Librecomputer Tritium H3

[Igor]

[WIP]

 

https://libre.computer/products/boards/all-h3-cc/

Schematics

Pros:

- yet another H3

Cons:

- micro USB powering

[TonyMac32]

There is an H5 one as well, correct? 

[chwe]

Cons:

• Voltage regulation for CPUX. The way I understand the schematics and @TonyMac32 'confirmed' it. There's no dynamic voltage regulation for feeding the CPU with 1.1V/1.3V. IMO thermal behavior of the board has to be checked before the board gets 'official support'.
• Yet another H3 board...   IMO there's 'not much' where this board shines for 'another H3 board'. 

Pros:

• eMMC additionally available 
• Heatsink which covers RAM& CPU
• All USB are deployed on type A connector

Questions:

• Did someone check their CSI connector? Schematics says: 'GC2035 200W'. What does this mean for  DVDD voltage? (GC2035 needs 1.8V, OV5640 needs 1.5V).  Does this board need a 'camera expansion board'? 
• Seems that libre computer rushes with a lot of different SoCs to the market no (LeKartoffel months ago, Tritium and Renegade now). From what I see, they maintain their products not bad in terms of software upstream. Do the devs think the same? This would decrease the 'amount of work' to maintain armbian on this board a lot. 

[TonyMac32]

In terms of "Le Potato" yes, Libre Computer is supporting development actively through the same team Amlogic is using to get their SoC's mainlined. It has been nice seeing the improvement from kernel to kernel, and they've been responsive to issues/feedback. If that's the rule for how they proceed, then I have some confidence in the boards barring any hardware mistakes. I think it would be best to verify the voltage situation, as I saw a few oddities on that portion of the schematic concerning layout and labelling. (Nothing "wrong" that I remember , but some labels and such)

Sent from my Pixel using Tapatalk

[mpmc]

54 minutes ago, chwe said:

Yet another H3 board...   IMO there's 'not much' where this board shines for 'another H3 board'. 

 

Exactly, I saw the board(s) and thought, why?! What does this do better than any Orange Pi H*?

[vgjdujdtcfhmrtsjy]

.

[guidol]

On 08/12/2017 at 5:32 PM, mpmc said:

 

Exactly, I saw the board(s) and thought, why?! What does this do better than any Orange Pi H*?

Running cooler?  Maybe....

The coolest here - at my home - are the NanoPi Neo Core2 and the Orange Pi One (both with mainline kernel)

against the hot Orange Pi Zero.

Between these are the NanoPi Neo2 and the Orange Pi Zero Plus H5.

 

Dont know why the Orange Pi One (which is a little bit older) is running so much cooler than the others (maybe the bigger PCB)?

[lanefu]

I was shocked with how pleased i was with my le potato. It Ended up being a much better kick starter gamble than espressobin. I have a h5 tritium pledge that should come soon.

The libre folks seem to have a few things going for them:

* they have real capital to pay hardware and software engineers
* they've embraced the rpi form factor as a standard
* minimalist designs to control complexity
* coherent communication

Naturally i wish they used gig Ethernet, but i understand their design philosphy and just have a use case that isnt network heavy. Ive got orange pis for that.

Sent from my SM-G950U using Tapatalk
 

[TonyMac32]

I was shocked with how pleased i was with my le potato. It Ended up being a much better kick starter gamble than espressobin. I have a h5 tritium pledge that should come soon.

The libre folks seem to have a few things going for them:

* they have real capital to pay hardware and software engineers
* they've embrassed the rpi form factor as a standard
* minimalist designs to control complexity
* coherent communication

Naturally i wish they used gig Ethernet, but i understand their design philosphy and just have a use case that isnt network heavy. Ive got orange pis for that.

Sent from my SM-G950U using Tapatalk

I have to agree le potato has proven itself to be simple and reliable. I've put the same heatsink on it as I put on the Tinker, it runs much cooler. The Tinker and it are more or less even on general use, each having strengths and weaknesses. I may need to get a Tritium H5, I passed on the funding run and shouldn't have, and I don't have any H5 boards.

Sent from my Pixel using Tapatalk

[lanefu]

I have to agree le potato has proven itself to be simple and reliable. I've put the same heatsink on it as I put on the Tinker, it runs much cooler. The Tinker and it are more or less even on general use, each having strengths and weaknesses. I may need to get a Tritium H5, I passed on the funding run and shouldn't have, and I don't have any H5 boards.

Sent from my Pixel using Tapatalk

The H5 has been growing on me. Container builds have been noticably faster on my opi PC2 than my opi plus 2E boards. I just added a prime to my nomad farm consequently

Sent from my SM-G950U using Tapatalk

[TonyMac32]

http://ix.io/18Fb

 

From an H3 and H2+ Tritium board, H3 first, H2+ last building against the WIP board definition.  Needs some work (stuck at 912 MHz, a couple other odds and ends) but not bad.  I'll play with the DT overlays on these as well, I have some hardware laying about that should work for testing.  I have an H5, need to figure out where to start with it.

[Igor]

4 hours ago, TonyMac32 said:

Needs some work (stuck at 912 MHz, a couple other odds and ends) but not bad. 

This is known bug/regression on all mainlined H3/H5 and we are waiting for next/better implementation.

 

Just check if there are any differences in the reference design. If we choose to keep them on a support list, then you need to define board config files (.wip) and add them to (next) web page. Not much hacking at this point. 

[TonyMac32]

23 hours ago, Igor said:

Not much hacking at this point. 

 

Right.  There won't be much hacking involved, I could replace the "add tritium H3" patches and support all 3 fairly easily, they share the entire device tree, only difference board to board is the board/SoC compatible, and at the latest I think 4.18 will support them without patches.  Other than the unexpected appearance of unknown hardware bugs, these boards are quite literally the SoC, the PMIC, the RAM, and the various plugs.  The support question comes down to dealing with the typical micro-USB issues, and probably folks who don't know which board they have since they are all exactly the same looking... 

[TonyMac32]

Hmmm, having a strange situation where, even using the appropriate device tree entries, HDMI fails on any images I make using the appropriate upstream device tree sources...

 

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/arch/arm/boot/dts/sunxi-libretech-all-h3-cc.dtsi?h=next-20180503&id=b8f4f1180726d53778771ebf8421bea13a63fc9b

 

This is a shared dtsi that each of the three boards includes, only defining the compatible string and board name, and including the correct proccessor dtsi. This works on 4.17, but for some reason, despite looking correct for our patched 4.14 doesn't seem to do the job...  It's worth noting the 2017.11 u-boot has no problem with this structure either.

 

[edit]  Forgot: http://ix.io/19mU

 

### Loaded modules:

Module                  Size  Used by
zram                   28672  4
ir_lirc_codec          16384  0
lirc_dev               20480  1 ir_lirc_codec
sun4i_codec            49152  3
sun8i_codec_analog     28672  1
snd_soc_core          155648  2 sun4i_codec,sun8i_codec_analog
snd_pcm_dmaengine      16384  1 snd_soc_core
sunxi_cir              16384  0
snd_pcm               106496  2 snd_pcm_dmaengine,snd_soc_core
snd_timer              32768  1 snd_pcm
sun4i_gpadc_iio        16384  0
iio_hwmon              16384  0
sunxi                  20480  0
musb_hdrc              98304  1 sunxi
industrialio           65536  2 iio_hwmon,sun4i_gpadc_iio
sun8i_dw_hdmi          16384  0
sun4i_tcon             28672  1 sun8i_dw_hdmi
dw_hdmi                28672  1 sun8i_dw_hdmi
cec                    53248  1 dw_hdmi
rc_core                32768  5 ir_lirc_codec,cec,lirc_dev,sunxi_cir
sun4i_drm              16384  0

hmmm

[Igor]

 

5 hours ago, TonyMac32 said:

HDMI fails on any images I make using the appropriate upstream device tree sources

You have HDMI out on 4.17 but not on 4.14.y?

[TonyMac32]

Correct. A bit strange. There is a dw-hdmi patch that needs a fixup, but I tried to to no effect.

@Igor my support patches for this board do the opposite of the OPi PC2 image I tried, I get debug console with uboot on mine, but no kernel hdmi, the opi image made today has no uboot hdmi but does have kernel.  I'm sure I'm doing something ridiculous...  Any argument about uploading as CSC even though it's headless?  it's only a csc, uboot and kernel patch to add the dts's?

[lanefu]

My Tritum h5 board came.   I just built an image from Dev branch using Tony's tritium-5.csc. w/ Kernel 4.14.43     Fired right up..... Even has console login on HDMI.   (although its not returning keystrokes at the console)

 

but I ssh'd in an all is well.    

 

Trying to understand a bit more about regression bug.   I'm running at 1.01ghz.  Is that related to regression bug?   Basically trying to understand on how I can help achieve parity with the Clock on my Opi PC2 or Prime. etc

[TonyMac32]

12 minutes ago, lanefu said:

Even has console login on HDMI.   (although its not returning keystrokes at the console) 

 

Well that's new, I had nothing on screen...

[lanefu]

1 minute ago, TonyMac32 said:

 

Well that's new, I had nothing on screen...

 

Yep all I did was look at it

[TonyMac32]

I've gotten tied up trying to gracefully switch to mainline u-boot on Le Potato, haven't had time to double back on this (hence CSC). 

[TonyMac32]

@lanefu try a different keyboard if you have one, I just logged in and got to desktop on the build I just made.  Color me surprised.

[lanefu]

Just now, TonyMac32 said:

@lanefu try a different keyboard if you have one, I just logged in and got to desktop on the build I just made.  Color me surprised.

 

Uhhhh yeah.. so uhm batteries.   Keyboard is good!

[TonyMac32]

http://ix.io/1be2

 

Error:  thermal thermal_zone0: failed to read out thermal zone (-110)

[    6.600383] sun4i-codec 1c22c00.codec: ASoC: /soc/codec-analog@01f015c0 not registered
[    6.600397] sun4i-codec 1c22c00.codec: Failed to register our card
[    6.618111] sun4i-codec 1c22c00.codec: Codec <-> 1c22c00.codec mapping ok

 

Not bad.

 

I believe there is a known issue involving the CPU Frequency.  Also remember this board does not have the "V" part of the DVFS, so it will be a little bit less flexible power wise.

[lanefu]

4 minutes ago, TonyMac32 said:

http://ix.io/1be2

 

Error:  thermal thermal_zone0: failed to read out thermal zone (-110)

[    6.600383] sun4i-codec 1c22c00.codec: ASoC: /soc/codec-analog@01f015c0 not registered
[    6.600397] sun4i-codec 1c22c00.codec: Failed to register our card
[    6.618111] sun4i-codec 1c22c00.codec: Codec <-> 1c22c00.codec mapping ok

 

Not bad.

 

I believe there is a known issue involving the CPU Frequency.  Also remember this board does not have the "V" part of the DVFS, so it will be a little bit less flexible power wise.

 

Okay so I searched around on the forums a while back for this and really struggled for a final conclusion.  Maybe you can help me understand the difference.

 

DFS - Dynamic Frequency Scaling ?? -- cpu changes frquency through magic?
DVFS - Dynamic Voltage Frequency scaling?? -- cpu changes uhhh frequency through a regulator?

 

[TonyMac32]

1 minute ago, lanefu said:

DFS - Dynamic Frequency Scaling ?? -- cpu changes frquency through magic?
DVFS - Dynamic Voltage Frequency scaling?? -- cpu changes uhhh frequency through a regulator?

 

SoC's have built-in support for changing frequency, generally.  So the "Magic" is a driver controlling a clock generator/multiplier.  However, most of the time, with reduced frequency you can reduce operating voltage a little bit to save on power and reduce heat.  In this case that isn't going to be an option, on the Allwinner-based boards that support it it's a bit of a hack in the first place, using a transistor on a gpio to change the target voltage of the system regulator by adding/removing a resistor from the regulator's "voltage setting".  This board does not support that.  On a board like the Tinker, Renegade, Rock64, there is a matched/recommended regulator chip PMIC that takes commands and adjusts the voltages accordingly in a much more controlled fashion.

[TonyMac32]

To update my thoughts/observations:

 

Pros:

   

• On 12/8/2017 at 8:36 AM, chwe said:

  • Yet another H3 board...   IMO there's 'not much' where this board shines for 'another H3 board'.

  • Exactly, to some extent this is the base model of Allwinner boards.  Short of a failure to build it correctly it is simply going to work.  (I drive a car without power windows or locks, so I'm partial to simple things)  Low development support footprint for something people will buy.
  • Using all SoC peripherals, no random changing of wireless chips or ethernet PHY's to cause wrinkles/confusion. Also low development/support footprint.
• Faithful replication of RPi form factor (for those who are interested in that, which is a few people)
• Upstream mainline activity is basically done, the device trees/etc are already available, no exotic hardware
• All 3 boards share the entire device tree other than the processor compatible/include.
  • Pin compatible, so the same board definition goes for all three.  Low development/support footprint again.

Cons:

• Typical "powered via micro USB, beware your system voltage" warning applies.
• Not sure how many devs have it/would be interested.  Somewhat mitigated by the "simplicity pro"
• No voltage scaling, although I'd have to see what the power costs truly are, and test stability.
• The biggest confusion I see is actually in the naming and marking scheme.  Users, after putting their heatsinks on, may forget which one they have and download the wrong image for their "ALL-H3-CC" (written on every board).  There are no distinguishing marks other than the giant "Hx" on the SoC itself, which you cover up.  I would like to think the average user could figure it out, but... 

I propose this board series be placed in WIP. Typical statement about gpio powering/"no single board computer really meets the USB host spec for power delivery" etc.  Once I remember my details I can make an entry for it on the page.

 

I will need to change the H3/2+ config away from the current device tree and substitute in the correct one now that it exists.  Do we want separate configs for the H2+ and H3?  Or would it make any difference other than the user being confuses at seeing "H3" in their welcome screen?

[tkaiser]

3 hours ago, TonyMac32 said:

on the Allwinner-based boards that support it it's a bit of a hack in the first place, using a transistor on a gpio to change the target voltage of the system regulator by adding/removing a resistor from the regulator's "voltage setting".

 

The first and also all the 'better' H3/H5 boards from Xunlong and FriendlyELEC use an I2C accessible SY8106A voltage regulator allowing to adjust voltage in 20mV steps. No real difference to a real PMIC wrt DVFS. The Orange Pi One was the first board with a more primitive voltage regulation scheme only supporting 2 voltages using SY8113B/AX3833 voltage regulators. Then came SinoVoip and forgot voltage regulation on their overheating BPI M2+, later FriendlyELEC decided to drop voltage regulation on their NEO2 since Allwinner's H5 BSP didn't support it any more and that was most probably also Libre Computer's 'motivation' to drop voltage regulation at the same time.

 

But while Allwinner themselves do not support voltage regulation any more with H2+/H3 in their BSP (it has also been removed in their 4.4 kernel code drop) and never supported it with H5 the mainline kernel code that properly implements voltage regulation with both SY8106A and SY8113B/AX3833 without using the ARISC core exists for over 2 years now.

 

In other words: the Libre Computer Allwinner boards are designed to run with Allwinner BSP based OS images (Android / crappy Linux) and not mainline kernel   This design decision allows for lower peak performance and higher idle consumption at the same time.

[tkaiser]

2 hours ago, TonyMac32 said:

Not sure how many devs have it/would be interested

 

Allwinner H3 was exciting back in 2015 and early 2016. Allwinner H5 was interesting since still inexpensive but full ARMv8 feature set (especially ARMv8 Crypto Extensions). Basically all H2+, H3 and H5 boards are the same: https://forum.armbian.com/topic/1351-h3-board-buyers-guide/?do=findComment&comment=44979

 

For an Allwinner H device able to draw some attention it needs either to be dirt cheap or needs interesting features like great CPU performance (needs DVFS to go beyond 1.3 GHz) or great amount of DRAM (impossible since 2 GB DRAM are the max) or great graphics capabilities (impossible, only boring Mali 4x0 and limited video engine).

 

The Tritium board (it's just one with 3 different SoCs soldered to it) was designed as an Android TV box without enclosure (closely following Allwinner's current reference design to be able to run their BSP stuff and therefore vendor's Android). No Gigabit Ethernet, no voltage regulation and more expensive than $10. No idea why I would choose this board for which use case...

 

It's 2018 now...

 

[tkaiser]

On 12/8/2017 at 8:19 AM, Igor said:

https://libre.computer/products/boards/all-h3-cc/

 

And also https://www.loverpi.com/collections/libre-computer-project/products/libre-computer-board-all-h3-cc?variant=3133794353165 do not look like a supported product.

 

At the time of the Tritium campaign the prices were attrative (especially the little $9 H2+ loss leader) but now with each board being $11 more...

[lanefu]

6 hours ago, tkaiser said:

 

Allwinner H3 was exciting back in 2015 and early 2016. Allwinner H5 was interesting since still inexpensive but full ARMv8 feature set (especially ARMv8 Crypto Extensions). Basically all H2+, H3 and H5 boards are the same: https://forum.armbian.com/topic/1351-h3-board-buyers-guide/?do=findComment&comment=44979

 

It's 2018 now...

 

 

@tkaiser Is there anything out there now that is as exciting as the H3 was?   

 

I got an h5 tritium because it has 2 gigs of ram, and I figured it would be fun to add to my nomad cluster (for stuff that doesnt need a lot of network).    I think my pre-order was cheaper than what the cost of a Opi prime would be.... PS Primes are back in stock.