19:31 <hexdump0815> digetx: tagr: kwizart: i did an install of the latest jetpack image for my 2gb nano which updated the u-boot version - as a result no more deferred devices and the thermal sensors are working

19:33 <hexdump0815> this is the good part - that bad part is that the performance is quite bad: "7z b" only gives around 4200 as total score and should give more like 5200 if running at 1.4ghz and around 6000 or more if run at 1.9ghz which seems to be the case with mainline

19:34 <hexdump0815> both according to cpufreq and via mhz program, which detects the real frequency quite exact in my experience so far - so something must be wrong that the performance is so low - maybe memory timing? at least there are still some errors in the u-boot and dmesg output

19:34 <hexdump0815> this is the current dmesg now: https://pastebin.com/raw/uXj8MS3E

19:36 <hexdump0815> anyone any idea what is wrong?

19:37 <hexdump0815> thermal throttling is out of the picture as there is a string fan on the heatsink and it remains cool even under load

19:37 <hexdump0815> *strong

19:49 <hexdump0815> maybe this is part of the problem? - "WARNING: phandle mismatch was found in node 'external-memory-controller@7001b000' ... It will not be updated in the destination DTB."

19:51 <hexdump0815> and resulting - "tegra-mc 70019000.memory-controller: no memory timings for RAM code 1 registered" and "tegra210-emc 7001b000.external-memory-controller: failed to get nominal EMC table: -19"

19:52 <DavidHeidelberg> hexdump0815: what `make dtbs_check` says? (or `make something.dtb`)

20:12 <hexdump0815> DavidHeidelberg: here is the output for the dtb used - https://pastebin.com/raw/0zc9CBtH

20:14 <hexdump0815> but my naive and completely wild guess is more that the bootloader tries to write the memory data somewhere into the devicetree and finds a different structure there than it expects - maybe legacy vs. mainline differences? - all just very wild guessing

20:18 <hexdump0815> some more observations: my 1920x1080 monitor is detected as 1024x768 - worked well so far connected to anything else

20:18 <hexdump0815> plus power down does not actually power down the nano

20:29 <kwizart> hexdump0815, it's weird because I was expecting upstream kernel was relying on upstream u-boot (which latest version were expected to allow EMC memory training tables unless I'm wrong).

20:30 <kwizart> maybe we need to check diffs with upstream vs downstream dtb wrt thermal nodes