| 14:08 | mhoney_work | Are there any standard linux distros that work with the cubox that don't need a custom compiled kernel? |
| 15:24 | rabeeh | moheny_work: you can use the CuBox installer to install few of those that already have a kernel |
| 15:45 | mhoney_work | Hi rabeeh: So I installed the debian hardfloat version with the installer, but where do I get the headers for the kernal? |
| 15:45 | mhoney_work | •eth00• Unable to locate package linux-headers-3.5.7-00735-g8dd8e33-dirty |
| 16:03 | ujee | Thanks for the cubox rabeeh, and thanks for the Xilka cbxbiker61, they are both awesome. I got a chance to compare Cubox with a board called "SabreLite". They were able to get it in my mailbox in ~1.5 days, but the documentation and community are nonexistant. (I expect I'd need ~6 months to figure the little missing details out). Cubox, on the other hand, took long to deliver, but it was nice and easy to setup. Thanks again. |
| 16:05 | rabeeh | ujee: nice to hear that. |
| 16:06 | rabeeh | did you buy the SabreLite from Freescale? The dual cpu or the single cpu? |
| 16:06 | ujee | the Quad one |
| 16:06 | ujee | (one cpu, 4 cores) |
| 16:06 | rabeeh | the imx6 version then |
| 16:07 | ujee | I got it from farnell, but they should be the same |
| 16:07 | rabeeh | with the screen? |
| 16:08 | rabeeh | i think there is a with screen and another without screen version |
| 16:08 | ujee | only the board, the cheapest model I think |
| 16:43 | dv_ | I have worked with the imx6 |
| 16:43 | dv_ | the version with the screen is the SD I think. it is also very expensive. |
| 16:44 | dv_ | the imx6 chip is very powerful, but can also get very hot. we did a stress test with 1080p video and a 3d scene rendering as fast as possible, and the chip reached 74°C |
| 16:45 | rabeeh | 74c where? |
| 16:45 | rabeeh | die? |
| 16:45 | dv_ | on the die, yes |
| 16:45 | rabeeh | ambient? |
| 16:45 | rabeeh | 74c on die is nothing |
| 16:45 | dv_ | just a few cm next to the die it already dropped to 55°C |
| 16:45 | rabeeh | is it the quad? or dual-lite? |
| 16:46 | rabeeh | is it a flip chip? |
| 16:46 | dv_ | quad, with one core active (the rest disabled) |
| 16:46 | dv_ | not sure about the flip chip. how do I recognize that? |
| 16:46 | dv_ | the die is in a metal casing |
| 16:48 | dv_ | its this one: http://at.farnell.com/freescale-semiconductor/mcimx6q-sdp/eval-i-mx-6-sabre-6quad/dp/2253177 |
| 16:49 | rabeeh | the quad is flip chip |
| 16:49 | rabeeh | flip chip is where you can see the die |
| 16:49 | rabeeh | bga or qfp is typically where you can't see |
| 16:50 | rabeeh | if there is a metal then it's probably a heat spreader on top of the die; if you remove it then you will see the die |
| 16:50 | dv_ | okay |
| 16:50 | rabeeh | http://www.flickr.com/photos/66086529@N08/8389836255/in/photostream/ |
| 16:50 | rabeeh | so the substrate is on top; so you should feel all the heat coming from the chip |
| 16:50 | dv_ | oh btw. rabeeh, is there a reason why the vivante graphics headers and libraries supplied by marvell are X11 only? in the older version included in the big tarball, there were some vdk headers, but they are essentially non-functional. also, in the current headers, there is a nativefb.h, which is useless, because the functions in it are only declared, not defined. |
| 16:51 | dv_ | ah |
| 16:51 | rabeeh | also the temperature junction where it's measured (74c) is really on top of the die |
| 16:51 | dv_ | yes |
| 16:52 | dv_ | i'm just being curious. 3d graphics work fine for me in X11. its odd though that an embedded platform wouldnt allow for non-X11, FB-only rendering |
| 16:54 | dv_ | also, the iMX6 also uses a vivante chip (the quad uses a GC2000). and there, both X11 and non-X11 rendering are possible, with the same set of libraries and headers (you just need to #define some macros). |
| 16:54 | rabeeh | dv_: there were some posts on the forums of people that were able to run directfb accelerated - right? |
| 16:54 | rabeeh | let me dig a big |
| 16:54 | dv_ | this isnt about directfb though |
| 16:55 | rabeeh | oh; the EGL on non-x11? |
| 16:55 | dv_ | yes |
| 16:57 | dv_ | one reason why I would be happy about a solution like on the iMX6 is that I have experienced problems with the xorg dove driver. I posted about it in the forums. I cannot set a resolution above 1024x768 in X. I can however use higher resolutions with the framebuffer. However, this might be because I am not using a "true" hdmi monitor, but one with a DVI connector, and a dvi-hdmi cable. |
| 16:59 | rabeeh | dv_: the hdmi transmitter is limited to certain frequencies in order to pass audio on it |
| 17:00 | rabeeh | the internal frame buffer with the clock generator are not limited. |
| 17:00 | rabeeh | there is a "trick" the NXP sent me to disable audio but whatever comes in the transmitter, is transmitted out |
| 17:00 | rabeeh | i can share this with you if you want. |
| 17:00 | dv_ | okay |
| 17:01 | rabeeh | this bypass mode can be enabled with the i2c_tools in Linux]\ |
| 17:01 | rabeeh | this bypass mode can be enabled with the i2c_tools in Linux |
| 17:01 | rabeeh | i mean, set with xrandr 1024x768@60/75 or whatever |
| 17:01 | rabeeh | you see nothing on the screen; then through i2c start changing registers inside the NXP to make it pass through mode (or bypass mode) |
| 17:01 | rabeeh | but this will disable audio support. |
| 17:01 | rabeeh | interested? |
| 17:01 | dv_ | okay |
| 17:02 | dv_ | I will anyway try it out with a proper hdmi monitor tomorrow, perhaps then the xorg driver works properly |
| 17:02 | dv_ | I might as well try this out too |
| 17:10 | rabeeh | dv_: http://pastebin.com/e0V6UVUL |
| 17:10 | rabeeh | that should open the HDMI transmitter (in DVI mode) to any resolution that is driven by the framebuffer. |
| 17:11 | dv_ | thanks |
| 17:11 | dv_ | I'll try these two things: 1) your trick 2) connecting a true hdmi monitor |
| 17:11 | dv_ | hopefully, the drivers arent broken at all, and I was just doing the wrong thing |
| 17:12 | dv_ | once this is done, I'll move on to the BMM and vMeta stuff (that is, integrate it in my yocto meta-cubox layer) |
| 17:31 | rabeeh | dv_: when you say it was hot; is it physically measured? or the temperature number? |
| 17:31 | dv_ | it was measured with a thermography |
| 17:32 | rabeeh | ok. fancy fancy :) |
| 17:32 | rabeeh | i have one here too :) |
| 17:32 | dv_ | ah btw, how hot does the cubox geht under maximum load? |
| 17:32 | dv_ | does it become a hot cube? :) |
| 17:33 | rabeeh | from the chip? |
| 17:33 | rabeeh | or the external box? |
| 17:33 | rabeeh | the external box max temp is beneath it; we measured ~42c |
| 17:33 | rabeeh | after ~1hr running 3d benchmark and the rest of the processor heavy L1 cache |
| 17:33 | rabeeh | with 25c ambient |
| 17:33 | dv_ | and the chip? |
| 17:34 | rabeeh | the chip die varies a lot from system to system. |
| 17:34 | rabeeh | it can go from 65c to 90c |
| 17:34 | rabeeh | but that's the internal of the internals of the die on the junction |
| 17:34 | rabeeh | if you measure that with thermography you would typically get 55-65c |
| 17:35 | rabeeh | thermography on the board. |
| 17:35 | rabeeh | it's pretty good design thermal wise. |
| 17:36 | rabeeh | notice that from spec point of view, all those chip starts to break on ~125c on the die (typically 85c outside) |
| 17:36 | dv_ | and what about its lifespan? isnt it shortened at high temperatures (but still <125c)? |
| 17:37 | dv_ | i think I remember that from some talks about CPU cooling |
| 17:37 | suihkulokki | lifespan should be within spec as long as temperature is within spec |
| 17:40 | rabeeh | 125 is within spec |