| 10:58 | shesselba | rabeeh: if we want to turn off SATA phy on CuBox, is there a way to detect eSATA hotplug event? e.g. is there some gpio/controller bit? |
| 11:14 | jnettlet | shesselba, there is no uevent that is broadcast? |
| 12:24 | rabeeh | jnettlet: i have been running the new ddr configuration for two days and runs perfectly fine |
| 12:24 | rabeeh | there is a major power reduction about 0.5W on the C1 board. I can send you my configuration. |
| 12:25 | rabeeh | just for the record; i'm running the same configuration on 500MHz clock rate (where the silicon officially supports 400MHz); it has been running for ~1hr; and keep it the whole day |
| 12:26 | shesselba | jnettlet: How should there be a software event, if sata phy is already powered down? |
| 12:28 | rabeeh | if sata phy is powered down there is no way you can detect presence of a drive |
| 12:28 | rabeeh | presence is detected only via SATA oob sequence (i.e. the differential line of TX AND RX are powered up) |
| 12:29 | rabeeh | there is a possibility to check presence every few seconds |
| 12:30 | dv_ | the c1 isnt even really out yet, and already people try to overclock it :P |
| 12:30 | shesselba | rabeeh: ok, thanks for the info |
| 12:30 | bencoh | :)) |
| 12:30 | rabeeh | dv_: for testing purposes my friend :) |
| 12:31 | rabeeh | dv_: but ddr overclocking should be easy for us since first we ship with 1300Mbps devices (i.e. 667MHz clock rate) and second our DDR topology is quite unique |
| 12:33 | dv_ | alright |
| 12:33 | dv_ | hm it would be interesting to re-run my transcoding experiments with this increased rate |
| 12:33 | dv_ | since I suspect memory transfer to be the main bottleneck |
| 12:37 | rabeeh | dv_: what is your CPU utilization on such a case? |
| 12:37 | rabeeh | i remember you said high numbers ! |
| 12:37 | dv_ | the CPU usage wasnt an issue actually |
| 12:37 | dv_ | I primarily measured the time it took to transcode |
| 12:38 | rabeeh | so it's either the dsp or the memory then. |
| 12:38 | dv_ | for a 30s h264 high profile input it took ~31s to transcode to h264 baseline |
| 12:38 | rabeeh | oh |
| 12:38 | dv_ | there was very little difference if I use baseline or high profile as input |
| 12:38 | rabeeh | the numbers changed then |
| 12:38 | dv_ | ? |
| 12:38 | rabeeh | it was 10 seconds content that goes to 12second |
| 12:38 | dv_ | ah that was the first test |
| 12:38 | dv_ | but that was too short |
| 12:38 | rabeeh | now 30 goes to 31; maybe getting the gst pipeline warming is the bottleneck? |
| 12:39 | dv_ | yeah but it seems to me it converges to around 100% |
| 12:39 | rabeeh | and are you sure you are running in benchmark mode? |
| 12:39 | dv_ | keep in mind though that this was 1080p data |
| 12:39 | dv_ | "benchmark mode" ? |
| 12:39 | rabeeh | i.e. the decoder is not syncing with the clock time |
| 12:39 | dv_ | I wasnt playing in real-time if thats what you mean |
| 12:40 | dv_ | oh. no it wasnt |
| 12:40 | rabeeh | ok. since 10 -> 12 and 30 ->31 are in the same ballpark |
| 12:40 | dv_ | yes. but if memory speed is increased, then I guess this will change |
| 12:40 | rabeeh | i.e. meaning the 1-2 seconds differences is just getting gst pipeline warm to start decoding and encoding |
| 12:41 | dv_ | I'm curious how it will behave on the i4pro |
| 12:41 | rabeeh | i'm sure there is a devmem command that you can run to make the ddr speed faster |
| 12:41 | rabeeh | dv_: even with the dual lite; if the DDR is the bottleneck then dual lite with 64bit should give another boost |
| 12:42 | rabeeh | dual and quad has native 533MHz DDR clock support which adds too |
| 12:42 | dv_ | whats the mem config of the c1? |
| 12:42 | rabeeh | dv_: i can send you that; i'm using freescale tool - |
| 12:42 | rabeeh | https://community.freescale.com/docs/DOC-96412 |
| 12:42 | rabeeh | i can send you my .xls and .inc files that provides you those. |
| 12:42 | rabee | 12:42 * rabeeh is sending |
| 12:43 | dv_ | alright. I wont be able to do anything until monday, but I will write it down. |
| 12:43 | rabeeh | ok. |
| 12:43 | dv_ | does the c1 have a 64 bit bus? |
| 12:43 | dv_ | or is it smaller? |
| 12:43 | rabeeh | anyhow; you need a special usb host to host cable to run those |
| 12:43 | rabeeh | c1 has 32bit ddr interface |
| 12:43 | dv_ | whew |
| 12:43 | dv_ | no wonder I see these figures |
| 12:43 | rabeeh | c1 + solo |
| 12:44 | rabeeh | :) |
| 12:44 | dv_ | that said, its actually nice to see that it is at 100% and not higher |
| 12:44 | dv_ | (with this hw config) |
| 12:44 | dv_ | this means that on i2 and i4 realtime transcoding should be doable |
| 12:45 | rabeeh | for the record; the USB host to host cable is a USB cable that has two host connectors on both sides but the +5V is wired through a 100Ohm resistor |
| 12:45 | dv_ | ah. is this a micro-A? |
| 12:45 | rabeeh | dv_: yes |
| 12:45 | rabeeh | no |
| 12:45 | rabeeh | simple host connector the you plug to your PC and another host connector that you plug to the upper USB port of the C1 |
| 12:46 | rabeeh | in this case you plug in the USB otg phy of the imx6 to the PC |
| 12:46 | dv_ | okay |
| 12:46 | rabeeh | the way to boot it is simply powering up C1 without the micro SD; imx6 will try to boot from micro SD and fail then it will auto switch to "USB downloader" mode |
| 12:47 | rabeeh | which is simply a USB HID device waiting for commands from a PC |
| 12:47 | dv_ | also speaking about booting, I was using a 1A power source before. could this be the reason why the c1 didnt start after rebooting sometimes? |
| 12:47 | rabeeh | the +5V of the PC should be never directly connected to the +5V of the C1 USB connector since it may damage |
| 12:47 | dv_ | (I had to wait a few minutes before retrying) |
| 12:47 | rabeeh | what we do is adding 100 Ohm in the middle to do some sort of current limiting |
| 12:48 | rabeeh | dv_: wife is waiting. Need to go |
| 12:48 | dv_ | alright, thanks. ttyl |
| 12:48 | rabeeh | C1 consumes ~600mA in full load |
| 12:48 | rabeeh | each USB host can consume up to 500mA |
| 12:49 | rabeeh | so meaning you USB power supply should be from 600mA( i.e. no USB hosts is powered) or 1600mA (600+500+500) |
| 12:49 | rabeeh | ttyl |
| 14:29 | jnettlet | rabeeh, do you want me to roll those ddr timings into u-boot? |
| 15:51 | elouazzani | does anyone know where can i buy a new adapter for my cubox |
| 15:51 | elouazzani | europe plug |
| 16:17 | dv505 | elouazzani: you can buy any universal adapter. Just make sure it support 5V and the current parity |
| 16:17 | dv505 | elouazzani: as an example: http://www.vivanco.com/Homepage/Products/Energy/Power-Supplies-und-Adapters/For-Cameras-und-Navigation-systems/vivanco-27827-ac-dc-power-adapter-for-cameras-and-navigation-systems.html |
| 16:18 | elouazzani | ok dv505 Thanku |
| 16:21 | dv505 | s/current parity/correct parity. + on the inside |
| 16:21 | dv505 | normal adapter is up to 2A, which not a lot of universals support |
| 16:21 | dv505 | but you would only need that if you attach USB devices |
| 16:21 | dv505 | without USB, even .5A might do |
| 16:31 | elouazzani | ok that means 2000mAh right? |
| 16:49 | dv505 | yes |
| 18:06 | elouazzani_ | dv505: thank u |
| 21:12 | svere | rabeeh: i eventually got my C1! :-)))) |
| 21:19 | svere | just for confirmation: 1) I do not need a heatsink and 2) a 5V 2A power supply is sufficient for the C1!??!? |
| 21:54 | Bluerise | svere: You think it's not enough? |
| 22:19 | jnettlet | svere, yes and yes |
| 22:24 | svere | Bluerise: i think more than enough, but better make sure |
| 22:24 | svere | reboot... back in a minute |
| 22:43 | sver | 22:43 * svere cries as he can not find a micro-usb power-supply |