Anyone noticing a slower UI compared to the TX2? (Solved)

Hi there,
just got time to check out the Xavier, didn’t reflash it so my L4T is version 31.0.0, I saw that there is a 31.0.1 in the download page, will probably reflash this afternoon.
I noticed the UI is way slower than the TX2, could that be due to having wayland was wm? Even when enabling jetson_clocks seems a bit slower than in the TX2.
Any thoughts?

I’ll update this once I reflash again.

yes, only got my Xavier yesterday and have re flashed it, but i agree, it feels more sluggish than my TX2, might be since it’s a early version of the Jetpack 4.0 ?

But i did compile a kernel on it last night, and that seems to have runned in about the same time as on the tx2 :)

18.04 uses a different UI than what comes with the current jetpack. If you boot up an xavier before you flash it with jetpack you will see it. Must be still working on it.

Ah, ok, nice to know @danpollock, i went straight to flashing so didn’t check it first :)

I always start them up without flashing just in case :)

I always start them up without flashing just in case :)

The main reason could be the default core enablement. Only 4 cores are enabled and can only run at 1.2G HZ and GPU running at 670M HZ. It did feel more sluggish for me.

The default power mode is 2. You can use nvpmodel command to change to power mode 0 which gives you full power of Xavier.

sudo nvpmodel -m 0

You can change back to 2 or other models at any time.

i will make a note of that until next time, lol

Doesn’t enable all cores at full ?

There is a table of allowed clock ranges, and also a table for enable/disable of cores. nvpmodel modifies the tables. For any given table, sets to the highest performance. If the table was designed to keep within a 15W power range, then will result in slower clocks than if nvpmodel was told to go ahead and allow 30W. “nvpmodel -m 0” enables Denver cores and sets the highest possible allowed clock; followed by “”, this is the highest performance you will get.

Ahh, very nice info linuxdev, thanks :)

Awesome, indeed after the reflashing and the nvpmodel the performance is way better. Thanks linuxdev!