<montjoie>
I have weird problem on all my board, getting mad
<montjoie>
network work but so slow
<montjoie>
even my old working kernel is slow
<wens>
mine is still at 2017.11-rc4 plus some patches, I just kept rebasing stuff
<montjoie>
I have the impression that network is random on it
<montjoie>
the kernel were I am sure to have network tested before sending patch, doesnt have network yesterday
<montjoie>
perhaps the power suply is dieing
awais has joined #linux-sunxi
<wens>
I'm using Anker 60W 10 port power bricks
<montjoie>
I have buyed compoments for using an ATX
voxadam has quit [Quit: WeeChat 1.9.1]
awais has quit [Read error: Connection reset by peer]
<montjoie>
I have problem with my pine64 with the same power supply...
voxadam has joined #linux-sunxi
awais has joined #linux-sunxi
IgorPec has quit [Ping timeout: 268 seconds]
JohnDoe71rus has joined #linux-sunxi
JohnDoe_71Rus has quit [Read error: Connection reset by peer]
diego_ has joined #linux-sunxi
diego_ is now known as diego71
chlorine has joined #linux-sunxi
<wens>
most changes in sunxi-next have landed in linus' tree; only missing clk changes
chlorine has quit [Ping timeout: 248 seconds]
yann has quit [Ping timeout: 255 seconds]
juri_ has quit [Ping timeout: 268 seconds]
awais_ has quit [Ping timeout: 258 seconds]
awais_ has joined #linux-sunxi
chomwitt has joined #linux-sunxi
anarsoul has quit [Ping timeout: 255 seconds]
anarsoul has joined #linux-sunxi
DullTube has joined #linux-sunxi
msimpson has joined #linux-sunxi
matthias_bgg has joined #linux-sunxi
lemonzest has joined #linux-sunxi
<wens>
are H3 and H5 supposed to be 40nm or 28nm?
kivutar has quit [Ping timeout: 248 seconds]
kivutar has joined #linux-sunxi
kivutar has quit [Ping timeout: 260 seconds]
<lvrp16>
40nm
kivutar has joined #linux-sunxi
chlorine has joined #linux-sunxi
chlorine has quit [Remote host closed the connection]
chlorine has joined #linux-sunxi
yann has joined #linux-sunxi
popolon has joined #linux-sunxi
IgorPec has joined #linux-sunxi
chlorine has quit [Remote host closed the connection]
Andy-D__ has joined #linux-sunxi
sr-digitronic has joined #linux-sunxi
Andy-D_ has quit [Ping timeout: 248 seconds]
chlorine has joined #linux-sunxi
chlorine_ has joined #linux-sunxi
chlorine has quit [Read error: Connection reset by peer]
awais_ has quit [Ping timeout: 268 seconds]
sr-digitronic has joined #linux-sunxi
sr-digitronic has quit [Remote host closed the connection]
dave0x6d has quit [Quit: Connection closed for inactivity]
awais_ has joined #linux-sunxi
kaspter has quit [Ping timeout: 248 seconds]
chlorine_ has quit [Remote host closed the connection]
chlorine has joined #linux-sunxi
chlorine has quit [Ping timeout: 260 seconds]
anarsoul has quit [Ping timeout: 240 seconds]
anarsoul has joined #linux-sunxi
OneSploit has quit [Ping timeout: 268 seconds]
awais_ has quit [Ping timeout: 240 seconds]
awais_ has joined #linux-sunxi
awais_ has quit [Ping timeout: 250 seconds]
afaerber has joined #linux-sunxi
swiftgeek has quit [Ping timeout: 260 seconds]
chlorine has joined #linux-sunxi
dev1990 has joined #linux-sunxi
dev1990 has quit [Remote host closed the connection]
dev1990 has joined #linux-sunxi
kaspter has joined #linux-sunxi
awais has quit [Ping timeout: 248 seconds]
awais has joined #linux-sunxi
hardfalcon1 has quit [Ping timeout: 255 seconds]
kaspter has quit [Ping timeout: 255 seconds]
tkaiser has joined #linux-sunxi
gnarface has quit [Remote host closed the connection]
hardfalcon has joined #linux-sunxi
DullTube has quit [Quit: Leaving]
gnarface has joined #linux-sunxi
gnarface is now known as Guest44576
Guest44576 is now known as gnarface
juri_ has joined #linux-sunxi
JohnDoe71rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
swiftgeek has joined #linux-sunxi
rexxster has joined #linux-sunxi
ernestask has joined #linux-sunxi
<tkaiser>
willmore: You have an OPi PC2 and a NanoPi NEO2 or Plus2, right?
jbrown has joined #linux-sunxi
matthias_bgg has quit [Ping timeout: 264 seconds]
cnxsoft has quit [Quit: cnxsoft]
<willmore>
tkaiser, results posted.
benettig has joined #linux-sunxi
<willmore>
I'm getting a NanoPi2 shortly, but that's not relevant. FWIW, I have a PC2 and a 0+
<willmore>
For H5 boards at least. I have some zero an one boards as well as a PC.
megi has joined #linux-sunxi
<tkaiser>
willmore: Thank you. So memory bandwidth with H5 regardless of single bank vs. dual bank DRAM is a lot better compared to H3
<willmore>
Yay!
<willmore>
I wonder if we're just misconfiguring the H3s...
<tkaiser>
Hmm... I don't remember having tested this with Allwinner's BSP DRAM initialization...
awais has quit [Ping timeout: 250 seconds]
awais has joined #linux-sunxi
dave0x6d has joined #linux-sunxi
chomwitt has quit [Ping timeout: 260 seconds]
<icenowy[m]>
evne the PC2 DRAM configuration may be not stable
<icenowy[m]>
segfaults are seen on my pc2 @ 672MHz
<icenowy[m]>
and memory corruption is detected by memtester
<icenowy[m]>
but @576MHz they disappeared
benettig has quit [Quit: Leaving]
<icenowy[m]>
my Prime has the same problem and I downclocked it to 624MHz
<tkaiser>
icenowy[m]: In Armbian we use also 624 MHz everywhere. The only round of tests with OPi PC one and a half year ago suggested that trusting in 'vendor numbers' is not a good idea. But since then this has been forgotten and these funny 672 MHz appear everywhere again :(
<icenowy[m]>
mripard suggested me to do the test again on BSP boot0
<icenowy[m]>
to ensure it's not problem by opensource DRAM initialization
<tkaiser>
Well, I tested with BSP boot0 on Orange Pi Zero Plus. Just to realize that there's 672 MHz written in the .dtb and 624 MHz used in reality.
<icenowy[m]>
did you do any stability test?
<lvrp16>
with the recent increases in ddr prices, there's significant bom pressure at the current price points
<icenowy[m]>
my test is to build Linux kernel, run tinymembench and run memtester parallelly
* willmore
is glad to test on PC2 and Zero+
<icenowy[m]>
lvrp16: yes. and for Orange Pi 3 it's noticeable that they used degraded chips ;-)
<tkaiser>
Sorry, no more time to waste, if upstream 672 MHz is set it gets patched away to 624 MHz immediately.
reinforce has quit [Quit: Leaving.]
<lvrp16>
whats the fastest stable speed for the ddr controller?
<tkaiser>
lvrp16: Don't know but remember a discussion between beeble and apritzel months ago. It was a lot about calibration and stuff... and choosing the right modules with right trace routing. Something Theobroma is doing but what not happens on cheap TV boxes and tablets
<tkaiser>
And at least newer BSP kernels have ddrfreq driver and DRAM clockspeed seems to be part of budget cooling. The values in the H5 BSP can get really low
<lvrp16>
ahh gotcha
Pe3ucTop has joined #linux-sunxi
<tkaiser>
At least with H5 a BSP Android does not clock DRAM above 624 MHz with BSP defaults. IIRC there were 3 values defined as available_frequencies or something like that below /sys/devices/1c62000.dramfreq/devfreq/dramfreq/
<tkaiser>
I only monitored cur_freq and this remained at 624000 all the time
TEKrantz has quit [Ping timeout: 240 seconds]
TEKrantz has joined #linux-sunxi
<tkaiser>
icenowy[m]: How long did you have to wait for a memtester error?
vagrantc has joined #linux-sunxi
<tkaiser>
This is how it looks like with H5 BSP (u-boot + kernel): cat available_frequencies --> 168000000 336000000 672000000
<tkaiser>
cat cur_freq --> 624000
<tkaiser>
cat governor --> adaptive
kloczek has joined #linux-sunxi
<willmore>
One starts to wonder if things are working when cur_freq is not in the set of abailable_frequencies...
<tkaiser>
willmore: Of course you can't write to any of these nodes
<tkaiser>
But tinymembench numers _look_ like it's really 624 MHz comparing with mainline u-boot where we can set the DRAM clockspeed as we want
deserter has joined #linux-sunxi
matthias_bgg has joined #linux-sunxi
<willmore>
Is there a standard for drivers like this? Are they allowed to return values that aren't even in the available_freq? Do they return values when they don't know for sure what's set? Like if the driver didn't change the setting since boot, do they measure/read or do they just assume that it's at default? It seems there's no standard.
<tkaiser>
You're talking about Alwinner's H5 BSP, the one not even able to get voltage regulation working ;)
<willmore>
In this case, sure. I was thinking in the general sense.
<willmore>
Yay, BSP....
matthias_bgg has quit [Ping timeout: 260 seconds]
reinforce has joined #linux-sunxi
netlynx has joined #linux-sunxi
<lvrp16>
tkaiser: i think most allwinner designs are moving away from programmable voltage regulator
<lvrp16>
because most people using these are looking for the cheapest bom
<lvrp16>
by most people i mean orders of 10k+
<willmore>
The performance and thermal benefits of the programmable regulator are pretty darn nice.
<willmore>
The performance diff between an Orange Pi One and a PC is notable.
<willmore>
The PC2 vs the Z+ isn't quite fair as they're both pretty messed up for clock speed and voltage right now, so I won't give the PC2 a clear win here even though it's way faster now.
<tkaiser>
lvrp16: But how much more do the resistors and GPIO access cost to implement primitive voltage switching? I'm somewhat fine for my use cases with 1.1V vs 1.3V
<lvrp16>
I think Allwinner only certifies the chips for 1GHz at 1.2V for H3 and 1GHz at 1.1V for H5. Any speed over 1GHz is overclocking and some part of the chip will fail to operate deterministically.
<lvrp16>
1.296 is almost 30% outside of the production verification so for Allwinner's big customers, they don't see the need for PVR
msimpson has quit [Quit: Leaving]
<lvrp16>
even if it a simple GPIO voltage toggle
<miasma>
willmore: it has nothing to do with memory bandwidth and on-board heatsink?
<willmore>
Aren't the chips normally sold with the processor in bundles? It might not be as simple as looking at the cost difference between the chips.
<willmore>
miasma, it could. I don't know the specifics of the memory interface between the boards. You make a very good point about the PCB being a good heatsink as well.
<miasma>
iirc opi pc had better cooling via the pcb than many other h3 boards
<willmore>
The voltage regulator may be peripherial to the reason they differ so much in performance.
<willmore>
In my case the PC has no extra heatsink while the One does.
<miasma>
i'm curious about the perf and pcb quality of the new kickstarter board
<willmore>
The PC is in an enclosed space while the One has air flowing over it.
<miasma>
it's a bit disappointing that these SBCs can't always power all usb hosts with 500mA
<willmore>
lvrp16, how's the board design?
<lvrp16>
there's barely any perceptable difference between the orange pi one and pc last time i ran benchmarks
* willmore
almost never plugs anything more than a keyboard or mouse into these boards.
<lvrp16>
maybe like a 2-3% difference at best with the same cooling
<miasma>
willmore: i usually have some issues with wifi, this one appliance had a usb scanner connected
<miasma>
it would jam with rpi
<willmore>
Interesting. I see more than that, lvrp16. And my One has all the advantages I can give it while the PC is as crippled as I can make it.
* willmore
groans at the bad pun
<willmore>
Bad miasma!
<miasma>
i fixed the issue with some polyfuse bypass wires
<willmore>
Evil.
<miasma>
with fuses you never get fancy light effects
<willmore>
Which is a good thing.
<willmore>
A clever board designer would put an LED and resistor in parallel with the fuse.
<lvrp16>
those are benchmarks I ran with standardized cooling
IgorPec has quit [Ping timeout: 240 seconds]
<willmore>
"Do you see flickers of light near your USB ports?" "Yes." "They you're overcurrenting them."
<lvrp16>
5-6% difference?
<willmore>
lvrp16, what tests did you run? PTS stuff?
<lvrp16>
yes
<willmore>
Ahh. I'll redo my tests when I get a moment. Which is to say, in a week or two. *sigh*
<lvrp16>
those tests were really old
<lvrp16>
idk if we have the same ddr chips
<lvrp16>
or if armbian updated the clocks
<lvrp16>
miasma: the kickstarter tritium boards cut no corners on design and component quality
<tkaiser>
lvrp16: willmore: When testing today you'll get same performance difference (below 7% for sure) at least with the dual-bank models. With H3 and single bank DRAM it's a bit different and depends on whether the benchmark benefits from memory bandwidth or not.
<miasma>
lvrp16: any idea about the GND/5V rails? e.g. I think rpi only provides something like 150 mA to the usb ports
<tkaiser>
miasma: no, more and even configurable in /boot/config.txt
<miasma>
tkaiser: i meant the original rpi, can't remember what config rpi2/3 have
<miasma>
on rpi model b, it's totally up to the polyfuses
<tkaiser>
Well, no idea. But on the newer one you get either 600mA or 1200mA depending on max_usb_current=1 or not
<lvrp16>
i think i tried to get powerdraw on the raspberry pi 3 past 1.5A
<miasma>
opi boards probably don't have any specs for usb power output, but i've noticed that the ethernet light becomes dim when it runs out of power
<lvrp16>
couldn't do it
chlorine has quit [Remote host closed the connection]
<lvrp16>
thats probably your power supply not having a high enough voltage
<tkaiser>
miasma: There are current limiters with IIRC 1.5A each. One for the vertical receptacle and one for the two other receptacles.
<lvrp16>
i hooked a raspberry pi to a programmable dc supply, i got the rainbow square if i didn't set input voltage above 5.2v
<miasma>
tkaiser: but i doubt the rails can handle that
<lvrp16>
even at 5.15v the rainbow square came on
<lvrp16>
the voltage at the 5V rail after the input polyfuses at 1.5A had like a 0.5A disparity
<lvrp16>
0.5V*
<lvrp16>
poor design for sure
<willmore>
miasma, looks like the B has similar limits but without the port specific polyfuse--the board has one and there's not a lot of headroom.
<miasma>
lvrp16: not really. in my case the problem was probably the musb cable. when i connect the supply to gpio pins, it can provide a bit more power. i suspected the onboard rails also add to the total resistance
<lvrp16>
the 5V pins on the pi are directly to the 5V rail
<lvrp16>
so you are not polyfused
<lvrp16>
very different from going into the microusb connector
<miasma>
yep, but with the opi board
<miasma>
if i connect to wifi dongles to opi pc, the musb cable needs to be really good, otherwise the ethernet port leds become dim with the dongles connected
<miasma>
the dongles draw 1A max
<lvrp16>
willmore: i tested the current on a pi 3 with ssvb's cpuburn and the current was 1.5A. I hooked up an empty battery to the MicroUSB port and the current was still 1.5A.
<lvrp16>
miasma: do you mean the rpi? or the opi pc?
<miasma>
opi pc
<lvrp16>
i thought that was barrel plug?
<willmore>
it is
<miasma>
oops right. but the cable wasn't too good either
<lvrp16>
???
<miasma>
some of the psp charging cables are too thin
atsampson has quit [Ping timeout: 252 seconds]
atsampson has joined #linux-sunxi
IgorPec has joined #linux-sunxi
<[TheBug]>
hmm I would suggest for OPi products ordering their power supply from them it works well and if 5v 3A if I recall
<[TheBug]>
so has plenty of headroom
<[TheBug]>
using third party chargers may not work as well for the reason you noted -- many have small gage cable
<[TheBug]>
gauge*
popolon has quit [Quit: WeeChat 1.9.1]
<miasma>
yes, there are all sorts of cables. I'm a bit disappointed with the ebay/amazon/ali psp charger cables which fit, but they're thin and maybe too long
<tkaiser>
I read a review recently from someone who bought counterfeit power cables on either eBay or Ali (1.5 mm diameter and nice 'gauge rating'). Since he always wondered why they melt he decided to open them just to have a nice laugh.
<willmore>
I made my own cables. They're short, high diameter wire and they go right to a very solid 5V supply. I have some dupont connectors on that same supply so I can go through GPIO if a board needs it.
anarsoul|2 has joined #linux-sunxi
<willmore>
tkaiser, you need me to run that deb?
Pe3ucTop has quit [Ping timeout: 248 seconds]
<miasma>
willmore: i haven't found any metal connectors for the dc plug. the plastic ones are really weak
<tkaiser>
willmore: Do you have a single-bank DRAM H5 board?
sr-digitronic has quit [Remote host closed the connection]
yann has quit [Ping timeout: 240 seconds]
jelly-home is now known as hurtigbuffer
hurtigbuffer is now known as jelly
Andy-D__ has quit [Remote host closed the connection]
JohnDoe_71Rus has joined #linux-sunxi
<deserter>
hello. I'm trying to add capture support to A20 audio codec to mainline kernel 4.13. Added enumerated widget for controlling "ADC source Mux" but amixer and alsamixer dont let me to change default value of "2" to any othe valid value. Permissoins are rw----, i can change other controls. Where to go look now?