| 10:13 | topi`> | jnettlet[m]: what kind of config do you have for hostapd to get 500Mbit/s rx/tx speeds? you'd have to enable VHT flags to get 802.11ac speeds |
| 10:14 | topi`> | thus, you need channels 36-60 and probably 40mhz/80mhz wide channels |
| 10:14 | topi`> | on my HB, the bloody driver/firmware/something prohibits the use of channels 36-60 |
| 10:14 | topi`> | so I only run it at 2.4ghz |
| 10:21 | jnettlet[m]> | topi`: one second let me grab it. |
| 10:22 | jnettlet[m]> | You need to make sure and set the region to EU. I |
| 10:24 | jnettlet[m]> | topi`: https://pastebin.com/iR7ENNS9 |
| 10:26 | topi`> | that's probably my pain point. The Compex card reports World domain |
| 10:26 | jnettlet[m]> | most likely yes |
| 10:26 | topi`> | so no matter what I try to do at user-space, it keeps on disabling those channels |
| 10:26 | jnettlet[m]> | topi`: fyi, that patch while it works, skips over all the latency work that has been done. |
| 10:26 | topi`> | also channels 12 and 13 are disabled, which are definitely in use in EU |
| 10:26 | jnettlet[m]> | I am trying to sort out the proper patch to fix everything |
| 10:27 | topi`> | I saw some transmit speeds of 130Mbit/s while running my WLE600VX in 802.11n mode (using 20mhz channels) |
| 10:27 | jnettlet[m]> | that sounds about right |
| 10:28 | topi`> | so for some reason, the 4.13 kernel was not affected |
| 10:28 | jnettlet[m]> | as long as you have pretty clean 2.4Ghz airwaves around you. |
| 10:28 | topi`> | lots of other problems, though. Like a deadlock issued immediately every time the first connection comes in |
| 10:28 | topi`> | oh, we have no neighbors:) |
| 10:28 | jnettlet[m]> | no you can get 130Mbits per second TX that is about what I would max out at. |
| 10:29 | jnettlet[m]> | you will probably get more with the patch |
| 10:29 | topi`> | but the Deadlock also causes some pain down the lane, at some point the whole Hummingboard freezes |
| 10:29 | jnettlet[m]> | also I recommend disabling cubic tcp congestion algorithm. There is another bug that effects that. Switch to reno or westwood (if you have lossy connections) |
| 10:29 | topi`> | and, what is most irritating, the RESET button (that, right down on the board) stops working then |
| 10:29 | topi`> | it literally does nothing |
| 10:30 | topi`> | works fine after I have power-recycled the board |
| 10:30 | jnettlet[m]> | which generally means the SDHC card has crashed |
| 10:30 | jnettlet[m]> | yep |
| 10:30 | topi`> | I run the board from eMMC |
| 10:30 | topi`> | maybe that 4.13 is still too unstable |
| 10:31 | topi`> | which kernel(s) have you had a stab at? |
| 10:31 | topi`> | 4.9.x seem to be fine |
| 10:31 | jnettlet[m]> | 4.9 is what I am working with currently. I will test 4.13 soon, getting ready for the 4.14 LTS |
| 10:32 | topi`> | I recently rebased from 4.9.21 to 4.9.31 |
| 10:32 | topi`> | I've been testing it |
| 11:31 | topi`> | this is what I get on dmesg: |
| 11:31 | topi`> | [ 77.563092] ath: EEPROM regdomain: 0x0 |
| 11:31 | topi`> | [ 77.563106] ath: EEPROM indicates default country code should be used |
| 11:31 | topi`> | so I guess the EEPROM needs fixing |
| 11:44 | jnettlet[m]> | topi`: check further down. it may come up as that and then program the country code |
| 11:46 | topi`> | I'm pretty sure it is still dumbed down as World regd, since hostapd can't use any of those 5ghz channels |
| 11:46 | topi`> | and normally hostapd can use channels 12&13, but not with this wifi card. |
| 11:51 | Ke | 11:51 * Ke puppy eyes jnettlet[m] for the cooling solution |
| 11:51 | Ke> | not urgent though |
| 11:52 | Ke> | topi`: what did you use for mbin cooling btw? |
| 13:26 | topi`> | Ke: a 12cm fan. overkill yeah ;) |
| 13:27 | topi`> | the heatsink works fine as long as running under 1.3ghz and not using all 4 cores at once |
| 13:27 | topi`> | so make -j3 was stable, make -j4 was not |
| 13:27 | Ke> | huh |
| 13:28 | Ke> | I was hoping to be able to run full load at 2GHz indefinitely |
| 13:28 | Ke> | obviously not a typical workload, but almost all my systems are self hosting when it comes to kernel |
| 13:29 | Ke> | I guess I could consider the water cooling, if air does not work out |
| 13:34 | topi`> | wow, hostapd started with jnettlet's config file. Excluding some odd HT_* and br_lan config items |
| 13:35 | topi`> | Ke: it seems to me the 8040 does not consume significantly more power when at 2.0ghz versus 1.3 |
| 13:35 | topi`> | at least when idling |
| 13:35 | topi`> | my unscientific test involved in removing the fan and seeing how rapidly the temperatures rise when loading cores |
| 13:35 | topi`> | and to which value they converge at idle |
| 13:36 | Ke> | last we discussed here, 1.3GHz limit is just a limitation, but has no other effects on the cpu |
| 13:36 | Ke> | ie. if you software limit to 1.3GHz with 2.0GHz hw limit, it should be completely equivalent to 1.3GHz hw limit |
| 13:37 | Ke> | but it is known that on some systems dynamic frequency scaling is not equivalent to hw frequency settings |