| 10:19 | rabeeh | ok. things config failed on 566MHz |
| 10:19 | rabeeh | that's about 40% more than it was configured for. |
| 10:28 | rabee | 10:28 * rabeeh reruns with configuration optimized for 533MHz DDR as a starting point. |
| 10:53 | rabeeh | reached 600MHz and still running |
| 11:00 | rabeeh | reached 621MHz and still running |
| 11:04 | suihkulokki | ricing ;) |
| 11:06 | rabeeh | didn't fail yet. passed 635MHz :) |
| 11:11 | rabeeh | ok. failed now on 654 MHz |
| 11:12 | rabeeh | this means on 400MHz we have about 60% up to 648MHz for overclockers to play with :) |
| 11:17 | _rmk_ | rabeeh: so why did it fail at 566MHz? |
| 11:18 | _rmk_ | rabeeh: also, how are you testing it? |
| 11:21 | rabeeh | the first configuration that failed on 566mhz is because it was configured to run at 400MHz (all the cas latency and other parameters) and i was able to stretch it to 566MHz |
| 11:22 | rabeeh | _rmk_: the second configuration was for 528MHz and I was able to stretch it to 648MHz |
| 11:23 | rabeeh | so for overclockers we have a huge place to work with |
| 11:23 | rabeeh | it will help a bit on the processor side; but the huge difference will be on the GPU and VPU |
| 11:24 | rabeeh | mainly on the GPU since in xbmc main gui with dirty regions disable i measured around 2GB/sec transfer which is almost the maximum phyiscal limitation of DDR 32bit when running at 400MHz clock rate (800Mbps) |
| 11:24 | jnettlet | rabeeh, how much extra power are you chewing up in those configurations? |
| 11:25 | rabeeh | jnettlet: from 400MHz to 648MHz the addition was 288mW |
| 11:25 | jnettlet | not too bad |
| 11:25 | rabeeh | sorry. 388mW |
| 11:26 | rabeeh | if you run at 400MHz and stretch it out; this is pure GPU performance |
| 11:26 | rabeeh | i mean this is pure performance addition that increments to fps |
| 11:27 | jnettlet | does xbmc run the fps uncapped? |
| 11:28 | rabeeh | yes |
| 11:28 | rabeeh | they have some cap in low power mode |
| 11:30 | jnettlet | too bad the ipu couldn't handling 1920x1080 then we could have xbmc render to a lower res and scale it up |
| 11:32 | rabeeh | jnettlet: that's what most 32bit ddr based xbmc machines do |
| 11:33 | jnettlet | makes sense. |
| 11:33 | rabeeh | we have tried that also on first CuBox; but the upscaler was so crappy |
| 11:47 | _rmk_ | rabeeh: you're talking about the upscaler in the LCD controller? |
| 11:47 | rabeeh | yes. |
| 11:48 | _rmk_ | hmm, I find that it fine if you turn on vertical interpolation |
| 11:48 | _rmk_ | I think the framebuffer driver doesn't (or didn't) |
| 11:48 | rabeeh | it has both linear and bilnear interpolation; but they didn't look good enough |
| 11:50 | jnettlet | how much are you scaling also? Large scales may require dual pass scaling. |
| 11:50 | _rmk_ | well, it is better than watching broadcast SD upscaled HD |
| 11:56 | jnettlet | rabeeh, should I put the new ddr config you sent out yesterday into u-boot? |
| 12:00 | rabeeh | 720p to 1080p |
| 12:01 | rabeeh | jnettlet: yes please |
| 12:01 | rabeeh | if you don't have time i'll take it |
| 12:01 | jnettlet | I can probably get it done this weekend. |
| 12:02 | rabeeh | jnettlet: if you would like to spend some time; i think spl is more important |
| 12:03 | rabeeh | i'll take the ddr configuration (and prepare and ifdef for overclockers) |
| 12:03 | jnettlet | okay I will poke at that a bit later. |
| 17:18 | rabeeh | jnettlet: i have the patch ready on your u-boot |
| 17:18 | rabeeh | i'll push it through github |
| 17:25 | hste | rabeeh: got the device. thanks. Havn't had time to play with it yet. Can't find my usb-sdcaBut it looks cool. |
| 17:26 | hste | rabeeh: is it oc-friendly without heatsink? |
| 17:27 | rabeeh | yes. |
| 17:27 | rabeeh | should work without issues |
| 17:28 | rabeeh | the product cubox-i1 power is even lower :) |
| 17:29 | hste | :) my quad-dongle got hot on heavy gpu work. cpu at 1.2Ghz was no problem but heavy gpu usage I got temp warnings |
| 19:37 | rabeeh | hste: did you upgrade the heatsink on the processor? |
| 19:38 | hste | rabeeh: on my gk802 I addede |
| 19:38 | hste | rabeeh: on my gk802 I added a heatsink |
| 19:39 | rabeeh | hste: have a picture? |
| 19:42 | hste | rabeeh: https://drive.google.com/file/d/0B7RFDJW3q4H9YzFVaWtCMDlnYVk/edit?usp=sharing |
| 19:43 | rabeeh | hste: is it really touching the imx? |
| 19:44 | rabeeh | do you have thermal grease? |
| 19:44 | hste | rabeeh: yes but I kept the foam tape between |
| 19:44 | rabeeh | hehe |
| 19:44 | rabeeh | that foam is an insulator |
| 19:45 | rabeeh | you need minimum thermal resistance |
| 19:45 | rabeeh | try putting the heatsink directly over the imx6 |
| 19:45 | rabeeh | if there is a short then put another metal in between and make sure thermal greese is really filled |
| 19:46 | rabeeh | or you can buy online thermal pads that are electrically isolators but conduct thermal |
| 19:46 | rabeeh | (typically in the z-axes; which is where you want it to conduct) |
| 19:47 | hste | yes I know :) |
| 19:48 | hste | but it was just an easy hack and way better than it was before mod :) |
| 19:49 | hste | before it ws just the foam that connected to the case... |
| 19:50 | jas-hacks_ | rabeeh: how was ddr3 testing? manage to reduce heat dissipation |
| 19:51 | hste | rabeeh: does the solo run under 1.2Ghz ? |
| 19:52 | rabeeh | jas-hacks_: depends how you measure and which benchmark |
| 19:52 | rabeeh | hste: i run on 1ghz |
| 19:52 | rabeeh | i haven't overclocked the processor |
| 19:52 | rabeeh | only memory :) |
| 19:54 | jas-hacks_ | can imx6q be over-clocked on ddr? |
| 19:54 | rabeeh | i haven't tried that yet. |
| 19:54 | rabeeh | imx6q clock is already 533Mhz |
| 19:55 | hste | rabeeh: how high have u clocked it stable? |
| 19:55 | rabeeh | jas-hacks_: any idea how to change screen resolution on the ubuntu image you have sent? |
| 19:55 | rabeeh | it doesn't look like xrandr is working |
| 19:56 | jas-hacks_ | only way is to pass value in uboot |
| 19:56 | rabeeh | hste: around 650mhz clock rate |
| 19:56 | jas-hacks_ | using boot.scr |
| 19:58 | jas-hacks_ | gk802 used samsung memory, so interesting to see how they perform in cubox-i |
| 19:59 | hste | rabeeh: do u think 650 would be stable on most of the c1 devices? |
| 20:00 | rabeeh | hste: no |
| 20:01 | rabeeh | jas-hacks_: do you have exact model? |
| 20:01 | rabeeh | we are using samsung too (not on the C1 you have) |
| 20:01 | rabeeh | the C1 you have is hynix but we switched to samsung |
| 20:01 | rabeeh | (production version) |
| 20:02 | hste | ok |
| 20:03 | jas-hacks_ | K4B2G1646C-HCH9 DDR3-1333 |
| 20:08 | jas-hacks_ | there were/are lots of thermal problems with gk802 but I think was more related to the CPU model |
| 20:09 | rabeeh | what do you mean? |
| 20:09 | rabeeh | you mean SoC version? |
| 20:11 | rabeeh | jas-hacks_: those are crappy ddr chips too |
| 20:11 | rabeeh | idd7 is ~250mA |
| 20:12 | jas-hacks_ | Soc version |
| 22:05 | _rmk | 22:05 * _rmk_ wonders what the maximum temp of the imx6s is |
| 22:19 | jas-hacks_ | i think it's 105 |
| 22:20 | jas-hacks_ | but kernel should stop it going over 90 |
| 22:20 | _rmk_ | only if you have a kernel which can do throttling which mine can't |
| 22:21 | _rmk_ | it sits around 75C while rhythmbox runs (with no gpu) |
| 22:21 | jas-hacks_ | it should get hotter through GPU throttling |