Memory error analysis for determining potentially faulty memory components

• US 7,404,118 B1
• Filed: 09/02/2004
• Issued: 07/22/2008
• Est. Priority Date: 09/02/2004
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Vecoven, Frederic Louis Ghislain Gabriel, Baguette, Benoit Leon Roger Ghislain
• Primary Examiner(s): Kerveros; James C
• Application Number:
  US10/932,768
• Time in Patent Office: 1,419 Days
• Field of Search: 714/718, 714/719, 714/723, 714/753, 714/764
• US Class Current: 714/723
• CPC Class Codes: G06F 11/1044 with specific ECC/EDC distribution
• CPC Class Codes: G11C 2029/0411 Online error correction
• CPC Class Codes: G11C 2029/4402 Internal storage of test result, quality data, chip identification, repair information

Abstract

Accordingly, there has been described a computer system with a plurality of memory components where individual bits from multiple words are distributed among the memory components. An error analyzer is operable to identify a memory component as potentially faulty by accessing a table mapping syndromes to memory components using generated syndromes.

Claims

1. A computer system comprising a plurality of memory components where bits are distributed among a plurality of the memory components and an error analyzer operable to, for an uncorrectable multiple bit error, identify a memory component as potentially faulty by using generated syndromes to access a table operable to map syndromes to memory components, wherein the syndromes are generated using bits of an error correction code, wherein the error correction code is a SEC-DED-S4ED code.

2. The computer system of claim 1, wherein the error analyzer is operable to use the results of table accesses for each of a plurality of generated syndromes to identify a memory component as faulty when it is identified as potentially faulty for multiple generated syndromes.

3. A computer system comprising:

a plurality of memory components, wherein bits from a group of bits that comprises one or more words are stored distributed between memory components, the group of bits comprising data bits and error correction code bits, wherein the error correction code is a SEC-DED-S4ED code;

a syndrome generator, operable to generate a syndrome from regenerated error correction code bits for the group of bits and stored error correction code bits for the stored group of bits;

a table comprising entries linking syndromes to memory components; and

an error analyzer operable to, for an uncorrectable multiple bit error, identify a memory component as potentially faulty by accessing the table using a generated syndrome.

4. The computer system of claim 3 wherein the table is generated from a list of syndromes for each memory component for all combinations of bits for the group of bits located in the memory component.

5. The computer system of claim 3, wherein the error analyzer is operable to record the results of the table access.

6. The computer system of claim 3, wherein the error analyzer is operable to compare the results of table accesses for each of a plurality of generated syndromes to identify a memory component as faulty when it is identified as potentially faulty for multiple generated syndromes.

7. The computer system of claim 3, wherein the table is a two dimensional table with one of said axes being representative of syndrome codes and the other axis being representative of memory components.

8. The computer system of claim 7, wherein rows in the table are representative of syndrome codes and columns in the table are representative of memory components.

9. The computer system of claim 3, wherein each memory component comprises a predetermined maximum number of bits of a group of bits according to a corresponding number of error bits detectable by an error correcting code of the error correction code bits.

10. The computer system of claim 3, wherein each memory component comprises a maximum of four bits of a group of bits.

11. The computer system of claim 3, wherein each group of bits comprises a quadword.

12. The computer system of claim 11, wherein each quadword comprises 144 bits and each memory component comprises four bits from a quadword, and wherein each quadword includes 128 data bits, 9 error correction code bits, and 7 memory tag bits.

13. The computer system of claim 3, wherein each memory component is a memory chip.

14. The computer system of claim 13 comprising at least one memory module, wherein the memory module comprises a plurality of memory chips.

15. The computer system of claim 14, wherein the memory module is a dual in line memory module.

16. A method of identifying a faulty memory component in a computer system, the method comprising:

generating a syndrome from regenerated error correction code bits for a group of bits and stored error correction code bits for a stored group of bits, wherein bits from the group of bits, which comprises data bits and error correction code bits in one or more words, are distributed between memory components, wherein the error correction code is a SEC-DED-S4ED code; and

for an uncorrectable multiple bit error, identifying a memory component as potentially faulty by accessing a table comprising entries linking syndromes to memory components using a generated syndrome.

17. The method of claim 16, wherein the table is generated from a list of possible syndromes for each memory component for all combinations of bits located in the memory component.

18. The method of claim 17, comprising recording the results of the table access.

19. The method of claim 16, comprising comparing the results of table accesses for each of a plurality of generated syndromes to identify a memory component as faulty when it is identified as potentially faulty for multiple generated syndromes.

20. The method of claim 16, wherein the table is a two dimensional table with one of said axes being representative of syndrome codes and the other axis being representative of memory components.

21. The method claim 20, wherein rows in the table are representative of syndrome codes and columns in the table are representative of memory components.

22. The method of claim 16, wherein each memory component comprises a predetermined maximum number of bits of a group of bits according to a corresponding number of error bits detectable by the error correcting code of the error correction code bits.

23. The method of claim 16, wherein each memory component comprises a maximum of four bits of a group of bits.

24. The method of claim 16, wherein each group of bits comprises a quadword.

25. The method of claim 24, wherein each quadword comprises 144 bits and each memory component comprises four bits from a quadword, wherein each quadword comprises 128 data bits, 9 error correction code bits, and 7 memory tag bits.

26. The method of claim 16, wherein each memory component is a memory chip.

27. A computer system comprising:

memory means for storing bits from a group of bits that comprises one or more words, wherein individual bits from the group of bits are stored distributed between memory components, said group of bits comprising data bits and error correction code bits, wherein the error correction code is a SEC-DED-S4ED code;

a syndrome means for generating a syndrome from regenerated error correction code bits for a said group of bits and stored error correction code bits for said stored group of bits; and

table means for entries linking syndromes to memory components; and

error analysis means for identifying a memory component, for an uncorrectable multiple bit error, as potentially faulty by accessing the table using a generated syndrome.

28. A computer program product comprising program code operable to perform the steps of:

generating a syndrome from regenerated error correction code bits for a group of bits and stored error correction code bits for a stored group of bits, wherein individual bits from the group of bits, which comprises comprising data bits and error correction code bits in one or more words, is distributed between memory components, wherein the error correction code is a SEC-DED-S4ED code; and

for an uncorrectable multiple bit error, identifying a memory component as potentially faulty by accessing a table comprising entries linking syndromes to memory components using a generated syndrome.