Cisco 6506E Supervisor Engine with bad RAM

[tc0fh]

Hello from Romania!

I have two Cisco 6506E switches with VS-S720-10G-3C supervisor engines that crash during boot. From the crash output, it looks like I'm hitting the problem described here: https://www.cisco.com/c/en/us/support/docs/field-notices/637/fn63743.html

Cisco says that some network equipment they sold has bad memory chips form a single manufacturer between 2005 and 2010. On the board in question there are two different memory chips: MT 46V16M16-6T (date code 0950) and Samsung K7N163631B-PC16 (date code 913).

From Cisco's 2019 anual report https://www.cisco.com/c/dam/en_us/about/annual-report/cisco-annual-report-2019.pdf

For example, in the second quarter of fiscal 2017 we recorded a charge to product cost of sales of $125 million related to the expected remediation costs for anticipated failures in future periods of a widely used component sourced from a third party which is included in several of our products, and in the second quarter of fiscal 2014 we recorded a pre-tax charge of $655 million related to the expected remediation costs for certain products sold in prior fiscal years containing memory components manufactured by a single supplier between 2005 and 2010.

My question for the eevblog readers is this: have you encountered similar issues with those RAM chips? I would like to know which one is the culprit in this case in order to avoid replacing the good ones. Who is the single supplier in question?

Thanks for your help!

[nuclearcat]

Probably issue is in proper power down sequence and device designers missed this moment or it was published too late in errata.
Something like this: (page 50 of MT42L128M32D1, LPDDR2)
"An uncontrolled power-off sequence can occur a maximum of 400 times over the life of the device."
As i understand, if power-off sequence is not done properly, latch-up(s?) happens and it burns slowly sensitive parts of chips.

[asis]

Hi,
I saw something similar with my own eyes in 2011. While inside the switching center, there was a power surge, UPS switched to batteries several times, but part of the network equipment still rebooted ...
Failure: WLC2106 Cisco Wi-Fi controller and switch Cisco C2960-48TC-L.
I read this bill:
https://www.cisco.com/c/en/us/support/docs/field-notices/637/fn63770.html
WLC2106 managed to exchange for a new one.
Switch Cisco C2960-48TC-L was not restored.
Moreover, in the course of life, other switch's began to come from other objects
(C2960G-24TCL, 2811..3500 ..).
About the RAM.
All the same, we are probably talking about NOR Flash-RAM , because the initialization by including by power ON.
There are no replicas on RS232 (ROMMON).
Make a clone using the programmer by reading the desoldered FRAM from the working device - it does not work out  because MAC address (S / N) is protected, there is a protection bit and there are areas of checksums.
In addition, at the POST stage, the MPU mask is checked, which is read from the OTP MPU area.
In my opinion, it’s not really possible to do anything.
 
Vladimir.

[tc0fh]

Hello!

Digging around point to the Micron as the bad memory.

On the PCB there are 46V16M16-6T (TSOP66, T_ck=6ns, CL=2.5) but I can't find the exact part. I can only find 46V16M16P-5B (TSOP66 LEAD FREE, T_ck=5ns, CL=3).

Can i use the -5B instead of -6T?

Thanks a lot!