memory device 410_0 may perform a read operation by recognizing the read operation RD at the point of time 802. Furthermore, at the point of time 802 when 3 clocks pass after the chip select signal CS is enabled at the point of time 801, the second volatile memory device 410_1 may recognize the read operation RD from the command/address CMD/ADDR. However, since an active operation had not been performed in the second volatile memory device 410_1, the second volatile memory device 410_1 may determine the read operation RD instructed by the command/address CMD/ADDR, as illegal, and may not perform a read operation. If dCAL is less than tRCD, a mis-operation may occur as the second volatile memory device 410_1 recognizes the active operation ACT instructed to the first volatile memory device 410_0. Such a mis-operation may be prevented when dCAL≧tRCD. Additionally, at a point of time 803 when 3 clocks pass after the chip select signal CS is enabled at the point of time 802, the second volatile memory device 410_1 may recognize the read operation RD from the command/address CMD/ADDR. However, since an active operation had not been performed in the second volatile memory device 410_1, the second volatile memory device 410_1 may determine the read operation RD instructed by the command/address CMD/ADDR, as illegal, and may not perform a read operation.
At a point of time 804, the chip select signal CS may be enabled, and a precharge operation PCG may be instructed by the command/address CMD/ADDR. Then, the first volatile memory device 410_0 may perform a precharge operation by recognizing the precharge operation PCG at the point of time 804. At a point of time 805 when 3 clocks pass after the chip select signal CS is enabled at the point of time 804, the second volatile memory device 410_1 may recognize the precharge operation PCG from the command/address CMD/ADDR and may perform a precharge operation. Since a precharge operation does not consider whether an active operation has previously been performed, the precharge operation may be performed even by the second volatile memory device 410_1.
At a point of time 806, the chip select signal CS may be enabled, and an active operation ACT may be instructed by the command/address CMD/ADDR. Then, the first volatile memory device 410_0 may perform an active operation by recognizing the active operation ACT at the point of time 806. If dCAL is set greater than tRP, a mis-operation may occur as the second volatile memory device 410_1 recognizes the active operation ACT instructed through the command/address CMD/ADDR and performs an active operation, from the point of time 806. Such a mis-operation may be prevented since dCAL<tRP.
At a point of time 807, the chip select signal CS may be enabled, and a write operation WT may be instructed by the command/address CMD/ADDR. Then, the first volatile memory device 410_0 may perform a write operation by recognizing the write operation WT at the point of time 807. At the point of time 807 when 3 clocks pass after the chip select signal CS is enabled at the point of time 806, the second volatile memory device 410_1 may recognize the write operation WT from the command/address CMD/ADDR. However, since an active operation had not been performed in the second volatile memory device 410_1, the second volatile memory device 410_1 may determine the write operation WT instructed by the command/address CMD/ADDR, as illegal, and may not perform a write operation. At a point of time 808 when 3 clocks pass after the chip select signal CS is enabled at the point of time 807, the second volatile memory device 410_1 may recognize the write operation WT from the command/address CMD/ADDR. However, the second volatile memory device 410_1 may determine the write operation WT instructed by the command/address CMD/ADDR, as illegal, and may not perform a write operation.
As described above with reference to
Configuration and Operation of Nonvolatile Dual In-Line Memory Module (NVDIMM)
In
Referring to
When power HOST_VDD and HOST_VSS of the host is normal, the register 950 may buffer a command, an address and a clock provided from the memory controller 9 of the host through a host control bus HOST_CMD/ADDR_BUS, and may provide the command, the address and the clock to first and second groups of volatile memory devices 911 to 914 and 921 to 924 through the control bus CMD/ADDR_BUS. When the power HOST_VDD and HOST_VSS of the host is normal, the first group of volatile memory devices 911 to 914 may transmit/receive data to/from the memory controller 9 of the host through the third data buses DATA_BUS3_1 to DATA_BUS3_4 respectively corresponding thereto, and the second group of volatile memory devices 921 to 924 may transmit/receive data to/from the memory controller 9 of the host through the fourth data buses DATA_BUS4_1 to DATA_BUS4_4 respectively corresponding thereto. That is, when the power HOST_VDD and HOST_VSS of the host is normal, first and second groups of volatile memory devices 911 to 914 and 921 to 924 may independently communicate with the memory controller 9 of the host through corresponding one among the third data buses DATA_BUS3_1 to DATA_BUS3_4 and the fourth data buses DATA_BUS4_1 to DATA_BUS4_4.
If the power fall detector 960 detects a fail in the power HOST_VDD and HOST_VSS of the host as levels of voltages forming the power HOST_VDD and HOST_VSS of the host become unstable, the supply of the power HOST_VDD and HOST_VSS of the host to the NVDIMM 900 is interrupted. Then, emergency power EMG_VDD and EMG_VSS of the auxiliary power supply 10 is supplied to the NVDIMM 900. The auxiliary power supply 10 may be realized through a large capacity capacitor, for example, a super capacitor, and may supply the emergency power EMG_VDD and EMG_VSS while the data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 are backed up in the nonvolatile memory device 930. Although it is illustrated in
When the fail in the power HOST_VDD and HOST_VSS of the host is notified from the power fail detector 960, control over first and second groups of volatile memory devices 911 to 914 and 921 to 924 is changed from the memory controller 9 of the host to the controller 940 of the NVDIMM 900. Then, the register 950 may buffer a command, an address and a clock provided from the controller 940, and may provide the command, the address and the clock to first and second groups of volatile memory devices 911 to 914 and 921 to 924 through the control bus CMD/ADDR_BUS. The first group of volatile memory devices 911 to 914 may exchange data with the controller 940 through the first data bus DATA_BUS1, and the second group of volatile memory devices 921 to 924 may exchange data with the controller 940 through the second data bus DATA_BUS2. The controller 940 may read the data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 through the control bus CMD/ADDR_BUS, the first data bus DATA_BUS1 and the second data bus DATA_BUS2, and may store that is, back up the read data in the nonvolatile memory device 930.
The data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 backed up in the nonvolatile memory device 930 upon occurrence of the fail in the power HOST_VDD and HOST_VSS of the host may be transmitted to and stored in first and second groups of volatile memory devices 911 to 914 and 921 to 924 after the power HOST_VDD and HOST_VSS of the host returns to a normal state. Such a restoration operation may be performed according to control of the controller 940, and, after restoration is completed, control over first and second groups of volatile memory devices 911 to 914 and 921 to 924 may revert back from the controller 940 of the NVDIMM 900 to the memory controller 9 of the host.
The first group of volatile memory devices 911 to 914 share the same control bus CMD/ADDR_BUS and data bus DATA_BUS1 in communication with the controller 940. Similarly, the second group of volatile memory devices 921 to 924 share the same control bus CMD/ADDR_BUS and data bus DATA_BUS2 in communication with the controller 940. Nevertheless, the controller 940 may independently access an individual volatile memory device among the first group of volatile memory devices 911 to 914, and may independently access an individual volatile memory device among the second group of volatile memory devices 921 to 924. In this regard, descriptions are made above with reference to
The first and second groups of volatile memory devices 911 to 914 and 921 to 924 may be DRAMs or may be not only DRAMs but also different kinds of volatile memory devices. The nonvolatile memory device 930 may be a NAND flash. However, the nonvolatile memory device 930 is not limited to such, and may be any kind of nonvolatile memory device, for example, a NOR flash, a resistive RAM (RRAM), a phase RAM (PRAM), a magnetic RAM (MRAM) or a spin transfer torque MRAM (STT-MRAM).
The components in the NVDIMM 900 shown in
For example, the controller 940, the register 950 and the power fail detector 960 may be configured through one chip or may be configured through a plurality of chips. Moreover, the numbers of the first group of volatile memory devices 911 to 914, the second group of volatile memory devices 921 to 924 and the nonvolatile memory device 930 used in the NVDIMM 900 may be different from the illustration of
The NVDIMMs 900 in
Through the multiplexers 1101 to 1104, the data pads DQ0 to DQ3 of the first group of volatile memory devices 911 to 914 and the third data buses DATA_BUS3_1 to DATA_BUS3_4 may be coupled when the first group of volatile memory devices 911 to 914 communicate with the memory controller 9 of the host, and the data pads DQ0 to DQ3 of the first group of volatile memory devices 911 to 914 and the first data bus DATA_BUS1 may be coupled when the first group of volatile memory devices 911 to 914 communicate with the controller 940.
Through the multiplexers 1105 to 1108, the data pads DQ0 to DQ3 of the second group of volatile memory devices 921 to 924 and the fourth data buses DATA_BUS4_1 to DATA_BUS4_4 may be coupled when the second group of volatile memory devices 921 to 924 communicate with the memory controller 9 of the host, and the data pads DQ0 to DQ3 of the second group of volatile memory devices 921 to 924 and the second data bus DATA_BUS2 may be coupled when the second group of volatile memory devices 921 to 924 communicate with the controller 940.
Since the NVDIMM 900 of
Power-Down Backup Operation
At step S1110, first and second groups of volatile memory devices 911 to 914 and 921 to 924 communicate with the memory controller 9 of the host at a normal time, and control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is performed by the memory controller 9 of the host in the NVDIMM 900 illustrated in
At step S1120, determination is made for whether a trigger condition may be satisfied for backing up the data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 in the nonvolatile memory device 930. For example, detection of a fail in the power HOST_VDD and HOST_VSS of the host may satisfy the trigger condition. Alternatively, when a backup operation is performed according to an instruction of the memory controller 9 of the host, the instruction by the memory controller 9 of the host for the backup operation may satisfy the trigger condition.
At step S1130, control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 may be changed from the memory controller 9 of the host to the controller 940 of the NVDIMM 900. Further, power used by the NVDIMM 900 is changed from the power HOST_VDD and HOST_VSS of the host to the emergency power EMG_VDD and EMG_VSS supplied by the auxiliary power supply 10. Moreover, as a control subject is changed to the controller 940, a data bus used by the first group of volatile memory devices 911 to 914 is changed from the third data buses DATA_BUS3_1 to DATA_BUS3_4 to the first data bus DATA_BUS1, and a data bus used by the second group of volatile memory devices 921 to 924 is changed from the fourth data buses DATA_BUS4_1 to DATA_BUS4_4 to the second data bus DATA_BUS2.
At step S1140, the controller 940 individually sets command address latencies CAL on the first and second groups of volatile memory devices 911 to 914 and 921 to 924 which share the control bus CMD/ADDR_BUS and the data buses DATA_BUS1 and DATA_BUS2.
Referring to
For example, this may be achieved by setting the command address latency CAL of target volatile memory devices such as, volatile memory devices 911 and 921 of the respective first and second groups of the volatile memory devices 911 to 914 and 921 to 924 to a first value for example, 0 and by setting the command address latency CAL of the remaining ones other than the target volatile memory devices 911 and 921 of the respective first and second groups of the volatile memory devices 911 to 914 and 921 to 924 to a second value for example, 3.
At step S1150, the controller 940 reads the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 by using the set command address latency CAL. For example, the controller 400 may read the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 by accessing the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 of which command address latency CAL is set to the first value for example, 0, through applying the command/address CMD/ADDR at the enable time of the chip select signal CS. Since the remaining volatile memory devices 912 to 914 and 922 to 924 other than the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 are set to the second value for example, 3 in the command address latency CAL thereof, they neglect the read command from the controller 940.
The scheme of step S1140 that the controller 940 sets command address latencies CAL independently on the first and second groups of volatile memory devices 911 to 914 and 921 to 924 which share the control bus CMD/ADDR_BUS and the data buses DATA_BUS1 and DATA_BUS2, and the scheme of step S1150 that the controller 940 reads data by accessing the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 which has a specified command address latency CAL may be understood from the descriptions made above with reference to
At step S1160, a data backup operation is performed as the data read from volatile memory devices are written in the nonvolatile memory device 930. For example, the data read from the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 may be backed up in a page of the nonvolatile memory device 930.
At step S1170, a determination is made as to whether the nonvolatile memory page is full (i.e. the data write of the page is completed). If the nonvolatile memory page is not full, the process may return to step S1140.
For example, if data stored in the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 remains, the controller 940 may perform the read operation for the remaining data stored in the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 by setting the command address latency CAL of the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 to the first value for example, 0 and by setting the command address latency CAL of the remaining volatile memory devices 912 to 914 and 922 to 924 other than the target volatile memory devices 911 and 921 to the second value for example, 3 at step S1140.
For another example, when all of the data stored in the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 are backed up, the controller 940 at step S1140 may set the command address latency CAL of another target volatile memory devices for example, the volatile memory devices 912 and 922 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 to the first value for example, 0 and may set the command address latency CAL of the remaining volatile memory devices 911, 913, 914, 921, 923 and 924 other than the target volatile memory devices 912 and 922 to the second value for example, 3. Then, at step S1150, the controller 940 may read the target volatile memory devices 912 and 922 through the setting of the command address latency CAL. Although not illustrated the selective reading of the first and second groups of volatile memory devices 911 to 914 and 921 to 924 which share the control bus CMD/ADDR_BUS and the data buses DATA_BUS1 and DATA_BUS2, through the setting of the command address latency CAL, may be performed to all of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 by individually selecting as the target volatile memory device each volatile memory device in the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924.
When it is determined at step S1170 that the nonvolatile memory page is full, the process proceeds to step S1180 where the nonvolatile memory page is programmed.
When programming the memory page of the nonvolatile memory device 930, it is necessary to check whether data not read from the first and second groups of volatile memory devices 911 to 914 and 921 to 924 still exists. Thus, during the program operation to the memory page of the nonvolatile memory device 930 of step S1180, the controller 940 may perform a refresh operation for the first and second groups of volatile memory devices 911 to 914 and 921 to 924. For example, a distributed refresh operation of uniformly distributed refresh cycles may be performed to the first and second groups of volatile memory devices 911 to 914 and 921 to 924 such that all rows are turned on before iterating a task and data is read when refresh is not performed in the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924.
When a new nonvolatile memory page is prepared and written (i.e. S1160-S1180), the first and second groups of volatile memory devices 911 to 914 and 921 to 924 may operate under a low power mode, in which the first and second groups of volatile memory devices 911 to 914 and 921 to 924 use a lower power than in a normal power mode. After the new nonvolatile memory page is prepared and written, when data to back up still remains in the first and second groups of volatile memory devices 911 to 914 and 921 to 924 and a memory page to program exists in the nonvolatile memory device 930, the first and second groups of volatile memory devices 911 to 914 and 921 to 924 are recovered to the normal power mode such that the operation of reading data to be backed up is performed continuously.
At step S1190, a determination is made for whether data to be backed up remains in the first and second groups of volatile memory devices 911 to 914 and 921 to 924. When data to be backed up does not remain, the power-down backup operation may end, and the NVDIMM 900 may be shut down. If data to be backed up remains, the process may proceed to step S1140, and the backup operation for the remaining data is performed.
Power-Up Restoration Operation
A power-up restoration operation may be performed when the power HOST_VDD and HOST_VSS of the host returns to the normal state or as the memory controller 9 of the host instructs a restoration operation. Since the power HOST_VDD and HOST_VSS of the host has returned to the normal state, the power-up restoration operation may be performed through the power HOST_VDD and HOST_VSS of the host.
In an example, the NVDIMM 900 may perform the restoration operation in the state in which the NVDIMM 900 is shut down after completing the backup operation described above with reference to
At step S1220, a determination is made for whether a restoration condition is satisfied, and, if the restoration condition is satisfied, restoration of data from the nonvolatile memory device 930 to the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is started.
At step S1230, the controller 940 individually sets command address latencies CAL on the first and second groups of volatile memory devices 911 to 914 and 921 to 924 which share the control bus CMD/ADDR_BUS and the data buses DATA_BUS1 and DATA_BUS2. As described above for the backup operation with reference to
For example, the command address latency CAL of the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 may be set to a third value for example, 0, and the command address latency CAL of the remaining volatile memory devices 912 to 914 and 922 to 924 other than the target volatile memory devices 911 and 921 may be set to a fourth value for example, 3.
At step S1240, data restoration to the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 may be performed by writing the data read from the nonvolatile memory device 930 into the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 through the command address latency CAL.
At step S1250, determination is made for whether data to restore remains in the nonvolatile memory device 930. If data to restore remains, the process may proceed to step S1230, and the restoration operation may be performed for the remaining data.
For example, when data restoration for the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 is completed, the controller 940 at step S1230 may set the command address latency CAL of another target volatile memory devices for example, the volatile memory devices 912 and 922 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924, to the third value for example, 0 and may set the command address latency CAL of the remaining volatile memory devices 911, 913, 914, 921, 923 and 924 other than the target volatile memory devices 912 and 922 to the fourth value for example, 3. Then, at step S1240, the controller 940 may restore the data read from the nonvolatile memory device 930 to the target volatile memory devices 912 and 922 through the setting of the command address latency CAL. The data restoration operation may be performed for all of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 by individually setting the command address latency CAL of each volatile memory device as the target volatile memory device in the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924, setting the command address latency CAL of the remaining volatile memory devices other than the target volatile memory device in the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 to the fourth value, and then restoring the data read from the nonvolatile memory device 930 into the target volatile memory device. The difference dCAL between the third value and the fourth value of the command address latency CAL may be set in such a way as to satisfy dCAL≧tRCD and dCAL<tRP.
When it is determined at the step S1250 that data to restore does not remain, to prepare for a case when the power HOST_VDD and HOST_VSS of the host is down again, it is necessary to secure sufficient storage capacity of the nonvolatile memory device 930 to back up the data stored in the first and second groups of volatile memory devices 911 to 914 and 921 to 924 before control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is changed to the memory controller 9 of the host.
At step S1260, a determination is made for whether erased or empty blocks are sufficient for the data backup in the nonvolatile memory device 930. For example, a determination is made for whether an amount of the erased blocks is sufficient to cover the entire capacity of the first and second groups of volatile memory devices 911 to 914 and 921 to 924 or the used amount or the valid range of data currently stored in the first and second groups of volatile memory devices 911 to 914 and 921 to 924 of the nonvolatile memory device 930. If sufficient erased blocks do not exist in the nonvolatile memory device 930, a new block is erased in the nonvolatile memory device 930 at step S1270.
When sufficient erased blocks exist in the nonvolatile memory device 930, control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is changed from the controller 940 of the NVDIMM 900 to the memory controller 9 of the host at step S1280, and the power-up restoration operation is completed.
Thereafter, the NVDIMM 900 may be used by the memory controller 9 of the host, and may operate in the same state as at step S1110 described above with reference to
Power-Down Interrupt Operation
When the power fall detector 960 detects occurrence of a fail in the power HOST_VDD and HOST_VSS of the host or the backup operation is instructed by the memory controller 9 of the host, the power-down backup operation is performed as described above with reference to
At step S1310, the power-down backup operation described above with reference to
At step S1320, a determination is made as to whether the power HOST_VDD and HOST_VSS of the host is recovered during the power-down backup operation. For example, when the power HOST_VDD and HOST_VSS of the host returns to the normal state and is supplied to the NVDIMM 900 or a signal corresponding thereto is received from the memory controller 9 of the host during the power-down backup operation, it may be determined that the power HOST_VDD and HOST_VSS of the host is recovered during the power-down backup operation.
During power-down interrupt operation, since the NVDIMM 900 has not completed the power-down backup operation yet, the NVDIMM 900 is not shut down and the first and second groups of volatile memory devices 911 to 914 and 921 to 924 are still storing data therein. Therefore, the data restoration process as in the power-up restoration operation may not be necessary. However, there is a chance that the memory page of the nonvolatile memory device 930 is occupied by the data of the first and second groups of volatile memory devices 911 to 914 and 921 to 924 during the data backup process of step S1310, therefore it may not be able to prepare for the next occurrence of a fail in the power HOST_VDD and HOST_VSS of the host. Accordingly, it may be necessary that control over first and second groups of volatile memory devices 911 to 914 and 921 to 924 is changed to the memory controller 9 of the host after securing sufficient space of the nonvolatile memory device 930 to back up data of the first and second groups of volatile memory devices 911 to 914 and 921 to 924 of the NVDIMM 900 for the next occurrence of a fail in the power HOST_VDD and HOST_VSS of the host.
At step S1330, a determination is made for whether erased or empty blocks are sufficient for the data backup in the nonvolatile memory device 930. For example, a determination is made for whether an amount of the erased blocks is sufficient to cover the entire capacity of the first and second groups of volatile memory devices 911 to 914 and 921 to 924 or the used amount or the valid range of data currently stored in the first and second groups of volatile memory devices 911 to 914 and 921 to 924 of the nonvolatile memory device 930.
When sufficient erased blocks exist in the nonvolatile memory device 930, control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is changed from the controller 940 of the NVDIMM 900 to the memory controller 9 of the host at step S1340, and the NVDIMM 900 may be immediately used by the memory controller 9 of the host.
However, when sufficient erased blocks do not exist in the nonvolatile memory device 930, a new block is erased in the nonvolatile memory device 930 at step S1350 to prepare for the next occurrence of a fail in the power HOST_VDD and HOST_VSS of the host.
Here, the block erased from the nonvolatile memory device 930 may have included the data backed up from the first and second groups of volatile memory devices 911 to 914 and 921 to 924. When a fail in the power HOST_VDD and HOST_VSS of the host occurs again during the power-down interrupt operation, rather than performing the entire power-down backup operation illustrated in
At step S1360, a determination is made for whether a trigger condition is satisfied for backing up the data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 in the nonvolatile memory device 930. As described above, the trigger condition may be detection of a fall in the power HOST_VDD and HOST_VSS of the host or backup instruction from the memory controller 9 of the host. When the trigger condition is not satisfied, the process returns to step S1330.
When it is determined that the trigger condition is satisfied, data of the first and second groups of volatile memory devices 911 to 914 and 921 to 924, which is backed up at step S1310 and then erased at step S1350, is backed up again at step S1370.
For example, it may be presumed that the data of the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 are backed up in the erased block of the nonvolatile memory device 930 at step S1310 and then the block storing the backed up data is erased at step S1350. Thus, the controller 940 of the NVDIMM 900 may set the command address latency CAL of the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 to a fifth value for example, 0. Then, after setting the command address latency CAL of the remaining volatile memory devices other than the target volatile memory devices 911 and 921 to a sixth value for example, 3, a volatile memory region which is storing data erased from the nonvolatile memory device 930 at step S1350 may be selected and read through the set value of the command address latency CAL. The read data are back up again in the nonvolatile memory device 930 at step S1370. After the selective backup operation of the step S1370 is completed, the power-down backup operation interrupted at the start time of the power-down interrupt operation may be resumed at step S1380.
Command/Address Snooping of NVDIMM
Referring to
The command and the address of the controller 940 outputted from the command/address control logic 1420 are transferred to a register clock driver (RCD) 1440 through a multiplexer 1450. The register clock driver 1440 may buffer a command, an address and a clock provided from the memory controller 9 of the host or the controller 940 of the NVDIMM, and may provide the command, the address and the clock to first and second groups of volatile memory devices 911 to 914 and 921 to 924 through the control bus CMD/ADDR_BUS. Furthermore, the register clock driver 1440 may have the function of restoring any distortion of the command and the address provided from the memory controller 9 of the host or the controller 940 of the NVDIMM. Hereafter, an embodiment of performing a power-down backup operation through command/address snooping will be described with reference to
Selective Backup Operation Using Command/Address Snooping of NVDIMM
Where the power HOST_VDD and HOST_VSS of the host is supplied normally, as aforementioned above, the first and second groups of volatile memory devices 911 to 914 and 921 to 924 are independently controlled by the memory controller 9 of the host. At step S1510, the controller 940 of the NVDIMM may snoop on the command and address inputted to the first and second groups of volatile memory devices 911 to 914 and 921 to 924 from the memory controller 9 of the host through the command/address snooping logic 1410.
At step S1520, the command/address snooping logic 1410 analyzes a valid area (i.e. an area where data is stored in volatile memory) of data stored in each volatile memory device of the first and second groups of volatile memory devices 911 to 914 and 921 to 924. The command/address snooping logic 1410 may analyze the valid areas of data stored in respective volatile memory devices and accumulate analysis results while control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is performed by the memory controller 9 of the host.
At step S1530, a determination is made as to whether a trigger condition is satisfied for backing up the data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 in the nonvolatile memory device 930. As aforementioned above, the trigger condition is a condition for backing up the data stored in the first and second groups of volatile memory devices 911 to 914 and 921 to 924 in the nonvolatile memory device 930. For example, detection of a fail in the power HOST_VDD and HOST_VSS of the host or instruction of a backup operation from the memory controller 9 of the host may satisfy the trigger condition.
When the trigger condition is satisfied, a volatile memory device having the valid area of data is selected at step S1540 based on the accumulated analysis results of step S1520, and the data stored in the selected volatile memory device is backed up in the nonvolatile memory device 930 at step S1550.
For example, it is presumed that the volatile memory device selected at step S1540 is the target volatile memory devices 911 and 921 of the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 described above with reference to
When data of valid areas are stored in some volatile memory devices among the first and second groups of volatile memory devices 911 to 914 and 921 to 924 which share the same control bus CMD/ADDR_BUS and the first and second data buses DATA_BUS1 and DATA_BUS2, only the volatile memory devices of valid areas are selected, and the command address latencies CAL of the selected volatile memory devices may be sequentially set to the first value, and the command address latency CAL of unselected volatile memory devices may be set to the second value. Therefore, by backing up only data of the valid area, it is possible to substantially shorten a time required to back up data.
Preferential Backup Operation Using Command/Address Snooping of NVDIMM
When the power HOST_VDD and HOST_VSS of the host is supplied normally, the first and second groups of volatile memory devices 911 to 914 and 921 to 924 are independently controlled by the memory controller 9 of the host. At step S1610, the controller 940 of the NVDIMM may snoop on the command and the address inputted to the first and second groups of volatile memory devices 911 to 914 and 921 to 924 from the memory controller 9 of the host through the command/address snooping logic 1410.
At step S1620, the command/address snooping logic 1410 analyzes an amount of stored data in each volatile memory device of the first and second groups of volatile memory devices 911 to 914 and 921 to 924. The command/address snooping logic 1410 may analyze the amount of stored data in respective volatile memory devices and accumulate analysis results while control over the first and second groups of volatile memory devices 911 to 914 and 921 to 924 is performed by the memory controller 9 of the host.
At step S1630, a determination is made for whether a trigger condition is satisfied for backing up the data of first and second groups of volatile memory devices 911 to 914 and 921 to 924 in the nonvolatile memory device 930. The trigger condition is a condition for backing up the data stored in the first and second groups of volatile memory devices 911 to 914 and 921 to 924 in the nonvolatile memory device 930. Detection of a fail in the power HOST_VDD and HOST_VSS of the host or instruction of a backup operation from the memory controller 9 of the host may satisfy the trigger condition.
When the trigger condition is satisfied, the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 may be prioritized in order of amount of stored data at step S1640, and the data stored in the volatile memory devices are backed up in the nonvolatile memory device 930 according to the priorities at step S1650.
For example, volatile memory devices having the largest amount of stored data are the volatile memory device 912 among the first group of volatile memory devices 911 to 914 and the volatile memory device 922 among the second group of volatile memory devices 921 to 924. The controller 940 may selectively read the volatile memory devices 912 and 922 having the largest amount of stored data by setting the command address latency CAL of the volatile memory devices 912 and 922 to a first value for example, 0 and setting the command address latency CAL of the remaining volatile memory devices 911, 913 and 914 and 921, 923 and 924 to a second value for example, 3. The read data are backed up in the nonvolatile memory device 930 as described above.
The backup operation may be performed for each volatile memory device in the respective first and second groups of volatile memory devices 911 to 914 and 921 to 924 according to the priority set at step S1640.
As is apparent from the above descriptions, when the NVDIMM 900 performs backup and restoration operations of data by a fail and a recovery of the power HOST_VDD and HOST_VSS of the host, the first group of volatile memory devices 911 to 914 of the NVDIMM 900 shares the control bus CMD/ADDR_BUS and the first data bus DATA_BUS1 in communication with the controller 940, and the second group of volatile memory devices 921 to 924 of the NVDIMM 900 shares the control bus CMD/ADDR_BUS and the second data bus DATA_BUS2 in communication with the controller 940. The controller 940 may back up and restore data by accessing independently the first group of volatile memory devices 911 to 914 through setting command address latencies CAL to different values. Similarly, the controller 940 may back up and restore data by accessing independently the second group of volatile memory devices 921 to 924 through setting command address latencies CAL to different values.
In one or more exemplary embodiments, the functions described herein may be realized in hardware, software, firmware or any combination thereof. If realized in software, the functions may be stored or transmitted as one or more instructions or codes on a machine-readable medium, that is, a computer program product such as a computer-readable medium. The computer-readable medium includes a communication medium including a computer storage medium and any medium that facilitates transfer of a computer program from a place to another place. A storage medium may be any usable medium that may be accessed by a computer. In a non-limiting example, such a computer-readable medium may be accessed by a RAM, a ROM, an EEPROM, a CD-ROM, an optical disk memory device, a magnetic disk memory device, a magnetic storage device or a computer, and may include any medium that may be used in carrying or storing desired program codes in the form of instructions or data structures. The disk and the disc as used herein include a compact disc (CD), a laser disc, an optical disc, a digital versatile disc (DVD), a floppy disk and a blue-ray disc, where the disk usually reproduces data magnetically but the disc reproduces data optically. Any combination thereof should be included within the scope of a computer-readable medium.
Although various embodiments have been described for illustrative purposes, it will be apparent to those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the following claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10-2016-0036643 | Mar 2016 | KR | national |
