While doing climate testing of our i.MX8QXP product, some issues have come up.
When SoC reports temperature above +85°C we started seeing glitches on the screen, and after some time Weston and/or any running application crashes with segfault.
We are running SCFW porting kit 1.6.0 with DCD Cfg settings generated from latest v15 of DDR Configuration spreadsheet, and linux-fslc-imx 5.4.74.
CPU revision is B0. Running 2GB LPDDR4 at 1200 MHz.
We have run DDR stress test tool in +90°C surrounding temperature at 1200-1300 MHz without issues. So the memory bus itself seems stable.
What does seem to solve the issue is to select "Apply maximum de-rate timings" in the configuration tool, then the board seems to be working fine even at +95°C SoC temperature.
We tried only disabling the automatic de-rate feature, but that made no improvement. Still the same glitching and crashing, only that the board no longer booted at high temperature.
What are the drawbacks of using maximum de-rate timings? Higher current consumption?
Also, are there any settings which can be tweaked for the de-rating? Since using maximum values at all times solves the problem, I'd guess that more de-rating would also work.
We have set T_RFC_NOM_X32 to 3904 as per the SDRAM datasheet nominal tREFiab. But of course the datasheet specifies much lower values for temperature above +85°C. Shouldn't this be considered anywhere?
Could one assume that this issue will be gone when moving on to C0 silicon?
