Integral post package repair

Patent metadata

Field	Value
Publication number	US-10042700-B2
Application number	US-201615168045-A
Country	US
Kind code	B2
Filing date	May 28, 2016
Priority date	May 28, 2016
Publication date	Aug 7, 2018
Grant date	Aug 7, 2018

Abstract

A post-package repair system includes a memory channel controller, a first error counter, a scrubber, and a data processor. The memory channel controller converts data access requests to corresponding memory accesses, and provides returned data to the host interface in response to responses received from a memory interface, wherein the responses comprise returned data and a plurality of error correcting code (ECC) bits. The first error counter counts errors in the returned data, and provides a control signal in response to reaching a predetermined state. The scrubber controls the memory channel controller to read data sequentially and periodically from a plurality of addresses of a memory system, and in response to detecting a correctable error, to rewrite corrected data. The data processor is responsive to the control signal to perform a post package repair operation with the memory system in response to the control signal.

First claim

What is claimed is:

1. A post-package memory repair system comprising:
a memory channel controller for converting data access requests received from a host interface to corresponding memory accesses provided to a dynamic random access memory (DRAM) memory interface, and providing returned data to said host interface in response to responses received from said memory interface, wherein said responses comprise returned data and a plurality of error correcting code (ECC) bits;
a first error counter coupled to said memory channel controller for counting errors in said returned data, and providing a control signal in response to reaching a predetermined state;
a background scrubber coupled to said memory channel controller for controlling said memory channel controller to read data sequentially and periodically from all memory locations of a memory system at a certain rate, and in response to detecting a correctable error, to rewrite corrected data; and
a data processor coupled to said memory channel controller and responsive to said control signal to perform a post package repair operation with said memory system in response to said control signal.

2. The post-package memory repair system of claim 1 wherein said data processor comprises:
a central processing unit (CPU) core coupled to said memory channel controller; and
a firmware memory coupled to said CPU core for storing a plurality of instructions which, when executed by said CPU core, cause said CPU core to control said memory channel controller to perform said post package repair operation.

3. The post-package memory repair system of claim 2 wherein said firmware memory stores a basic input-output system (BIOS) and said plurality of instructions are part of said BIOS.

4. The post-package memory repair system of claim 2 wherein said data processor increases a scrub rate of said background scrubber in response to said memory channel controller detecting a predetermined level of errors.

5. The post-package memory repair system of claim 1 wherein:
said background scrubber reads a data element from memory of a predetermined size less than a size of a corresponding row in said memory system; and
said memory channel controller has a read/write buffer of said predetermined size for temporarily storing said data element.

6. The post-package memory repair system of claim 1 wherein said predetermined state is reached when said first error counter reaches a terminal count.

7. The post-package memory repair system of claim 1 wherein said memory channel controller comprises:
an ECC check circuit responsive to said returned data and said plurality of ECC bits to detect an error in said returned data and to determine whether said error is correctable.

8. The post-package memory repair system of claim 7 further comprising:
a plurality of additional error counters,
wherein in response to detecting an ECC error in a memory access, said ECC check circuit further provides one of: a first increment signal to said first error counter, and a plurality of additional increment signals to a corresponding plurality of additional error counters selected according to a region of said memory access in said memory system.

9. The post-package memory repair system of claim 1 wherein said DRAM memory interface is adapted to be coupled to an interface compatible with the double data rate (DDR)-physical interface (DFI) Specification.

10. The post-package memory repair system of claim 1 wherein said background scrubber is implemented as a hardware circuit.

11. The post-package memory repair system of claim 1 wherein said background scrubber is implemented by software executing on said data processor.

12. A memory controller comprising:
a memory channel controller for converting data access requests received from a host interface to corresponding memory accesses provided to a dynamic random access memory (DRAM) memory interface, and providing returned data to said host interface in response to responses received from said DRAM memory interface, wherein said responses comprise returned data and a plurality of error correcting code (ECC) bits;
wherein said memory channel controller detects an error in said returned data by performing an error check based on said plurality of error correcting code (ECC) bits corresponding to said returned data, provides a first control signal in response to detecting said error in said returned data, and provides a count signal in response to detecting said error in said returned data is a correctable error; and
a first error counter coupled to said memory channel controller for counting correctable errors in response to said count signal, and providing a second control signal in response to reaching a predetermined state.

13. The memory controller of claim 12 wherein said second control signal comprises a machine check architecture interrupt.

14. The memory controller of claim 12 wherein said memory channel controller comprises:
an ECC check circuit responsive to said returned data and said plurality of ECC bits to detect said error in said returned data and to determine whether said error is correctable.

15. The memory controller of claim 14 further comprising:
a plurality of additional error counters,
wherein said ECC check circuit further provides a selected one of said count signal to said first error counter and a plurality of additional increment signals to a corresponding plurality of additional error counters in response to detecting an ECC error in a memory access, wherein said selected one is selected for said memory access according to a region of said memory access in a memory system.

16. The memory controller of claim 12 wherein said DRAM memory interface is adapted to be coupled to an interface compatible with the double data rate (DDR)-physical interface (DFI) Specification.

17. A method comprising:
converting data access requests received from a host interface to corresponding memory accesses provided to a dynamic random access memory (DRAM) memory interface;
providing returned data to said host interface in response to responses received from said DRAM memory interface, wherein said responses comprise said returned data and a plurality of error correcting code (ECC) bits;
background scrubbing all memory locations of a memory system periodically at a certain rate for errors according to said returned data and said plurality of ECC bits, and repairing correctable errors; and
performing a post package repair operation on said memory system in response to a predetermined number of errors in a region of said memory system during said background scrubbing.

18. The method of claim 17 wherein said converting comprises:
storing said data access requests in a queue; and
arbitrating among a plurality of data access requests stored in said queue to provide said memory accesses at times determined by predetermined criteria.

19. The method of claim 17 wherein said background scrubbing comprises:
providing a read command to a next address in said memory system;
receiving said response from said memory system;
performing an ECC check according to said returned data and said plurality of ECC bits; and
incrementing an error counter if said performing said ECC check indicates an error.

20. The method of claim 19 further comprising:
performing said post package repair operation on said memory system in response to said error counter reaching a terminal count.

21. The method of claim 17 wherein said

Assignees

• Advanced Micro Devices Inc

Inventors

• Brandl Kevin M

Classifications

• G11C29/42

  using error correcting codes [ECC] or parity check · CPC title

• G06F3/064

  Management of blocks · CPC title

• G11C29/78 Primary

  using programmable devices · CPC title

• G06F11/106

  Correcting systematically all correctable errors, i.e. scrubbing · CPC title

• G11C11/4093

  Input/output [I/O] data interface arrangements, e.g. data buffers · CPC title