| 15:31 | Ke> | jnettlet[m]: btw. puri.sm claims to be shipping iMX 8M devkits https://puri.sm/posts/june-1st-last-call-for-librem5-devkit/ |
| 15:32 | Ke> | jnettlet[m]: btw. any chance of getting more ram for the new cubox |
| 15:49 | jnettlet[m]> | Ke: we will support 4GB, but not right away |
| 15:51 | jnettlet[m]> | right now RAM prices are ridiculously so the cost really stacks up. |
| 15:52 | jnettlet[m]> | I am a bit concerned about their choice of the iMX8M...it is definitely not the lowest power of chips. |
| 15:52 | bencoh> | well, iMX8X hasn't been released yet, so ... |
| 15:53 | Ke> | I have a phone with quad a53, works ok |
| 15:53 | bencoh> | (and at the time they chose imx8m, X wasn't even in preprod phase) |
| 15:53 | jnettlet[m]> | they would be smart to implement an eDP screen with panel self refresh support |
| 15:53 | jnettlet[m]> | I am also wondering how they are doing power over USB-C, since the SOC's implementation doesn't support it. They must be using a third party arbiter |
| 15:54 | jnettlet[m]> | the iMX8X is even higher powered |
| 15:54 | jnettlet[m]> | the performance of the iMX8M is quite nice so far. But you are talking about a device up in the 6-8W range |
| 15:54 | bencoh> | the question is "how does it behave when idling" |
| 15:55 | jnettlet[m]> | definitely not even close to the same power numbers as the iMX6 |
| 15:55 | bencoh> | seriously? :/ |
| 15:55 | jnettlet[m]> | although I still need to test the latest kernel which does have some power savings improvements according to NXP |
| 15:55 | bencoh> | and there I thought imx6 was really bad at idling |
| 15:56 | bencoh> | (compared to imx7, imx6 is pretty bad, afai could tell) |
| 15:56 | bencoh> | (when talking about idle/suspend) |
| 15:57 | jnettlet[m]> | depends what you are doing. like I posted the iMX6 is idling with about 80-90mA power draw...no screen |
| 15:57 | jnettlet[m]> | suspend I have an iMX6S on a HB-Pro with the power LED removed and my Suspend to RAM is 19mA |
| 15:58 | Ke> | they actually said they couldn't use iMX 6 due to power use |
| 15:58 | bencoh> | we reached 10~25mA idling here with no screen on imx7 |
| 15:58 | jnettlet[m]> | idle on that board is under half a watt |
| 15:58 | bencoh> | (10~25mA @3.7V) |
| 15:59 | jnettlet[m]> | with wifi and bt on? |
| 15:59 | bencoh> | no, wifi/bt is much more needy here |
| 16:00 | bencoh> | once the wifi driver is inserted (or once the wifi interface is up, don't remember), core/bus start working at a higher freq |
| 16:00 | jnettlet[m]> | yeah with the wilink8 all my numbers are with wifi/bt enabled |
| 16:00 | bencoh> | so it'll reach 60~80 |
| 16:02 | jnettlet[m]> | so actually right about where the iMX6S/DL is right now. |
| 16:02 | bencoh> | if what you're saying about 8m is true, then I'm a bit disappointed |
| 16:03 | bencoh> | the true difference between 6s/dl and 7/7d is the possibility to run at really low core/bus freq (but then you can't much of the peripherals, if any) |
| 16:05 | bencoh> | (and the fact that they separated more clocks / clock sources from each other) |
| 16:05 | bencoh> | (as far as I remember) |
| 16:07 | jnettlet[m]> | well and no GPU |
| 16:07 | jnettlet[m]> | just the PXP engine |
| 16:09 | bencoh> | yeah |
| 16:10 | bencoh> | right, we didn't use it anyway, so ... :) |
| 16:13 | jnettlet[m]> | although I have been playing around with the TI configuration options. They have a power saving AP mode where you can limit the broadcast duty cycle. So 50% etc. I have't seen how much it shaves off the power consumption when idle |
| 16:14 | bencoh> | TI? |
| 16:22 | jnettlet[m]> | Texas Instruments |
| 16:23 | bencoh> | yeah, I just didn't understand why you were referring to TI here ... which component is from TI? |
| 16:24 | bencoh> | (the wifi module?) |
| 16:33 | jnettlet[m]> | the Wifi/BT |
| 16:33 | jnettlet[m]> | our Rev 1.5 som switched to the Wilink8 |
| 16:34 | bencoh> | ah |
| 17:22 | Ke> | now I got wired internet back, perhaps I could try the rmk's tree again |
| 17:25 | Ke> | definitely linux-4.16 (debian) does not allow 1G connection to work even with ethtool setting -s eth1 speed 1000 duplex full autoneg off |
| 17:36 | jnettlet[m]> | Ke: yeah bootlin did a fairly good job breaking Armada support in the 4.16 release. |
| 17:36 | Ke> | perhaps, has not worked for me ever |
| 17:36 | Ke> | but only tested twice |
| 17:38 | jnettlet[m]> | well rmk is still waiting for Bootlin to finish the patch series that they claim they have, so he doesn't have to keep rebasing his work. |
| 17:38 | jnettlet[m]> | it has only taken 2 years to this point, what is another 2-3 releases |
| 17:50 | Ke> | but only tested twice |
| 17:50 | Ke> | sorry |