EEVblog #883 - Orange Pi One vs Raspberry Pi 2

[EEVblog]

Dave takes a look at the Orange Pi One $10 single board 4 ARM Cortex A7 processor computer and compares it to the Raspberry Pi 2.
Coving the install of the ARMbian flavour of Debian.
Beware the security bug!
And what is the power/watt efficiency between the two?

http://www.orangepi.org/orangepione/
Schematic: http://linux-sunxi.org/images/7/7e/ORANGE_PI-ONE-V1_1.pdf
Allwinner H3 datasheet: http://linux-sunxi.org/File:Allwinner_H3_Datasheet_V1.2.pdf
Open source Allwinner H3 community:
http://linux-sunxi.org/Main_Page

[Vgkid]

Thanks for the video. At that price, it might be worth playing around with it.

[Whales]

Re 4K h265 decoiding @30FPS: whether or not you can use this comes down to whether or not drivers for it exist openly   Sometimes they keep the good drivers under wraps/for specific customers.

[Athanasis]

$10 for a Compute? This is fucking cool deal...  But who the hell sells this?

[Kjelt]

"Unfortunately you can not power the board through the mini USB connector you have to use the DC jack"
I wish the Pi had a DC connector like the beaglebone, who in *&^*&^ sakes has ever thought of powering a motherboard through a serial connector, its not a portable device people, it is a static computer.
I always solder my PSU directly to the PI pcb, not use that mechanically crappy usb port and I just hope that some day they will put just a decent Wago pcb connector or DC connector on it.

[Karel]

There are lots of boards around with better specs than the RPI.
The main reason why the RPI is a better choice, is the software eco-system and community.

[andiz]

A few corrections regarding the video:
The RPi Zero uses the ARM1176 core (1.25 DMIPS/Mhz), not the Cortex-A7 (1.9 DMIPS/Mhz).
The Allwinner H3 is rated at 1.3 GHz and the 1.6 GHz stated for the OrangePi PC are pure marketing-BS, because this totally overclocks the chip. See https://linux-sunxi.org/H3

And although the datasheet of the Allwinner H3 is open source, Allwinner is known for repeated GPL-violation and usually only provides closed source binary-blobs for their hardware (the same applies to the ARM Mali graphics), which are bound to a certain kernel version (which makes them in fact not open source). So you are stuck to an outdated kernel version as opposed to the RasPi with its Broadcom-chips.  https://linux-sunxi.org/GPL_Violations

Also the datasheet for RPi1 SoC BCM2835 is available and the newer versions only differ in the CPU-module, the peripherals are identical.

Bottom line: Allwinner-chips should be avoided if possible. You pay less, but in the end the disadvantages of these chips are not worth it.

[rx8pilot]

I couldn't assemble these for $10. Is there a single genuine part on the board? Is it being assembled by slaves in a dungeon? I love a bargain, but this is so extreme  - someone is paying the price for these even if the consumer is not. I can't support any of this, humans and businesses are being taken advantage of. Crap.?

[nctnico]

I couldn't assemble these for $10. Is there a single genuine part on the board? Is it being assembled by slaves in a dungeon? I love a bargain, but this is so extreme  - someone is paying the price for these even if the consumer is not. I can't support any of this, humans and businesses are being taken advantage of. Crap.?

If you producte 50000 pieces (or way more) you can have an SMT assembly line run very efficiently. Ofcourse the wages are low in China but boards like these have to be assembled using machines since you can't really do anything manually except for feeding PCBs and components and taking the finialised product from the assembly line and put it in a box.

[rx8pilot]

I am not talking about hand assembly - I am talking about building the whole product with genuine parts here in the USA. Even with volumes of 50k (which seems high, but I don't really know) - there is no way I could build it for $10 much less make any profit. The assumption here is that they are building it for less than what they sell it for.

Fake parts and slave labor is the only way to make this happen. Honorable mentions: No QC, bad software support, mediocre sales chain, etc. If I had time, I would love to estimate the BOM cost and assembly labor using real parts. Is paying as little as possible so important to people that they are willing to support shitty companies like this?

[edavid]

Fake parts and slave labor is the only way to make this happen. Honorable mentions: No QC, bad software support, mediocre sales chain, etc. If I had time, I would love to estimate the BOM cost and assembly labor using real parts. Is paying as little as possible so important to people that they are willing to support shitty companies like this?

You have no basis for that accusation.  I don't know if it's racism or just ignorance, but it's completely unfair.

And speaking of ignorance, do you consider Broadcom to be an ethical company?  I know I would much rather support Allwinner, despite their GPL failings.

[roli]

I've been trying to find a decent single board computer/SoM for quite a while. I think I've now gone through quite a few boards (at least 5) and I still can't find a solution.

I would really like something with Wifi and BLE built in. RPI3 has it, but it is quite expensive and bulky. I want something in the size range of RPI zero with those features. Mostly because of the thickness. Something like that for around 10-20$ would be just awesome. I am sure that Pi guys can add those features to the Zero for around 5-10$ extra. But they probably won't any time soon.

I would make one myself, but getting a SoC in a QFP or other usable package is pretty much impossible. Except older Allwinners. BGA is sadly a no-go if you want a cheap PCB.

[rx8pilot]

You have no basis for that accusation.  I don't know if it's racism or just ignorance, but it's completely unfair.

What race are you referring to? I did not mention any race. Are you being racist? How is it unfair to be upset that a company miraculously is able to sell products at a cost that way below what it costs to build using genuine parts?

And speaking of ignorance, do you consider Broadcom to be an ethical company?  I know I would much rather support Allwinner, despite their GPL failings.

What does broadcom have to do with my comments? I have no idea what they do or don't do. If they use fake parts and slave labor to meet unrealistic consumer price demands - then I would not consider them ethical. I have no specific information other than history and an understanding of what it takes to manufacture things.

[Kjelt]

How is it unfair to be upset that a company miraculously is able to sell products at a cost that way below what it costs to build using genuine parts?

Fair or unfair I can attest from experience that if you want to buy a $5 / 1000 pieces microcontroller from a company , they will drop to $3 if you buy 100000 and even lower than $2 depending on the contract if you go into millions. So is that unfair? That is called business I am afraid and the small fish get eaten by the big fish.
Besides this , these parts can come from an overproduction run, second source (over date), prototyping who knows where it comes from and what they paid for it. Maybe they break even and just want to move the products that were in their inventory and make a profit of a few % instead of the tens of % that western companies like to make.

[rx8pilot]

Here is my prediction:

As long as the world consumers demand ultra-cheap crap and are willing to look the other way when manufactures cut costs by any way imaginable - we (society) will have to live with the long term consequences. The consequence is that, as consumers, will ONLY have shit to choose from. Shit A or Shit B. That is because the innovators and above board manufacturers will have no interest in the economy where they make $.01 profit per unit. The companies that innovate will deal with commercial, industrial, and ,military applications where they can get there money back.

Products were once designed to solve a problem. Now they are designed to be cheap. It is fun for  a while, until that is all we have. A race to the bottom.

[bbandi]

I think the Broadcomm soc on the raspberries are completly open now: http://arstechnica.com/information-technology/2014/02/raspberry-pi-marks-2nd-birthday-with-plan-for-open-source-graphics-driver/

While as others mentioned, even though you got datasheet for this allwinner, the quality of the drivers are pretty bad, probably because of the not-so-open-if-you-look-into-it documentation...

Also, it's not really fair to compare this brand new board to the previous generation of the pi, the pi3 is said to be about 50% faster, so it would came pretty close to this allwinner. However this board seems to have the lowest $/speed ratio.

[rx8pilot]

How is it unfair to be upset that a company miraculously is able to sell products at a cost that way below what it costs to build using genuine parts?

Fair or unfair I can attest from experience that if you want to buy a $5 / 1000 pieces microcontroller from a company , they will drop to $3 if you buy 100000 and even lower than $2 depending on the contract if you go into millions. So is that unfair? That is called business I am afraid and the small fish get eaten by the big fish.
Besides this , these parts can come from an overproduction run, second source (over date), prototyping who knows where it comes from and what they paid for it. Maybe they break even and just want to move the products that were in their inventory and make a profit of a few % instead of the tens of % that western companies like to make.

With that in mind - I am super curious what the worldwide demand for these things are. They are toys/learning platforms aimed at tinkerers and education.

What does it take (beyond hard volumes) to get the price down to where these could be built for maybe $7? PCB, connectors, assembly - everything has to be the lowest of the low cost. When cost is the primary driver for the product, the goal is to find the lowest cost parts. In my experience - knock-off's and counterfeits are the de-facto way to get the cheapest price on parts. Paying people $1/day is another way to save money. Regardless of how they are able to pull it off, it is reasonable to say that corners are being cut at every opportunity without any consideration for the long term consequence. Suppliers, sub-suppliers, and sub-sub raw material suppliers are all cutting corners to meet the low cost demand. Consumers love it because they get the $10 thing. A decade or two from now, and all we will have access to is old stale tech because the big innovators are always being undercut. Good luck world. If your target is hitting the bottom, it's a pretty big target and you are sure to hit it. With a thud.

[apis]

Here is my prediction:

As long as the world consumers demand ultra-cheap crap and are willing to look the other way when manufactures cut costs by any way imaginable - we (society) will have to live with the long term consequences. The consequence is that, as consumers, will ONLY have shit to choose from. Shit A or Shit B. That is because the innovators and above board manufacturers will have no interest in the economy where they make $.01 profit per unit. The companies that innovate will deal with commercial, industrial, and ,military applications where they can get there money back.

Products were once designed to solve a problem. Now they are designed to be cheap. It is fun for  a while, until that is all we have. A race to the bottom.

It's not because consumers demand ultra cheap crap, it's because when you are choosing what fridge to buy (for example) the average consumer have no idea about any non-superficial performance parameter, so they just go with the cheapest one that is big enough. Since no other parameter determines what the customer choose that is the parameter the producer will optimize.

In some markets there are monopolies/oligopolies so you get crap but you have to pay out of your nose anyway because you have no alternatives.

Problem is that if free markets are to be efficient policy makers need to deal with these kind of problems but they typically do not since it's not in the existing manufacturers best interest (and they are the ones with influence), instead they are asking for less regulation since that is what is in their owners best interest. The result is that we get more and more overpriced junk.

[rx8pilot]

It's not because consumers demand ultra cheap crap, it's because when you are choosing what fridge to buy (for example) the average consumer have no idea about any non-superficial performance parameter, so they just go with the cheapest one that is big enough. Since no other parameter determines what the customer choose that is the parameter the producer will optimize.

Well stated. It becomes a slow decent that is not noticed until everything looks, smells, and performs like shit. We are moving down with the illusion of moving up - best magic trick in history.

[Artlav]

A supercomputer cluster?
Sorry, but that is a bad idea.
I got one, out of 6 PIs, and it measures 100 times slower on most tests and real tasks than a single quad core core i7.
For the same money (~US $200-ish with all the loose ends) you can get 3 or 4 used Celeron G1610 systems, each about 20 to 50 times more powerful than the whole 6-PI build.

The $/FLOPS of these small boards is embarrassingly low.

[filssavi]

A supercomputer cluster?
Sorry, but that is a bad idea.
I got one, out of 6 PIs, and it measures 100 times slower on most tests and real tasks than a single quad core core i7.
For the same money (~US $200-ish with all the loose ends) you can get 3 or 4 used Celeron G1610 systems, each about 20 to 50 times more powerful than the whole 6-PI build.

The $/FLOPS of these small boards is embarrassingly low.

I guess what you are seeing here is the effect of Interconnects on RPi you have no way of connecting the nodes in a really efficient way, the Ethernet is connected through USB2.0 so taking inro account USB overheads, Ethernet, tcp/ip framing etc the usefull data throughput is not that great in fact if we look at real top clusters (like in the top 500) the majority uses infiniband fabric as Interconnects because it has a muchlighter network stack trading of flexibility with peeformances
Than there is then the software side of which i'm sot an expert but i know it has to be aware he's running on a cluster and acting accordingly and i don't know if raspbian has this kind of support

[f4eru]

One of the big problems of the raspberry is the crooked I2C : it does not support clock stretching, so it cannot talk to a microcontroller.
And broadcom refused to update the silicon, so it's stuck.

[Karel]

..., so it cannot talk to a microcontroller.

And UART or SPI is not a valid way to talk with an MCU?
Never had any problems with communication between RPI and an MCU.

[Fungus]

..., so it cannot talk to a microcontroller.

And UART or SPI is not a valid way to talk with an MCU?

And a lot lot easier than using USB, IMHO.

[altaic]

What does it take (beyond hard volumes) to get the price down to where these could be built for maybe $7? PCB, connectors, assembly - everything has to be the lowest of the low cost. When cost is the primary driver for the product, the goal is to find the lowest cost parts. In my experience - knock-off's and counterfeits are the de-facto way to get the cheapest price on parts. Paying people $1/day is another way to save money. Regardless of how they are able to pull it off, it is reasonable to say that corners are being cut at every opportunity without any consideration for the long term consequence. Suppliers, sub-suppliers, and sub-sub raw material suppliers are all cutting corners to meet the low cost demand. Consumers love it because they get the $10 thing. A decade or two from now, and all we will have access to is old stale tech because the big innovators are always being undercut. Good luck world. If your target is hitting the bottom, it's a pretty big target and you are sure to hit it. With a thud.

You're missing a big part of global economics in your rather cynical analysis. It costs a lot more to produce stuff in the US because the source materials bounce all over the world. In China, materials come from themselves, and, as such, Chinese manufacturers get special deals. Other nations do not usually get such deals. Your righteous indignation is misplaced: China manipulates their prices and their currency to encourage the world to rely on them; at any time, they could decide otherwise. They're successfully moving the world's manufacturing infrastructure into their control, and it will not be a good day when they decide to squeeze.

Tangential anecdote: A while back, there was an international investment shit storm which was precipitated by a leaked internal Chinese government document describing their outside investor policies. It translated as, "Let the dog in, beat the dog up, let the dog out." The Japanese were furious, but pretty much everyone was like

Back to the point: Cheap stuff is correlated to bad quality and slave labor, but is not necessarily causally related. And that's not even taking into account that the cost of living in China is absurdly low. Even describing the USD equivalent of wages there is bound to be fallacious in countless respects. That said, everyone buying cheap stuff from China may lead to a rather nasty outcome.

[knks]

What's next? Apple Pi? 

[NiHaoMike]

I am not talking about hand assembly - I am talking about building the whole product with genuine parts here in the USA. Even with volumes of 50k (which seems high, but I don't really know) - there is no way I could build it for $10 much less make any profit. The assumption here is that they are building it for less than what they sell it for.

Fake parts and slave labor is the only way to make this happen. Honorable mentions: No QC, bad software support, mediocre sales chain, etc. If I had time, I would love to estimate the BOM cost and assembly labor using real parts. Is paying as little as possible so important to people that they are willing to support shitty companies like this?

If you can wait for a deal, you can get a basic quad core smartphone for $10. Basically 1/4 the cost of a Pi 3 for more or less the same CPU capabilities.

[samnmax]

What's next? Apple Pi? 

http://hackaday.com/2016/04/01/apple-introduces-their-answer-to-the-raspberry-pi/

 

[Athanasis]

What's next? Apple Pi? 

Nerds don't bother with Apple crap because they like hacking, modifying, adding, and be marketable as well. So no, don't expect an Apple pi to establish itself. We already have Linux which is fully opened source and customizable. Apple crap is only for crappy assholes like dead Steve Jobs and soon all his history will end up with him in his grave.

[jancumps]

I take it you're not a fan?

[Artlav]

I guess what you are seeing here is the effect of Interconnects on RPi you have no way of connecting the nodes in a really efficient way,

Not exactly. I mostly use it for explicitly parallel problems, where there is no communication between nodes, and the results are kinda sad once you get over the coolness factor.

[Athanasis]

What's next? Apple Pi? 

Nerds don't bother with Apple crap because they like hacking, modifying, adding, and be marketable as well. So no, don't expect an Apple pi to establish itself. We already have Linux which is fully opened source and customizable. Apple crap is only for crappy assholes like dead Steve Jobs and soon all his history will end up with him in his grave.

It was only a joke. Don't do yourself any injury.

It was a joke but the Apple pi exists like u see above in alink

--------------------------
I keep track of your replies with my android devices using the app "Tapatalk".

[steverino]

What's next? Apple Pi? 

Nerds don't bother with Apple crap because they like hacking, modifying, adding, and be marketable as well. So no, don't expect an Apple pi to establish itself. We already have Linux which is fully opened source and customizable. Apple crap is only for crappy assholes like dead Steve Jobs and soon all his history will end up with him in his grave.

Count me as a crappy asshole. (actually, aren't assholes supposed to be crappy?)

[drussell]

What link?

The one from April 1, 2016 in post number 28

[Athanasis]

What's next? Apple Pi? 

http://hackaday.com/2016/04/01/apple-introduces-their-answer-to-the-raspberry-pi/

 

that link

--------------------------
I keep track of your replies with my android devices using the app "Tapatalk".

[f4eru]

..., so it cannot talk to a microcontroller.

And UART or SPI is not a valid way to talk with an MCU?
Never had any problems with communication between RPI and an MCU.

SPI does not have a properly ACKed handshake.
UART does not have proper handshake either.

So you have to make the communication as slow as your slowest interrupt on the micro.
And if you fail that timing, you don't have a warning of the corruption.
That's bad.

It can be avoided with i2C, but only if the master handles the handhake properly. The raspberry doesn't.

Conclusion : there is no safe way to transfer a bit more data from raspberry to a micro.

[Franz Zinn]

Though the data sheets are available which is more than can be said of the Broadcom equivalent, Allwinner is not Open Source compliant either. Please see http://linux-sunxi.org/GPL_Violations .
This is not to detract from any of the merits of the Orange Pi. It is just a statement of fact about the unfortunate state of affairs that there is as yet no fully open source, embedded board with a modern GPU. Though the situation is improving as there are open source video drivers being developed for both VC4 and Mali-400.

[Fungus]

Conclusion : there is no safe way to transfer a bit more data from raspberry to a micro.

No safe way ... apart from programming the micro correctly so it can keep up with the data stream.

UART does not have proper handshake either.

You could always send a special character to say "I'm getting full!" and a different character to say "OK, start sending again". 

[apelly]

UART does not have proper handshake either.

You could always send a special character to say "I'm getting full!" and a different character to say "OK, start sending again". 

Oh ^S it! You're killing me.

[4gent5mith]

If you go to the Orange Pi site and try to download the Kali image for the Orange Pi one it takes you to a mega. co. nz folder. Which has all images/md5. Strange part is that it has images they don't list on their site. I downloaded all images (6.81GB worth); this might be very interesting.

I ordered one so we'll see how it goes. I'll likely post my discoveries (if any) on the forum.

[arekm]

That board have much more features available in form of ... pads:

http://forum.armbian.com/index.php/topic/755-orange-pi-one-adding-analog-audio-out-tv-out-mic-usb-and-ir-receiver/

Data lines for USB #3: points 1,2
Data lines for USB #2: points 3,4
1 - IR receiver RX line
2,3,4 - MIC1P/MIC1N/MIC-MBIAS
5,6 - LINEOUTR/LINEOUTR
TVOut signal and MUTE

[f4eru]

No safe way ... apart from programming the micro correctly so it can keep up with the data stream.

Yeah.
Shure.
Ensure that a 20 MHz micro reacts fast enough compared to a 500 MHz SOC.
I plan to do more useful things with my micros than to wait for data from the host.
Especially when servicing time critical tasks like multiple regulation loops in the 1000-2000 interrupts/seconds range.
 

You could always send a special character to say "I'm getting full!" and a different character to say "OK, start sending again". 

Suuure.
Yeah.
Does not work here.
The micro cannot react and do SW handshake before corruption happens, because it has no Hardware buffer, so it needs to service each byte with an interrupt....


So the conclusion is still : i2C is the only option for a communication with a host including a positive hardware ack.... as long as you don't use a RPI

[NiHaoMike]

The microcontroller supports real time programming. You can limit the data rate on the Pi side so the buffer never overflows.

[Fungus]

No safe way ... apart from programming the micro correctly so it can keep up with the data stream.

Yeah.
Shure.
Ensure that a 20 MHz micro reacts fast enough compared to a 500 MHz SOC.
I plan to do more useful things with my micros than to wait for data from the host.
Especially when servicing time critical tasks like multiple regulation loops in the 1000-2000 interrupts/seconds range.

Ummm... last time I checked RS232 had baud rates plenty slow enough for a 20MHz micro to respond to and still have time left over for other things. It's all down to good programming.

You could always send a special character to say "I'm getting full!" and a different character to say "OK, start sending again". 

Suuure.
Yeah.
Does not work here.
The micro cannot react and do SW handshake before corruption happens, because it has no Hardware buffer, so it needs to service each byte with an interrupt....

I don't know what 'micro' you're referring to. The micros I use (AVR based) have a one byte buffer in hardware which gives you plenty of time to respond to an interrupt. I believe this is a common in other chips, too.

If your micro doesn't have a 1-byte hardware buffer you could toggle a data pin in the interrupt handler when, that lets the sender know you're ready for the next byte and would hardly slow things down at all (what's the interrupt response time on a 20MHz micro? Less than a microsecond?).

Give me a few seconds and I could come up with half a dozen other ways to make it work, too. It's all down to what I call "good design and programming".

So the conclusion is still : i2C is the only option for a communication with a host including a positive hardware ack.... as long as you don't use a RPI

Or... you can use I2C and rely on clock stretching for flow control (assuming your 'micro' can do clock stretching in hardware).

I don't get your assertion that a Raspberry Pi can't do I2C. "I2C master" is one of the easiest protocols to do in software.

[hukuzatuna]

Than there is then the software side of which i'm sot an expert but i know it has to be aware he's running on a cluster and acting accordingly and i don't know if raspbian has this kind of support

I have a 128-core cluster of Adapteva Parallella boards, Raspberry Pi 2s, and Raspberry Pi 3s. I use this for developing prototype distributed AI code. Raspbian supports MPI, though I don't use the Debian package. I install MPICH-3.2 and use that. I've also gotten Apache Spark to run on the cluster with a lot of customizing the Java runtime environment to squeeze the heap into that tiny 1 Gig chip. It's an interesting experiment, but not particularly performant because of the Java ecosystem running on limited hardware. MPI, however, does have reasonable performance (within the limitations of the hardware). Once I get things working on this cluster I'm pretty comfortable moving it into a distributed environment using much larger machines.

Cheers,
Phil

[Okabe]

We should make a EEVblog SETI group XD

[System Error Message]

if you're interested in many core clusters TILE has an interesting one which facebook uses as PCIE cards in their servers to handle the processing bits such as firewall and web server. Their dev machines arent cheap but you can get one with 8x36 cores in a 1U config with the usual stuff you find in servers like lots of ram slots and also even has its own PCIe slot too just like a regular x86 server. I have one of these many core CPUs in the router i've  been using, never goes beyond 4% under the worst case (internet + firewall + filtering and proxy + unoptimised management + network services, no hardware acceleration used) and much cheaper than any consumer router in the price/performance, beats ARM A7s and A9 by a large margin in per core and per clock throughput. Best part is the TILEs tend to come with 10Gb/s NICs while the raspberry pi or pi clones have 100Mb/s, only 1Gb/s for some and thats only on the usb2 interface rather than PCIe.

As far as SETI computing power goes if it ever became available for these routers it would be a huge improvement as ISPs have been buying them so it is quite funny when you tell your ISP you have the same router they use in their edge gateway or router.