Creating an NDI 1080@30 media box

(Erik Szabo) #1


I am somewhat new to the single board computer scene. I am planning to make a media box for screens. This solution would use the newly popular NDI transferring method, so I was looking for a board with gigabit connection at least as the raw video and sound have traffic ~ 130 MBps. The plan is to achieve ~0 sec delay as the live footage reaches the screens. I was wondering if the BPI-M3 could handle this workload. Any input is welcomed on this topic.

Best regards, Snowmaker

(David Coles-Dobay) #2

Not yet

The situation with the M3 is that consumer expectations are much greater than the ability of the small development team can satisfy.

The chip is actually limping with software mashed up from other kernel sources. I am endeavoring to create a separate source tree that will not have the same constraints that the current one has. Once complete it should handle the throughput no problem. I have a different product using this SOC chip that I am working on fixing this product is a gift back to the guys and girls who struggled so long. The through put is actually evident but the software is not in the best of shape.

(Erik Szabo) #3

Well hello there.

Let me thank you for an official statement on this question :slight_smile: Other alternative, I sat up an rtmp server, what board should I use for for an ffplay -nobuffer application? I managed to tune the delay down to 1.5sec that should be okay for now.

Best regard,

(David Coles-Dobay) #4

I was wondering about this, so you are able to do fullscreen 1080i at 3.5MBs or is it just a scale up?

And do you know if RTSP will work as well ?

Oh and I got the GCC 8.0 and Binutils 2.28 finished the other day.


Hi Erik,

The M3 has 8 cores, but only one (1) that is responsible for graphic. If you look for example at BPI-M2U, this chip comes with: MALI-400 MP2 @ 500MHz, but M2U has not voltage correction for lower clocking so it gets hot.

In any case, GPU and VPU are not so good supported on ARM except you stay with android. Read, before you buy. It is worth the time :wink:

and a nice overview of Linux development: also a nice place to search

(David Coles-Dobay) #6

Actually it is a little different than what was described by Tido. The A83T does not use the Mali architecture at all it is a different GPU called SGX544MP1 it has two function paths one H264 the other is for other codecs. It does not use one of the processor cores for video as stated.

The issue with video on this chip has been IMO poor implementation of the XSGX kernel extensions and mismatched software. By default the downloadable images seem to all require the use of X11 video rather than vidpau or other video methods. So IMO the graphics cores have not been properly even turned on for use. The architecture of the chip is different than the mali varieties and I have not seen a clean implementation of the XSGX yet on these chips.

I have seen slower less capable SOC provide excellent graphics and video on other XSGX type chips. My thought is that it should only be a case of properly porting the improved source code onto the linux trunk to get the correct code.

I am working on building a toolchain based off the GCC 8 and linux 4.11.11 releases to run Debian 9 binaries. This will fork or branch completely off from the current sunxi distro / group / whatever that has removed armv7L from the binutils and GCC recognized processor platform list.

On compiling a different tool chain and linux distro. I am unsure of what rules or other impediments have been placed on the different code repos. I do not intend to do up commits back to the original code but to provide them as a separate repo.

The source code as it stands now has code for every supported architecture and every supported protocol. This is a lot of unused and unrelated code. An example is there is code associated with Firewire on Sun Sparc systems in the archive. While that is great for Sparc owners it has no benefit in the repo for the A83T. What I will be doing is paring down the source code to only what is able to be used by the processor and chip set available.

Paring down the repo will also allow automated builds and the ability to track changes only occurring to the same files in mainline.

Also code not currently included in the main repo can be included to augment the new toolchain from other repo sources. So the faster more current development of the PowerV GPU source code can replace the code in mainline included in the autosevo repo.

Yes it may take a few months but there will be a toolchain for the A83T that is up to date and correct for the SOC.


Nope. A 3D accelerator does exactly nothing video related. You even fail to understand the difference between encoding (h.264 only) and decoding. Good luck to you!

(David Coles-Dobay) #8

No charles I am not talking about the 2 pipelines setup for encoding H264 I am talking about the 2 out of 4 pipeline setup for output decoding one is multiformat the other is h264.

And with code with this processor you can use any pipeline in either direction encode or decode. The processor for the GPU is different than what you are aware of charles.

The new code should ( if I get the blobs right ) have four independent channels available with one tied to the camera in by physical hardware.


Dear auto,

understood, you believe more in wrong block diagrams than reality. Maybe important to you: Allwinner will soon start mass production of an A83T successor where A7 cores have been exchanged with A53 ones. VR9 will still use an Imagination GPU bust most probably not slow SGX544MP1 as on H8vr (AKA A83T) again. But that’s only 3D acceleration and not related to video :slight_smile:

(David Coles-Dobay) #10

Thanks for the heads up on the A53 I think that the pi board is getting a different soc as stated here repeatedly. Regardless I am ordering a 5 trays of them for my first build anyway. Sinovoip will setup and build on 500 quantity.

But the block diagram comes from the Imagination core manufacturer and was included in the A83T however the source code in the mainline and sunxi and linero is a much older rev sgx530 or earlier. There was some argument on some forum about ffmpeg code in the blob and lacking a functional open source driver for the powerVR chip. There is a blob and source code with the Imagination API kit and I will be including it in my build tree. It is unknown at this point if the GPU is fast or slow as it has not even been turned on in any of the linux images . It looks pretty good on the Android 5.1 tablets this chip is used in. And the PowerVR GPU seem to work well in the iPhone and iPad products.

Announcements of chips Allwinner Technology has just introduced its new octa-core H8 System on Chip (SoC) for “high-end” gaming consoles and video OTT (over-the-top) boxes, right before the Hong Kong Electronics Fair taking place on October 13-16. The processor feature eight Cortex A7 cores clocked at up to 2.0 GHz, coupled with Imagination Technology PowerVR SGX544 GPU @ 700 MHz with support for OpenGL ES 2.0/1.1, OpenCL 1.1 APIs.

R58 looks identical

Not seeing any Announcement of a A53 upgrade anywhere. Kind of doubt it as it is a different architecture and pinout than the A83T, H8, R58 SOC

A53 is 64bit the current series 32bit

(David Coles-Dobay) #13

Ok so do I buy from Damian or Eva Wu or do I buy from the manufacturer of PowerVR series GPU that are in most of the Apple products and they are inferior to what you like because you say so and some random people say so.

I get it you are the God king of your world.

And Damian has the real information not the manufacturers information and EvaWu has the real announcements not the manufacturers announcements.

I do not need 64bit nor is it any benefit for what I am doing so your info means nothing.

Thanks I am buying the A83T from Sinovoip.


Above you quote Charles to be a Troll.

If I read his hidden posting:

While Charles embraces SoC - spends each day hours about this topics, what are you @autosevo ?

The one who takes out for him to judges peoples he has never seen in person ?

(David Coles-Dobay) #16

A few people here have had bad results with the things they are doing.

I am the one who have not had same bad results.

This individual on this forum selects to condemn and attack anyone not having his experience. It is kind of like a person who can not ride a bike attack those who are having no problem riding around him.

There are official documents from the vendors being provided by the vendors proper that I have found to be correct. There are forum posts from other than the official sources that I have not found confirmation of. I choose to put more weight on what is working than what is not officially stated.

It is really simple Charles had a bad experience with how he was choosing to use his board. He selected to find fault in the board and refute anyone who had solved the very issues he was having. This became a vendetta some point before my interaction with him.

Doing exactly what Charles did produces failure at every try.

Doing something different produced results that are beneficial.

At some point it becomes nothing more than vandalism to insist that all must do something incorrectly to obtain the reality he has found himself in.

I am the guy who is having success with these boards maybe you should listen.


I have ordered my BPi-R1 Jan. 24 2015 - a few months later I came across the person behind Charles. I have no clue how he can spend so much time for and about SoC’s, but he knows early and a lot about it.

In online communities we like to judge a person maybe to fast.

There are people like BrianBeuken, who I had to Mute - because their nonsense that we were wrong was just so much xxx that I couldn’t stand anymore words of him.

What I want to say, the conclusion, it was not about to offend you. He just explained to you that all the love and spare time you invest is never reflected in their (SinoVoip) repository. So all people coming after you, will face the exact same problem - no matter what you do. His anger is probably better described as a desperation.

And do not think, he has not tried. And if you ever curious how deep into detail he goes to: OpenMediaVault

How much conversation with the OMV guys ? to me this is a performance you have to proof first. Once you got the M3 up and running as promised, I grant you that :wink:

Last but not least, if you master that - over at armbian smart guys are always welcome :sunglasses:

(David Coles-Dobay) #18

The wonderful thing about open source hardware and software is that you do not have to rely on a group or a specific vendor. If you disagree you fork or simply create your own repo it does not mean that you violate some ones copyright you simply reflect your changes on your repo or remove the software if that’s not good enough.

If you demand things from people or expect them to act within the confines of your expectations you really do need to pay them.

Often people think that it is their calling to have vengeance on something they disagree with. It provides no benefit and all it does is cause harm.

Open Media is probably a great product. For me it didn’t pass as useable for my needs so I have not considered it since I first looked at it. Not that Open Media is unworthy in someway but of all of the different file sharing and file vault software out there I did not find one that fit my particular needs. That’s ok they are fine but I wrote my own that does what I want it to do.

So when I say “I don’t care” understand I have no control over anything that I did not personally make so why would I care ?

The whole point of what I am doing is not for any group, manufacturer or other entity but my own. If I share my efforts it is because I desire to. NO OTHER REASON.

My sharing what I am doing is just that sharing.

What benefit does joining some other group provide me ?

Do they provide me what I need … NO

So looking at the forum I see the factory rep sinovoip who may or may not have a team, trying to build a team. It looks like he has his hands full. This guy is not the company sinovoip he is the forum moderator.

Why there has not been a separate distro setup or at least a separate image M3 repository for the base images I do not know. However I was able to post my cleaned out images on a free web drive and make them available, why haven’t more people done the same ? I know some of the groups have published images and I am unsure of the issues if any.

I know that I have not had the same experience others have had mainly because I do things differently than others.

I have not joined any group of developers because I do not need to and I only share what I want to.

Maybe if you understood my creed you might understand my activity.

  1. I do not have a need to be ruled.
  2. I do not have to be forced to participate in common good.
  3. I have a god given right to my own tools.
  4. I have a god given right to my own body.
  5. I have a god given right to my own mind.
  6. I have a god given right to the products of my mind.
  7. I have a god given right to the products of my labor.
  8. I do not see a reason for law givers and law keepers.
  9. I take full responsibility for my actions.
  10. I share any of it on my own terms.


Hi David, you still didn’t understand my comments. Never mind and let it be. Cheers Tido

(David Coles-Dobay) #21

No actually Tido I do comprehend your comments. I do not stand under them.

It is the fact that some believe that their conclusion is so absolute or so popular as to be absolute, both are equal to arrogance when the expectation is that anyone viewing the same information would have the same conclusion. Some are so sure of their conclusion that they must try and convince or belittle anyone who does not share the same belief. Yes I know you have not had good luck and find that frustrating. That does not give you any rites of any kind. When someone else does not have the same conclusion as you it becomes threatening to your self. It really should not.

What ever the reasons for trying to force another to submit to your will is nothing more than an attempt at oppression.

As far as the SINOVOIP inc. provided images well they seem to be better than the ones supplied by anyone else. So simply do better yourself and stop being a bully.


Hi David,

I see your point. And you are right with what you wrote, but no rule without exception. Here is one.

physics. If a plug is designed to reach its limit with 5V @ 1.8 Amp (P = UxI (5x1.8 = 9Watt) this is a fact. We humans know the electricity so well to design it this way.

When electricity is transported, you lose energy (voltage) for transport. The thinner the lead, the higher the loss (the same applies to water lines). Two (2) major problems (cheap cable, wrong connector) that occur all the time. SinoVoip could fix this in serveral ways eg. sell a MicroUSB cable in their shop with thicker copper lines, just as an example.

To improve and do better does not take a lot, one more example ? see next lines

Github Pull request, armbian building tools. Both are designed for collaboration - but SinoVoip must collaborate.

SinoVoip truly, but sadly does not.

So it is not about arrogance or conclusion of me and others - it is about seeing your (SinoVoip) errors and to improve.

SinoVoip and Foxconn work in collaboration, I wrote that a while ago

(David Coles-Dobay) #23

I agree that the USB power supply is not enough.

It seems to me that a design for a regular product they produce has an upgraded processor placed. This is actually kind of cool like buying a regular VW golf with a GTI engine. So yes to get the GTI level of drivability you have to upgrade the clutch and the suspension.

The offered USB wall wart seems to be the SKU that BPI has in stock for the other boards.

I wrote a response to a question about NO START condition a while back.

In that post I gave the part number for a known good power supply and instructions to feed the power into the header rather than the USB port. I also mentioned that the barrel connector was enough but the end on the power supply would have to be changed.

I also mentioned the Heatsink to assure stable and reliable operation with this board. I am getting around to sending a macaque to BPI to see if they can make them inexpensively.

This is the nature of open source open hardware.

In my conversations with BPI I found the cost associated with customizing a production run for my needs. It was well within the budget for my project. I found the staff at BPI helpful and providing good service.

(David Coles-Dobay) #24

This is a mistaken concept the idea that anyone MUST do something.

It is a choice of the individual to participate or not. The very idea of MUST is repugnant to free minded peoples.

If Armbian or any group wishes to force another to do anything than payment would be in order. However if Armbian or any other group wished to create products or open source for the BPI products there is nothing preventing them from doing so.