Banana Pi BPI-R2 Open Source smart router with MTK 7623N design

Hello all.

Introduction begin I’m orange pi user so I’ve bought an Orange pi plus. I felt wondered at its abilities such as Gigabit Lan, mali400, DVFS, shutdown and reset buttons and other good things. But something gone wrong during the testing. Turned out that Gigabit Lan is connected thru USB2.0 (400Mb/s), mali400 has NO driver on SUPPORTED OS based on Linux, Hardware H265 video decoding wasn’t provided (H3 SoC supports it of course). DVFS does’nt work on little frequencies (There is some hiddend DVFS with higher priority). The board buttons are not programmed.

That is not a H3 computer that supports linux or Gigabit Lan (same success is to solder butterfly knife there - still knife but useless). What we (OPI buyers) get is deception. Introduction end

I really hope you making product that meet its description. I want to buy it. So please let us buy your product - show us it’s performance

  • show us Mali work on linux
  • show us hardware video decoding H264 and H265 on linux
  • show us power consumption at different run levels
  • show us working of programmed buttons

Otherwise Your product does not costs its parts, ICs. I dont want buy Jet engine on horse cart for price of jet liner, in this case price must meet its abilities - price of horse cart.

Thank you.

It seems you tried the wrong OS images? Everything is working when you use ‘Ubuntu Desktop’ from here: https://www.armbian.com/orange-pi-plus/ (most users stopped using the ‘official’ OS images for their Bananas and Oranges long ago)

Thank you for trying to help me, Charles. Directly from your link: Mainline kernel images (all boards) (next/dev branches)

  • No Mali drivers
  • No hardware accelerated video decoding
  • schedutil CPU governor may cause clicks and pops on audio output – change(d) to ondemand to work around this issue


Really I did not use this image, There may work DVFS and board buttons. I guess that is not OPI developer’s work.


But the main problems are still relevant:

  • theoretically maximum speed of Gigabit Ethernet is about 50MB/s
  • no mali drivers
  • no hardware H265 decoding
  • theoretically maximum speed of SATA is usb2.0 speed same as ethernet fail

So no any software will be able to fix Gigabit Ethernet and SATA connection thru USB2.0.

I hope Banana Pi developers will not do such scam.

Yes, you’re using still the wrong image as already said. Choose “Ubuntu desktop – legacy kernel” and you’re done. Gigabit Ethernet is fast but SATA is slow since it’s not SATA but GL830. Start reading from ‘Beware’ here: http://linux-sunxi.org/Sunxi_devices_as_NAS#Requirements_.2F_which_device_to_choose

I hope Banana Pi developers will not do such scam.

They did, look at BPI M3: same slow/defective GL830.

Can you please provide target date and price?

end of this month will public sale .and price is about 85 USD

Banana pi BPI-R2 run ubuntu linux with kernel 4.4.59 test

Unbelievable. Instead of showing anything relevant (boot log, at least some performance numbers for Ethernet and SATA or any other important stuff) you show a boring video without SATA/PCIe connected and a terminal with just a ping running.

Do you really call this a ‘test’? Where are the kernel sources? Where is schematics? You don’t have them right? Even the BLOBs you got from MTK are still outdated (4.4.43-BPI-R2-Kernel.tgz)

This will gonna be the same fiasco clusterfuck as with R1 in the beginning.

Haters gonna hate. What is the “fiasco” with r1? Works well for me, an issue with switching “during boot” is as paranoid as aliens attack for use cases of this device (home, smb, small branch router/non critical server/pbx) )) SATA performance? Really? I don’t think putting the production database here will be the best practice or what else use case do plan and care for SATA performance? 2sinvoip - good job, use several r1, waiting eagerly for an updated version.

For anyone else making the same mistake: The R2 is NOT an upgrade but an entirely different product backed by a different SoC manufacturer (MTK and not Allwinner) and backed by a different developer community (this time NONE vs. linux-sunxi community before).

It took Sinovoip a year to release schematics for R1 while at the same time a lot of stuff simply did not work at all.

And now two and a half years later 3 community projects included all the knowledge and fixes (Bananian, Armbian and David Bentham’s OpenWRT) but the whole initial try&error show has been forgotten and people laugh about users that went through this unbelievable situation back then and have a reason to ask for schematics (without it you’re more or less lost when dealing with hardware issues) and ask for sources (without it you’re more or less lost to dealing with software issues).

Good luck to anyone happy with a closed source product (no schematics, no sources) without any developer community around! Keep in mind: community fixed ALL issues with R1 back then, Sinovoip did nothing to help.

Which software do you run on your R1?

For the moment Raspbian

  • -fw(iptables)/nat/switch/ap(hostapd)
  • -ipsec site to site and client vpn(libreswan)
  • -pbx(asterisk) with usb gsm dongle(chan_dongle)
  • -some “smart home” self made gpio software
  • -video recording from IP cameras (yet only implementing, even in my most budget solution already ordered several external usb SATA bays with external power for ~15 bucks per slot, even had no thoughts of using internal sata - 2.5 inch TB+ are more expansive then 3.5 and latest need 12v anyway, one hdd is definitely not enough for most reliable persistence task and you will definitely not use fast 2.5 ssd on on-board port at least maybe some big data lab…).

community fixed ALL issues with R1 back then, Sinovoip did nothing to help.

I thought that why we have the community and hardware manufactures (both low (Allwinner/MTK) and high (Sinovoip) levels). Actually I had no major unresolved issues with the configuration above (at least those I know about) and I think it is more then enough for hw costs. We may also blame the distributor for not contributing. Reverse engineering and closed schematic is crap, yet we see the fact - r1 exists and works…

Just IMHO, maybe I am wrong, you seems like an experienced guy in sbc just don’t understand such hate for Sinovoip :slight_smile:

I don’t need any test video R2 is running Linux. If i want a mini ARM PC, i will buy another board. I need you show me how this work as AP mode runs OpenWrt/LEDE firmware, is it OK?

Which kernel?

3.4.112-sun7i

Welcome to current R2 situation

It is not yet released. So if I understand you correct - it was totally the same situation with R1 and in several years we now have a perfect home router/server with only some geek problems like no VLAN port mapping during reboot and slow onboard SATA which nobody use anyway.

Why so many haters in this community. If you aim using OpenWrt firmware why bothering waiting R2 - you can use R1 or any other Asus or whatever.

They can’t even provide boot logs so how should they show you any functionality besides a ping and usless desktop environments?

  • Have you seen a single time a wifi antenna connected to this board?
  • Have you seen on any picture Ethernet cables connected to both the WAN port and one of the switch ports?
  • Aren’t you worried that on the pictures at the top of this thread not two Ethernet cables are connected to this board and no wifi antenna but there’s an USB wifi dongle inserted (they ‘test’ their board without using its core functionality)?

If you know these guys a bit longer you know that they haven’t the slightest problem to sell unfinished products with 100% broken software. That’s why it is so important to make people aware and force SinoVoip to show boot logs (they can’t since dmesg will still show TONS OF ERRORS proving that software isn’t working properly), release kernel sources (they can’t since they did not pay/sign the MTK NDA so they have no sources) and release schematics to check for the usual hardware flaws these guys introduce each time they release a new board.

But potential customers are happy to listen to the same marketing chitchat over and over again. :joy:

will update image soon. for BPI-R2 will ready hardware 2.0 this month.

we coworker with MTK , now we update to linux 4.4.70, and all driver working fine .:slight_smile:

all source code all have update to github, and just private for now , when we ready ,will public it .for we now need optimize and test driver.

here you can see log.this is you need ??

Yes, thank you. So we know now at least a little bit:

  • /cpus/cpu@0 missing clock-frequency property means cpufreq not working yet? What’s the clockspeed the cores are running with?
  • An ASM1062 is providing SATA functionality. Have you or MTK already explored the NCQ problem when accessing two disks in parallel. Does it need echo 1 > /sys/block/sdX/device/queue_type to slow the disks down when 2 disks are accessed in parallel? Have you ever tested with more than one disk?
  • sda is a 32 GB SDCZ80 USB3 thumb drive. 93.3 MB/s write performance looks ok until you look how fast this device is in reality: http://usb.userbenchmark.com/SanDisk-Extreme-USB-30-32GB/Rating/1466 (so your dd test is xxx, you test buffers not the device and you miss either conv=fdatasync or oflag=dsync flag)
  • sdb is a 320 GB SATA disk connected to ASM1062. 65.2 MB/s is horribly low given that you test wrongly and the thumb drive scores better
  • it seems WAN port is eth0 and LAN ports appear as single eth1 port (or vice versa). Random MAC-Address with both interfaces. Will you fix this?
  • Starting LSB: daemon to balance interrupts for SMP systems... really bad since does nothing and contains a memory leak eating up all the RAM over time (your Ubuntu rootfs from 2 years ago already contained this flaw, reported several times in forum, happily ignored every time)

If you would not have hold back this log for almost half a year this feedback could’ve arrived earlier (useless anyway, you ignore community feedback usually). Now imagine if you would immediately open your github repo to allow third parties to look through R2 kernel sources to improve stuff?

But it wouldn’t change anything since even if community tries to help you you simply ignore it. Here a long time linux-sunxi community member took your kernel sources for this unfortunate BPi M2U and improved it in many ways: https://github.com/dan-and/BPI-M2U-bsp What did you do with this valuable information?

Ha ha, if it have 2 blank mini Pcie slot BPI-R2 will be better in alot of situation. I know SOC has 3 pcie lane but R2 use only 2, one for sata chip We can chose MT7603 and MT7612 for Dual band wifi (cheaper than MT7615

Sorry i did not read boot log i will do it now.

Ok. I read

i read both datasheet of ASM1062 & ASM1061 when i read about Witi Board, ASM1062 can use one or two lane PCIe, ASM1061 use one. But two lane for ASM1062 dont need because of limitation of gigabit port. Correct me if i’m wrong.

Exactly! ‘what change ? ?’ What’s the difference between 1.0 and 1.1? Why hardware had to be changed and in which way? These are the questions (you’ll never answer of course).

Look how hardware manufacturers that really follow ‘open source’ principles do this: Every single revision gets documented even if it’s just an internal prototype: https://github.com/OLIMEX/OLINUXINO/tree/master/HARDWARE/A64-OLinuXino

From time to time progress will be published, discussed with customers/community to eliminate mistakes and to pick up good ideas to improve the product: https://olimex.wordpress.com/?s=A64-OLinuXino

@sinovoip claims doing ‘open source hardware’ but does only the opposite. Everything is closed. No real open source hardware documents published, not even schematic, not a single line of code so far, this whole ‘open source’ babbling is Just Not True. But fortunately for you, your customers are happy with that. They love to be told ‘wait and see’ over and over again. You’re the most lucky guy on earth with such customers :joy: