Problema Memoria

Check the RSA logs to see if there are any PFA errors, single bit errors (Memory ProteXion Events on xSeries 440 and 445), or Excessive or Correctable Errors logged. Any of these errors should warrant a service call for FRU replacement. Check the POST error log for error message 289. If the DIMM was disabled by a system-management interrupt (SMI), replace the DIMM. If the DIMM was disabled by the user or by POST, proceed to step 9.

To enhance reliability, servers use ECC (Error Checking and Correcting) memory and Chipkill. ECC memory is fault tolerant and single bit errors can be corrected without shutting down the server. Chipkill corrects multiple single bit errors. If a memory chip error occurs, Chipkill will automatically take the failed memory chip offline while the server continues to run. Chipkill provides protection for memory similar to RAID protection for disks. Data bits and parity are written across multiple chips on the DIMM, and the controller is able to reconstruct a missing bit from a failed chip and continue working. Chipkill support is provided by the memory controller and is implemented using standard ECC DIMMs

Check for POST numeric or beep codes at system startup. A startup beep code of 1-3 typically indicates a memory problem. If you are receiving an error at bootup, read the rest of the steps in this document, then refer to Troubleshooting bootup issues.

Make sure your system is at the latest BIOS, Systems Management firmware, and diagnostics. These updates ensure that new or added components are recognized and used correctly by the system.

Download latest BIOS and support files for your system

For example:

xSeries 365 - May not boot if DIMM pair J1 and J5 is not installed, or if BIOS not at latest level

xSeries 365 - DIMMS disabled with downlevel BIOS

xSeries 200 - latest BIOS required to recognize 512MB DIMM

Check the IBM ServerProven compatibility Web site to ensure that the memory has been tested and certified as being supported for your system. Most third party memory does not meet the stringent performance and quality guidelines required by IBM, and thus is not supported in IBM systems.

SDRAM memory in a server can be either registered or unbuffered. The two types cannot be mixed. In registered memory, the registers provide a controlled delay in communication with the chipset. This delay is one clock cycle. This ensures that the clock edge collects all the data, and ensures that the data is accurate and complete in heavily loaded memory systems. In unbuffered memory, the chipset controller deals directly with the memory. No registers are between the chipset and the memory as they communicate with each other.

Using incompatible memory is the most common reason memory upgrades do not work. While incompatible memory may work to some extent, it can lead to unpredictable system behavior including data corruption and poor system performance including blue screen stop errors, NMI errors, and intermittent lockups.

Memory installed after receipt of the system should be verified as fully compatible with the system. It is important to make sure that you are using the correct part numbers when selecting computer memory. Any third party memory must be a direct replacement for the IBM option part number or replacement (FRU) number.

When adding memory to a system that uses interleaving, the memory modules must be added in sets of two. Memory interleaving enables memory banks to be accessed simultaneously rather than sequentially. Interleaving can only occur between identical memory modules. Banks of memory are combined and organized into even and odd bytes. In this arrangement, a data request can be made to one bank, and while that request is pending, a second request can be made to the other bank. Interleaving access to memory increases system performance, because read and write activity occurs nearly simultaneously across multiple memory modules.

Reseat the memory modules firmly in their slots to ensure a firm and proper connection. Memory and other components frequently become dislodged when other work is done inside the system or when the system is moved from one location to another.

request to check more information with the link

https://www.ibm.com/support/pages/troubleshooting-memory-issues