[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference wonder::turbolaser

Title:	TurboLaser Notesfile - AlphaServer 8200 and 8400 systems
Notice:	Welcome to WONDER::TURBOLASER in it's new homeshortly
Moderator:	LANDO::DROBNER

Created:	Tue Dec 20 1994
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	1218
Total number of notes:	4645

1185.0. "Memory POST test results differ 350 vs 440 Mhz CPU" by GIDDAY::FLAWN () Fri Apr 25 1997 10:08

Hi,

A customer's 5/350Mhz 8200 system was showing bad memory (at power up, and 
with simm_callout on a failing SIMM was shown). When the memory module was 
moved to a 440Mhz system also on the customer site the same module didn't
show the fault, though it was verified (by moving the SIMM on the module
and reinstalling on the 350Mhz system) that a SIMM was at fault.

This came about because of reconfiguration of the system (which ended up 
taking longer than expected as a consequence) and the customer is interested in
why this would be the case.

Both systems run 4.1 console.

Can anyone indicate if there are any specific CPU model dependencies involved 
in the memory tests ? I assume the test which detects this sort of failure is
a march test across all memory so I would have expected the same results.

Regards and thanks,
Dave Flawn
CSC Sydney

T.R Title User Personal
Name Date Lines

1185.1 DANGER::HARTWELL Fri Apr 25 1997 11:32 26

T.R	Title	User	Personal Name	Date	Lines
1185.1		DANGER::HARTWELL		`Fri Apr 25 1997 11:32`	26
	- Was the memory module at fault located in the same slot with the exact same memory and processor configuration between the 350 and 440? If the 2 systems were not EXACTLY the same it is possible that self-test would not pick up the failure due to a specific data pattern/patterns that cause the failure. If you can state that these two systems were exactly the same then it needs to be looked at. Remember self-test is a quick verify test, not a all inclusive diagnostic. It is designed to run fast and initialize the module with good ECC. It is not designed to find all faults. Why?, because no customer would be willing to wait the required time needed to fully test a memory module upon each power-up/reset. Just to run a single pass of a memory pattern such as Read data, write 1's, read 1's, write 0's read 0's would require about 2.6 minutes per each 4 gig option. If your customer had a 16 gig system, your looking at over 10 minutes to execute a simple all 0's all 1's pattern. Hope this helps /Dave

    
    - Was the memory module at fault located in the same slot with the
      exact same memory and processor configuration between the 350 and
      440? If the 2 systems were not EXACTLY the same it is possible that
      self-test would not pick up the failure due to a specific data
      pattern/patterns that cause the failure. If you can state that these
      two systems were exactly the same then it needs to be looked at.
    
      Remember self-test is a quick verify test, not a all inclusive
      diagnostic. It is designed to run fast and initialize the module
      with good ECC. It is not designed to find all faults. Why?, because
      no customer would be willing to wait the required time needed to
      fully test a memory module upon each power-up/reset. Just to run
      a single pass of a memory pattern such as Read data, write 1's, 
      read 1's, write 0's read 0's would require about 2.6 minutes per
      each 4 gig option. If your customer had a 16 gig system, your
      looking at over 10 minutes to execute a simple all 0's all 1's
      pattern. 
    
    
    
      Hope this helps
    
    
    						/Dave