We're using custom-made SoM based on Olinuxino Micro design in our products.
Couple of months ago we purchased the next small batch of SOCs. They turned out to be mcimx233cag4b , not mcimx233cag4c we successfully (500+ boards) used before.
We see memory problems with these i.mx233 rev. B chips.
The problems manifest itself during memory testing using memtester (https://packages.debian.org/jessie/memtester ). The boards either
1) freeze during the test
2) or catch weird kernel OOOPS linked to NULL pointer derefence (which is clear sign of memory corruption)
Some of the boards (10% - 20% or so) pass this test fine. Some work worse than others, i.e. freeze during first seconds of test, some of the boards can last dozens of minutes before error occurs.
Again, we observed nothing like this in our previous batch, which was based on mcimx233cag4c. The PCBs and other components are identical (same batch).
The memory chips (Alliance AS4C32M16D) are the same as before. Changing memory chips to tested ones doesn't help. Changing Alliance memory to Hynix or Samsung doesn't help.
We also tried to play with memory settings, like changing CAS latency, frequency and so on - still nothing.
So, do you have any ideas what can possibly be wrong? Were there any changes related to DDR between rev. B and rev. C which are not covered in errata?