Browse Prior Art Database

Method of Collecting and Tracking Memory Module's Field Performance Data In a Personal Computer

IP.com Disclosure Number: IPCOM000013615D
Original Publication Date: 2000-Jan-01
Included in the Prior Art Database: 2003-Jun-18
Document File: 2 page(s) / 49K

Publishing Venue

IBM

Abstract

The disclosed method provides a simple and effective way for collecting and tracking memory module's field performance data

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 58% of the total text.

Page 1 of 2

  Method of Collecting and Tracking Memory Module's Field Performance Data In a Personal Computer

The disclosed method provides a simple and effective way for collecting and tracking memory module's field performance data

in today's Personal Computers. Data is stored in a non-volatile area of the memory module itself, independent to the system

hardware and error log. That means that the data stays intact with the memory once it's replaced or removed from the system.

Years of field data indicates that over 80 percent of the memory modules returned from the field are NDF (no defect found).

Once a suspected defective memory module is removed from the failing system, probability of repeating the same failure in a similar

system or on a memory module tester is often difficult, without knowing the nature and condition of the failure. This is mainly caused

by the following problems:

1. Marginal hardware design on part of the system, or the memory module, or interaction between the system and the

memory module.

2. Poor FRU (Field Replacement Unit) isolation on the system part.

* 3. Mis-pulling (Arbitrary replacement of parts) by the on-site service personnel.

* 4. Data is loss once memory module is removed from the system.

5. Random failures caused by DRAM soft errors.

As a result, these NDF memory will be recycled back to the general memory population, until it fails again, and again! This

paper provides a solution to address problems *3 and *4.

Majority of the Intel platform Personal Computers and Servers today use PC100 or PC133 standard memory; by design, this class

2 of memory provides 128-bytes of non-volatile memory user space, that is accessible via IC bus (Inter Integrated Circuit Bus).

This paper describes the method of using this free space to implement a defined data structure, providing a mechanism to track field

performance data, as well as performing failure re-creation, failure analysis, trouble-shooting and design enhancement.

First, the 128-byte user space on each memory module is partitioned into 3 blocks:

1. Block number 1 has two entries for storing memory related failures in the Manufacturing Test Process; each entry is subdivided

into 3 field:

- System Serial Number, 7 bytes long. (Example: 23MM105) - Manufacturing memory error code, 5 bytes long. (Example: 00000 (no error) - Date of test, 8 bytes long. (Example: 05/14/99)

2. Block number 2 is reserved for encoding or separator (can be used for authentic memory verification). The field is 16-byte long.

3. Block number 2 has 3 entries for storing memory related failures as reported by Diagnostics test program and Service Processor

(if installed); each entry is subdivided into 3 fields:

- System Serial Numbe...