LX2162A DDR max speed

Hey!
I’m using LX2162A with ClearFog dev board.
According to datasheet maximum DDR speed for LX2162 is 2900


Using DDR 3200 causes problems with our test software - Linux freezes when using 3200. Downgrading to DDR speed 2600 resolved all problems.
Is there any reason for providing prebuilt images with DDR speed 3200 for LX2162 (could be found here SolidRun Images)?

1 Like

I am checking with the bring up team regarding this discrepancy. I know that our current product and BSP have passed all our internal burn-in and validity tests with the existing BSP settings.

Thank you for reporting this!

The default settings were an oversight and based on lx2160a com-express + sdram module characteristics.

We have reduced both the ddr and bus clocks according to lx2162 datasheet for our default images:
https://images.solid-run.com/LX2k/lx2160a_build/20240927-97d77e3

Further please confirm whether your development board has 8GB or 16GB of RAM.

In case of 16 we found wrong timings in the SPD eeprom.
If you are feeling adventurous you may use the attached script matching your memory size, to test an updated configuration. It has not yet been validated under computational and thermal load.
In case of errors however you are responsible for unbricking your device, we will not RMA it.

lx2162-som-v11-8gb-spd.sh.txt (1.9 KB)
lx2162-som-v11-16gb-spd.sh.txt (1.9 KB)

In case the SPD got corrupted, a special image ignoring SPD altogether may be booted from sd-card:
lx2160acex7_2000_700_3200_LX2162A_CLEARFOG_18_9_0-2a7ab21_sc08g_draft2.img.xz
lx2160acex7_2000_700_3200_LX2162A_CLEARFOG_18_9_0-2a7ab21_sc16g_draft2.img.xz

I have just noticed that 3200 images no longer exist, and found this post. Is it confirmed that using a DDR speed of 3200 causes issues? I have been using Honeycomb with 3200 DDR4 as my primary development system for QNX for the last couple of years without issues (though there have been some anomalies). We have also deployed a couple of dozen machines for testing.

–Elad

At SolidRun we have not seen any failures due to using the 3200MHz memory speeds. The change was purely to keep in line with NXP’s documentation and recommendations. In general you will not see much of a performance difference in switching to 2900 from 3200 if you are concerned, maybe a 100-200MB/s peak single core throughput.