Home > AMLogic, Android, Hardware, Linux, Ubuntu > The First Amlogic S912 Development Board is Coming Soon with Khadas VIM2

The First Amlogic S912 Development Board is Coming Soon with Khadas VIM2

We have a decent choice of Amlogic S905 development boards like ODROID-C2 or NanoPi K2, but I was recently asked whether I knew of any Amlogic S912 development boards. I’m sure Amlogic has one for internal development, but those are hard to get, and probably expensive, and while you could probably get an S912 TV box board those lack I/Os, and software support may truly be a challenge. So I’m pleased to announce that Shenzhen Wesion will soon provide an update to their Khadas VIM Pro board with Khadas VIM2 powered by Amlogic S912 octa-core processor.

Click to Enlarge

The company will actually offer three variants of Khadas VIM2 boards (Basic/Pro/Max) specifications with highlights in bold showing differences with Khadas VIM Pro board:

  • SoC –  Amlogic S912 octa core ARM Cortex-A53 @ up to 1.5 GHz with ARM Mali-820MP GPU
  • System Memory
    • Basic – 2 GB DDR4
    • Pro/Max – 3 GB DDR4
  • Storage
    • micro SD card and 2MB SPI flash
    • eMMC Flash – Basic: 16GB; Pro: 32GB; Max: 64GB
  • Video & Audio  Output – HDMI 2.0a up to 4K @ 60 Hz with CEC support
  • Connectivity
    • Basic – Gigabit Ethernet with WoL support, 802.11 b/g/n/ac and Bluetooth 4.1 via Ampak AP6356S module
    • Pro/Max – Gigabit Ethernet with WoL support, 802.11 b/g/n/ac with RSDB and Bluetooth 4.2 via Ampak AP6359SA module
  • USB – 2x USB 2.0 host ports supporting 900mA and 500mA loads, 1x USB 2.0 type C port supporting power and data only
  • Expansion header
    • 40-pin 2.54mm pitch header with USB, UART, I2C, ADC, PWM, I2S, SPDIF, and ISO7816
    • 10-pin FPC connector with I2C and IOs
    • 8 “pin” pogo pads array with USB, I2C, DVB bus, and I/Os
  • Misc – Blue LED, white LED, dual channel IR, power/function/reset buttons, header for RTC battery, fan header
  • Power Supply –  5V to 9V via USB type C, 4-pin VIN 1.25mm pitch header, or pogo pads for VIN (5V recommended for better efficiency); programmable current limit switch up to 4A (Set to 3A by default)
  • Dimensions – 82.0 x 57.5 x 11.5 mm (4x M2 mounting holes)

We can see that it’s not just a processor update with many new features added to the new boards. If like me, you’ve never heard about RSDB, it stands for Real Simultaneous Dual Band, and allows to use both 5 GHz and 2.4 GHz at the same time, while for most dual band modules only one frequency can be used at a given time. That’s a clear advantage if you’re going to use the board as an access point.

Click to Enlarge

The company will provide Android 7.1 Nougat and Ubuntu 16.04 or greater operating systems and SDKs for the board, and work on UEFI support is in progress. The board will be launched last that month, and for now, the only VIM2 specific documentation provided is Amlogic S912 datasheet, but you can be sure there will be a lot more on Khadas Docs page at launch, although I expect many of the instructions available for Khadas VIM (Pro) will still work on VIM2 board.

You’ll find more details on the announcement forum post, such as Linux OpenGL ES not working natively, i.e. without libhybris and Android libraries, and the board has been designed with micro servers in mind with features like WoL and SPI flash for network boot, as well as UEFI support.

Thanks to Geokon for the tip.

  1. Hoangdinh86
    July 2nd, 2017 at 11:40 | #1

    Sadly, no Usb 3.0

  2. July 2nd, 2017 at 11:41 | #2

    @Hoangdinh86
    That’s because there’s no USB 3.0 interface on Amlogic processors for now.

  3. Mr Netflixer
    July 2nd, 2017 at 12:08 | #3

    without libhybris (GPU drivers) and Android libraries

    This is a major problem for Linux/LibreELEC images on the AML S912 – there are hacks out there to provide for GPU video output but this leads to GUI tearing.

  4. KiLLWarez
    July 2nd, 2017 at 14:13 | #4

    Price?

  5. July 2nd, 2017 at 14:36 | #5

    @KiLLWarez
    We’ll have to wait a few more days or weeks for that.
    But if we’d want to speculate, Khadas VIM2 Basic should sell for around $70 based on the specifications above. The VIM2 Max version could be close to $100. Those price should include shipping, and USB C cable.

  6. theguyuk
    July 2nd, 2017 at 15:37 | #6

    I was just reading the szwesions 912-development-board page, it notes they have a lite version with s905D but sadly only for ODM

  7. kszaq
    July 2nd, 2017 at 15:54 | #7

    @cnxsoft Linux hardware video decoding works natively, the only thing libhybris is needed for is OpenGL.

  8. blu
    July 2nd, 2017 at 16:08 | #8

    I always wondered how HardKernel can provide GLES support without libhybris, and others cannot..

  9. tkaiser
    July 2nd, 2017 at 17:47 | #9

    Nice details (at least 2T2R MIMO and 5GHz Wi-Fi, WoL, no brain damaged powering scheme) but why’s there no huge heatsink by default or at least mounting holes for one, what’s that thing looking like a SATA connector on the lower PCB side and how does it work to provide 2 type A USB receptacles, USB2 also on the type-C connector, also on the 40 pin header as well as pogo pins?

    Prices will start at $80 or even more?

  10. kszaq
    July 2nd, 2017 at 17:51 | #10

    @blu
    Hardkernel doesn’t have S912 board. Every S905(X) board has GLES support without hybris. S912 GLES in Linux is not possible without hybris because Amlogic did not pay for Mali DDK for Linux, they only have Android license.

  11. Slawek
    July 2nd, 2017 at 19:07 | #11

    No USB 3.0 and native sata port.
    Its a prank Bro !

  12. July 2nd, 2017 at 19:30 | #12

    Hi, everyone, this is Gouwa from Khadas Team.

    Any questions about Khadas VIM2 can be asked here or Khadas Forum, and I will try to answer all the questions 🙂 .

    The retails price needs to wait for the final launch in the next weeks.

    Thanks CNX-Software for the article.

  13. July 2nd, 2017 at 19:39 | #13

    @tkaiser

    1) I’m not sure where you figure out, but there is no SATA connector on VIM2.
    2) It’s reversion1.0 showed on the picture, only 8-Pin pogo pads on the 1st version, and the final release version will be 20-pin Pogo pads, 7-pin pogo pad for a on-board programmable MCU, and a 6-pins pogo pads for extra power sources.
    3) There are two solutions for heat sink: use silica gel directly, or design a specified heat sink for VIM2 which mount at the four M2 screw holes.

  14. Willy
    July 2nd, 2017 at 19:46 | #14

    Looks possibly interesting. It might be difficult to fix a heat sink but it’s nice to see a board with a “wide” range of input voltage. Letts see how much it costs ; for quite some time the NanoPI-M3 has been available with 8 A53 at 1.4 GHz (real), and it seems that octocores of 1.4 GHz or more still remain rare. The low profile form factor is also much appreciated.

    @tkaiser: it’s not a SATA connector, it’s the other side of the low-profile ethernet jack.

  15. tkaiser
    July 2nd, 2017 at 19:58 | #15

    @Gouwa
    When you run cpuburn-a53 on all 8 CPU cores… how low does S912 throttle down after 5 minutes? Below 600MHz?

    And what about the USB2 ports? How many host and OTG ports does S912 provide and how is USB ‘routed’? Can i use USB on all receptacles, GPIO pins and pogo pads simultaneously or not?

  16. July 2nd, 2017 at 20:18 | #16

    @kszaq
    What I mean by not working “natively” in Linux is that there aren’t any Linux drivers, and instead you need to use libhybris + Android libraries/drivers.

  17. July 2nd, 2017 at 20:41 | #17

    @tkaiser

    As still Sunday here in China, I will do some simple testing tomorrow and update here.
    I think users should install a heat sink (actually, a 3-level cooling fan is also available on VIM2) when full on all 8 cores.

    For Amlogic S912, which is different with S905(D/X), the SoC featured with dual channels USB2.0 and one OTG2.0. Yes can be simultaneously.
    USB details for VIM2:
    1) OTG2.0 is on USB-C port
    2) one USB2.0 for USB Host port(with 900mA and closed to USB-C port one)
    3) another USB2.0 channel is leaved for a four-ports USB HUB, and for USB-Host Port, 40-Pin GPIO, 20-Pins Pogo-Pads.

  18. tkaiser
    July 2nd, 2017 at 20:47 | #18

    @Gouwa
    Nice to see that you expose one USB host port directly so it has not to share bandwidth (many many other board designs fail with this). Looking forward to some ‘thermal performance’ numbers tomorrow 🙂

  19. Paul
    July 3rd, 2017 at 01:25 | #19

    Gouwa :
    Hi, everyone, this is Gouwa from Khadas Team.
    Any questions about Khadas VIM2 can be asked here or Khadas Forum, and I will try to answer all the questions ….

    Can you give us some typical 802.11n data transfer performance figures in Client mode ?

  20. kszaq
    July 3rd, 2017 at 03:47 | #20

    @cnxsoft
    I know what you mean and I’m pointing out that hardware video decoding works natively in Linux for S912, like it does for any other Amlogic SoC – using libamcodec.

    GLES is the only thing you need libhybris for.

  21. July 3rd, 2017 at 09:04 | #21

    @kszaq
    Oh thanks. I thought both GLES and VPU libs had to go through libhybris.

  22. tkaiser
    July 3rd, 2017 at 13:11 | #22

    Paul :
    Can you give us some typical 802.11n data transfer performance figures in Client mode ?

    That’s close to impossible since it depends on so many factors and most of these are external (the environment the device operates in and the antennas used which will add significantly to costs if you choose appropriate ones) 🙂

    Maximum throughput numbers can be calculated but always keep in mind that PHY rate is not directly related to any real world throughput, see http://80211notes.blogspot.com/2014/03/phy-rate-and-udp-throughput.html for example.

    But as soon as distance increases or walls are in between all depends on antennas and TX power settings (adjustable or not?). But since we could use 2T2R MIMO robustness is better and with appropriate antennas positioned appropriately also performance will be much better compared to 1T1R implementations. Speaking about real-world performance: Your AP must also support at least 2T2R, 5GHz and also beamforming.

  23. blu
    July 3rd, 2017 at 13:17 | #23

    @kszaq
    Ok, I understand. Amlogic need to address that for their own sake. Let’s see when that will happen.

  24. Mum
    July 3rd, 2017 at 14:04 | #24

    > micro SD card and 2MP SPI flash

    @cnxsoft typo here. I assume you mean 2MB (16MBit) SPI flash?

  25. Stephen
    July 3rd, 2017 at 18:55 | #25

    Did the max memory spec for the S912 change since last year? The spec was 2GB, but the pro/max has 3GB.

  26. tkaiser
    July 3rd, 2017 at 19:04 | #26

    @Stephen
    Nothing changed. By googling for ‘S912 3GB’ or something like that you’ll get a lot of TV boxes targeting clueless people.

  27. July 3rd, 2017 at 19:08 | #27

    @Stephen
    In the first leaked specifications, memory was shown to be limited to 2 GB, but eventually it changed, and many Amlogic S912 TV boxes now come with 3GB RAM.

  28. July 3rd, 2017 at 23:35 | #28

    @tkaiser

    As I can’t paste picture here, so I create a topic in Khadas Forum for the CPU thermal testing:
    http://forum.khadas.com/t/khadas-vim2-thermal-testing/852

  29. crow
    July 4th, 2017 at 04:59 | #29

    @kszaq
    but not with mainline kernel just with the 3.14.29 kernel right?

  30. July 4th, 2017 at 10:16 | #30

    @Gouwa
    It’s possible to post pictures using HTML code.
    The comment took time to be approved, as I don’t use my computer between around 21:00 to 9:00.

  31. Paul
    July 4th, 2017 at 10:19 | #31

    tkaiser :

    Paul :
    Can you give us some typical 802.11n data transfer performance figures in Client mode ?

    That’s close to impossible since it depends on so many factors ….

    I realise performance depends on many factors, antenna gain, position, if-line-of-sight, distance & interference environment, 802.11 MCSrate etc. However, websites like ‘SmallNetBuilder’ routinely conduct meaningful comparative tests of commercial Wireless Router performance down to N300/$21, showing graphs of transfer rate v distance/attenuation. Such websites usually have a page describing the test environment. To minimise difficult to measure variables of crowded environment, SNB use screened room and simulate Distance using an Attenuator, so for example, on 2.4GHz the RT-AC68U achieves about 100Mbps downlink from short to medium distance, then dropping to 26Mbps Uplink, 11Mbps downlink before dropping out at 63dB Atten. Other test sites attempt approximate real-world conditions showing large home plan view with Reference router and Router under-test in different rooms; other sites try line-of-site to 3 to 5 different distances usually in uncrowded location.

    My point is, if websites provide useful comparative test results for boxed routers, (sadly, as yet don’t test cheaper unboxed WiFi boards), it should be possible for Gouwa/Khadas to provide 2.4GHz performance measurements in some specified environmnet, be it screened room, a building with qty of drywalls, floors, or in the open, line-of-site. As with SNB, test setup can be described further, eg throughput of Reference Router, Channel 6, HT20, OS version/driver, Board-under-test has Bluetooth & other services disabled or default settings etc.

    In this case the photo of the Khadas board implies 2 External Antenna will be used so it would be reasonable for the maker to conduct a performance test using typical pair of 6dBi antennae Router makers often supply as standard. The Tx power setting is easily stated.

    Your link to factors explaining why UDP throughput is much less than theoretical at PHY was very interesting thanks.

    My own interest is connecting to public or semi-public APs at a distance too far for the Intel WiFi in my miniPC to reach, but where my phone WiFi can just about reach the nearest AP. I am certain all of the nearby public AP’s are 2.4GHz only (5GHz would be less range surely) and almost certain HT20 GI 800ns with no beamforming & being public, out of my control, so perhaps throughput no better than MSC7 (65Mbps) ? I already have 1 x 15dBi omni Antenna – not expensive from Aliexpress, and a spare pair of, I think 6dBi Antennae.

  32. July 4th, 2017 at 11:20 | #32

    @Paul

    Yes, I will do the WIFI performance testing, but still need wait for the final board arriving(few days later).

    Will update here then with the WIFI testing details.

  33. tkaiser
    July 4th, 2017 at 14:12 | #33

    @Gouwa
    Thank you but

    is not a real test but just a ‘load illusion’. I was asking for cpuburn-a53 for a reason: since this is the heaviest ‘CPU only’ load generator I’m currently aware of. Another pretty nice variant to test for the efficiency of heat dissipation is cpuminer (since once throttling occurs it shows performance numbers — this is not meant as a generic benchmark but only to check for the efficiency of a cooling solution). And I would also recommend to immediately install RPi-Monitor in case you want to explore further thermal behaviour (it’s something completely different looking at numbers a scripts spits out or getting nice graphs you can analyze later).

    I was also surprised that S912 seems to be a big.LITTLE design? If that’s the case it makes reliable performance testing much worse, see here for example https://discuss.96boards.org/t/odroid-xu4-cortex-a15-vs-hikey-960-a73-speed/2140/26 (there are also a lot of external ressources mentioned like cpuminer and a nice 64-bit NEON optimized Linpack suitable to tune dvfs settings).

    Anyway: Since idle temperature is 22°C above ambient temperature and with such a light load as above it gets already close to throttling tresholds shipping with a huge heatsink seems mandatory to me. BTW: fans blowing from the side over a heatsink should be a lot more efficient. Just let cpuminer show you, installation/use is very easy and outlined in the link above.

  34. July 4th, 2017 at 16:58 | #34

    @tkaiser

    Thanks for the professional feedback & suggestions, I will check out the approaches you mentioned to do further testing.

    At the same time I will push our team to design a specified heat sink for both VIM and VIM2.

  35. tkaiser
    July 4th, 2017 at 17:07 | #35

    Paul :
    My point is, if websites provide useful comparative test results for boxed routers, (sadly, as yet don’t test cheaper unboxed WiFi boards)

    Those websites collect absolute numbers within given/reproducable environments to provide listings and graphs (and hopefully some insights too). The only meaning of their absolute numbers for any reader is the ability to do a relative performance ranking afterwards most probably already ignoring some of the prerequisits for those numbers (eg. 3T3R MIMO being basic requirement on both client and AP for the numbers shown).

    So while such overviews using always the same environment testing through a bunch of different devices provide some useful numbers (since allowing for relative comparisons) I doubt that’s possible when someone starts to test his own device. Maximum throughput numbers need no tests since the result of simple calculations and everything else depends on too much else.

    So wrt relative numbers it’s easy. If you can choose between 1T1R and 2T2R you always choose the latter (of course needed on both sides of the connection to be of any use), single vs. dual band is also a no-brainer and if both AP and client support RSDB that’s also great since enhancing performance and range if both sides of the connection make use of it.

    But for your special use case (public AP far away) I doubt you get any meaningful numbers when Khadas does some testings. If I were you I would get the cheapest 3 antenna 2.4Ghz AP allowing to run LEDE/*WRT and able to be configured as client bridge or even wireless repeater.

    BTW: there’s a good reason to not test ‘cheaper unboxed WiFi boards’ since average readers are not willing to accept reality anyway and always think about misusing SBC as wireless access points. Even those better dual-antenna/dual-band Wi-Fi chips on some Android devices and now also on some SBC like the VIM2 here are not made for AP mode (crippled firmware limiting max client connections, no beamforming).

  36. tkaiser
    July 4th, 2017 at 17:26 | #36

    @Paul
    Almost forgot since you mentioned ‘Antenna – not expensive from Aliexpress’. I also ordered a few some time ago and did some quick testing immediately thinking having found just another ‘buy cheap buy twice!’ category. But to be honest: I still lack the knowledge to interpret the numbers so I postponed all of this to get more insights first. But since we started to replace 10GbE at customers with 40GbE in the meantime dealing with Wi-Fi feels broken anyway (same with wired networks below Gigabit Ethernet of course).

  37. tkaiser
    July 4th, 2017 at 17:59 | #37

    @Gouwa
    Good to hear that you’re exploring a heatsink option. Maybe it would be also an idea to think about two heatsinks with different scenarios in mind: heatsink only and heatsink + fan. For passive cooling the fin distance is important since if too tight convection won’t work. Such a heatsink like on the MACCHIATOBin should work pretty nice with both scenarios and a top mounted fan while classical heatsinks with fins might work better with a lateral fan. When designing a heatsink to be used together with a fan I always would prefer a large and slow rotating over the small annoying ones (but that’s obviously just me, many tinkerer love leds blinking and fans making noise).

    Back to software/testing: the great thing about cpuminer is that it generates a lot of heat so you can easily test efficiency of hardware (heatsink) and software (cpufreq/dvfs settings) and that it also provides real performance numbers. So if you prevent throttling (killing all CPU cores except one) you can easily walk through all available cpufreq OPP and once you get no better khash/sec values reported by cpuminer you know that performance does not further increase (thinking about the ‘Amlogic desaster’ last year with all cpufreqs above 1.5GHz being faked or RPi 3 still cheating on us and reporting incorrect clockspeeds when throttling occurs).

    Hardware (different heatsinks, convection only or combined with some airflow) performs differently and you can use cpuminer to reliably test for this (since producing constantly performance numbers and you need to run some tests for a longer time since otherwise thermal effects aren’t noticeable). But same applies to software too. Inapproriate throttling settings eg. only jumping between 1.5GHz and 1.0Ghz when throttling treshold temperature is reached show worse performance than better throttling settings allowing to use as much cpufreq/dvfs OPP as possible (not jumping only between 1.5GHz and 1.0GHz but choosing a stable 1.3GHz for example).

    You might want to look into http://linux-sunxi.org/User:Tkaiser#Reliability_testing_on_Pine64 since there’s a simple daemon collecting ‘minerd –benchmark’ numbers able to be fed into RPi-Monitor with an appropriate template. This helped as a lot improving throttling settings on a few platforms and I think it’s also a great help when providing a PWM controller fan with variable speed.

  38. Paul
    July 4th, 2017 at 19:32 | #38

    Gouwa :
    @Paul
    Yes, I will do the WIFI performance testing, but still need wait for the final board arriving(few days later).
    Will update here then with the WIFI testing details.

    Pleased to hear that. Can you also advise which way the 2 external Antennae are organised, eg both for WiFi 2.4 & 5GHz, or 1 for Bluetooth, 1 for WiFi 2.4GHz + 1 for Wifi 5GHz etc.

  39. July 4th, 2017 at 20:50 | #39

    @Paul

    The one close to USB port is for WLAN & Bluetooth shared, and the other one is for WLAN only.

  40. tkaiser
    July 4th, 2017 at 21:23 | #40

    @Paul
    Adding to @Gouwa’s answer: you get 2×2 MIMO in both Wi-Fi bands as with all other AP635x with some theoretical potential for interferences on one antenna due to Bluetooth’s adaptive frequency hopping spread spectrum implementation (IIUC).

  41. Paul
    July 11th, 2017 at 10:02 | #41

    tkaiser :

    Paul :
    My point is, if websites provide useful comparative test results for boxed routers, (sadly, as yet don’t test cheaper unboxed WiFi boards)

    But for your special use case (public AP far away) I doubt you get any meaningful numbers when Khadas does some testings. If I were you I would get the cheapest 3 antenna 2.4Ghz AP allowing to run LEDE/*WRT and able to be configured as client bridge or even wireless repeater.
    BTW: there’s a good reason to not test ‘cheaper unboxed WiFi boards’ since average readers are not willing to accept reality anyway and always think about misusing SBC as wireless access points. Even those better dual-antenna/dual-band Wi-Fi chips on some Android devices and now also on some SBC like the VIM2 here are not made for AP mode (crippled firmware limiting max client connections, no beamforming).

    Thanks for your very helpful thoughts on the subject. I realise the SNB & other test site environments are unlikely to match typical use environments, but test results at least, as you admit, provide a RELATIVE performance, which I think helps users to pick a model for their needs. Without that test data, the only other guide to POTENTIAL performance is, as you state, hardware spec, be it 1T1R, 2T2R, internal or external antenna, WiFi chip etc.

    Although I thought buying a Wireless Router a bit OTT, cost and energywise, purely to extend the range of my miniPC WiFi connection to reach some distant public APs, as I was told a modern Wireless Router would do it, in my ignorance I did in fact buy a ASUS RT-AC58U some time back on account of its spec showing long range & low energy consumption, not knowing then, that it didn’t support ‘Client mode’ or after discovering my error, didn’t support OpenWRT either !! I was able to confirm the AC58U took about 5W and with its 4 Ext. Ant. was sensitive, able to ‘see’ 3 times more APs, than my phone, but the UI had no method of connecting to them as a Station.

    Re. your suspicion the VIM2 may not do the job either due to ‘not made for AP mode (crippled firmware limiting max client connections, no beamforming)’, remember I primarily want ‘Client, not AP mode’ – the onward connection to my miniPC can be via Ethernet. If the VIM2 can support simultaneous ‘Client + AP mode’ that would be a luxury, as I would probably setup a 5GHz AP. in which case I don’t envisage more than 3 local clients. Do you think the VIM2 likely to allow this ?

    @cnx I hope you will receive a board to test soon ?

  42. July 11th, 2017 at 10:33 | #42

    @Paul
    I have not been contact to get a board yet. The next board’s reviews will be “Le Potato”, and then Rock64.

  43. tkaiser
    July 11th, 2017 at 12:18 | #43

    @Paul
    Wrt ‘not made for AP mode (crippled firmware limiting max client connections, no beamforming)’ — this was not mentioned with your use case in mind but what happens in general if people see an SBC with better Wi-Fi specs. Many immediately think such a SBC would make up for a great AP which is not the case. Any cheap dedicated router able to run a free OS (LEDE/*Wrt) will do a better job.

    The only ‘client implementations’ I know of that perform rather well as AP are Atheros mPCIe cards (AR9380 or better — you get them for 10 bucks on eBay) with ‘unlock_atheros_regulatory_restrictions.patch’ applied.

  44. July 26th, 2017 at 22:35 | #44

    @tkaiser

    VIM2 Wi-Fi Performance testing is available here:
    http://forum.khadas.com/t/vim2-wi-fi-performance-testing/922

  45. July 26th, 2017 at 22:36 | #45

    @cnxsoft

    Hi, cnxsoft, we will contact you in these days for VIM2 review 🙂

  46. Paul
    July 29th, 2017 at 08:20 | #46

    Gouwa :
    @tkaiser
    VIM2 Wi-Fi Performance testing is available here:
    http://forum.khadas.com/t/vim2-wi-fi-performance-testing/922

    Thanks for the link to WiFi Performance testing. I didn’t realize you also have Internal Antennas. At first I assumed your 81.5 Mbps on 2.4GHz used some specified External Antennas until I saw your note at the bottom “If with external Wi-Fi antenna should be with better Wi-Fi performance.” Were these results obtained with the ‘Basic’ (AP6356S) or the ‘Pro’ (AP6359SA) ?

    Can you publish 2.4GHz results using a pair of typical 5dBi External Antennas and let us know which version of the VIM2 is used ?

    If you will be supplying a mains adapter, can you also let us know what the energy consumption at the wall, (2nd best, into the board) both at ‘Idle’, and when WiFi is transfering at maximum data rate ?

  47. tkaiser
    July 29th, 2017 at 15:42 | #47

    @Gouwa
    Thank you. Output from ‘iwconfig’ on VIM2 and ‘wdhelper’ (or screenshot of settings when you option-click on the Airport menu item) would help to understand settings. I would assume your router only allowed HT20 with 2.4GHz since otherwise iperf would be better?

    Paul :
    Can you publish 2.4GHz results using a pair of typical 5dBi External Antennas and let us know which version of the VIM2 is used ?

    What do you expect? More than 90 Mbits/sec when HT20 with 2×2 MIMO is used in 2.4 GHz band? Better antennas should only make a difference if distance will be increased.

  48. Paul
    July 31st, 2017 at 06:51 | #48

    tkaiser :
    @Gouwa

    Paul :
    Can you publish 2.4GHz results using a pair of typical 5dBi External Antennas and let us know which version of the VIM2 is used ?

    What do you expect? More than 90 Mbits/sec when HT20 with 2×2 MIMO is used in 2.4 GHz band? Better antennas should only make a difference if distance will be increased.

    As the link to their test results showed distance to router only 7metres I am sure you are right. But when I looked in Gouwa’s link for the test conditions to get 81.5Mbps, I saw the note “If with external Wi-Fi antenna should be with better Wi-Fi performance.” – I took him at his word to ask how much better.

    Refering back to our conversation in this thread 4-11 July, in my use-case of using a cheaper unboxed WiFi board to extend the range of my miniPC, as range is my problem, clearly I need a board with External Antennas to extend range as far as possible. If the VIM2 can function in ‘Client mode’, has the range and data rate, comes in at an attractive price, then it could be the solution I seek. At my current location where I seek to connect to distant public APs, there is only 1 broadband provider, whose fastest dataplan 35Mbps. Using my phone WiFi to nearest public AP tethered to miniPC on USB3, the best I ever saw was a peak of 48Mbps just once, usually peaks below 32Mbps. So in my case, any cheap WiFi board approaching 81.5Mbps in 2.4GHz TCP/IP ‘Client mode’ is more than enough. I am not knowledgeable about typical public AP characteristics, but as stated in my post #31 assume worst-case HT20 GI 800ns, no beamforming. Certain there no 5GHz AP in range.

    Your post #43 recommended mPCIe cards Atheros AR9380 or better. My miniPC is a v. compact NUC difficult to modify to use Ext. Ant. & prefer not swap out the standard fit Intel 7260HMW card. (I never discovered whether it was the NUC Int. Ant. or card responsible for weak WiFi). I have just 1 spare USB port, so rather connect by GbE or WiFi. (I guess 5GHz best as VIM2 has RSDB) For above use case have you any thoughts on the ‘Basic’ (AP6356S) or the ‘Pro’ (AP6359SA) Wifi chip performance, perhaps in relation to the AR9380 you mentioned ? I gather the overall transfer rate from remote AP thru the WiFi board to the PC, be it via GbE or 5GHz RSDB, will be lower than the 81.5 Mbps iperf result. Is this reduction likely to be significant for a VIM2 class of product ?

    Re. #35 your suggestion to buy “cheapest 3 antenna 2.4Ghz AP allowing to run LEDE/*WRT and able to be configured as client bridge”, I got put off dedicated routers after my big mistake buying the RT-AC58U. Having missed the fact most Routers don’t support ‘Client mode’, I chose the AC58 on account of its low energy consumption yet long range. Having read many router reviews before buying the AC58, most consumed more power than my NUC, and many a huge size -not good for travellers. I looked in OpenWRT HW Table – a dauntingly long list with no way to filter on the criteria of interest: 2.4GHz range & energy consumption. Have you any suggestions on where to hone in on a ‘cheapest 3 antenna 2.4GHz LEDE/*WRT capable Router’ with consumption <6W ? My Android 6 phone easily configures as a WiFi Client to my PC. As the VIM2 comes with Android 7.1 surely simplest solution is to also configure as a WiFi Client to my PC ?

  49. Paul
    July 31st, 2017 at 07:35 | #49

    @Paul
    Last question written in a hurry. Using same approach as on my Android phone assumes there is a local UI on the WiFi board – in practice a keyboard & monitor, which I prefer not to. One reason I prefer a low energy consumption, is for economical 24/7 running, so I could connect a KB & monitor briefly to set it up, or maybe there is a boot time config file to execute the commands required to run in WiFi Client mode ?

  1. No trackbacks yet.