If you’re mainly using your Android mini PC or STB to stream videos over the web or your local network, Wi-Fi performance is actually more important than CPU performance, and I’ve recently had quite a few issues with Wi-Fi with both T428 (RK3188 / Broadcom Wi-Fi module) and CS868 (AllWinner A31 / Realtek Wi-Fi module), which were only fixed after upgrading my router firmware. So I’ve decided to re-test most of the mini PCs I own with the upgraded router firmware, and the number are highly confusing, and I found out there’s probably be no easy to way to estimate the overall Wi-Fi performance of any device, and it’s most probably highly dependent on the router used, and its firmware version.
My router is TP-LINK WR940N, a 300 Mbps Wireless N router with three antennas. Since I bought it, I had never upgraded the firmware (3.9.18 Build 100104 Rel.36350n) from 2010. When I test Android mini PCs or media players, I always place the device is the same area which is about 5 meters from the router, but not line of sight, and I transfer a 278 MB file from a SAMBA share in a PC running Ubuntu. I use ES File Explorer to perform the transfer, and I check the time with a stop watch. I write the time, and calculate the throughput. In theory, the NAND flash write speed could also be a bottleneck, but considering the throughput are always less than 3MB/s, and the file is only 278MB, which is less than the amount of RAM, I believe it should not impact the results.
Here are the results I got with the old router firmware since I started doing this test.
Matricom G-Box Midnight MX2 is the fastest by a wide margin, and T428 and CS868 have pretty poor performance because they often lost the signal during the transfer.
As I explain in CS868 review, upgrading the router firmware completely changes the performance of CS868, and improves T428 significantly. So equipped with my “new” router with firmware 3.13.9 Build 120201 Rel.54965n, I tested the devices again to see if results would be much different, and if we could find some patterns, e.g. Realtek Wi-Fi module perform better or not. I performed the file transfer three times on each device, to make sure there was not too much variation between transfers.
The updated results show CS868 far ahead, while G-Box Midnight MX2 is not that good anymore, and may have been negatively impacted by my router firmware upgrade. That’s the complete opposite of earlier results.
The table below shows more details about the tests and the devices.
| Product | Time 1 | Time 2 | Time 3 | Throughput (MB/s) | Connection Speed / Signal Strength |
Wi-Fi Module | Firmware / Build Number | Remark |
| Hi-802 (Freescale I.MX6) | 03:32 | 04:33 | 04:25 | 1.112 | 150Mbps (Good) | Realtek RTL8188CUS | R13.4-rc2 V1.65.3 | Slow microSD, high CPU usage (around 6 to 7) |
| G-Box Midnight MX2 (AML8726-MX) | 04:20 | 03:44 | 03:44 | 1.178 | 150Mbps (Excellent) | Realtek RTL8188EUS | JOP40D.20130422 | Highly Variable speed (Up to 2.6 MB/s) |
| T428 (RK3188) | 03:02 | 03:15 | 03:32 | 1.416 | 65Mbps (Excellent) | Broadcom AP6330 | Rk31sdk-eng 4.2.2 JDQ39 eng root 20130419 | Highly variable speed (Up to 1.7 MB/s), lost signal once, and cancelled test |
| Tronsmart Prometheus (AML8726-MX) | 03:05 | 03:10 | 03:08 | 1.481 | 150Mbps (Excellent) | Realtek RTL8188EUS | JOP40D.20130307 | Constant speed around 1.5MB/s, high CPU usage (4 to 5) |
| MK908 (RK3188) | 02:47 | 02:30 | 02:47 | 1.723 | 150Mbps (Excellent) | Broadcom AP6210 | mk908_100j1 15-04-2013 | Constant speed around 1.7MB/s |
| CS868 (AllWinner A31) | 01:30 | 01:30 | 01:28 | 3.112 | 150Mbps (Good) | Realtek RTL8188EUS | fiber_onda809q9-eng 4.1.1 MASTER eng ok 20130506 | Constant speed around 3MB/s |
| Measy U2C (RK3066) | 02:40 | N/A | N/A | – | 135Mbps (Excellent) | Mediatek MT5931 | B12_RK3066_Public_V411_121224 | Worked the first time, but then failed to transfer the file completely 4 times, so I gave up |
I dropped E-Great H5 and Droid Stick A2 from the tests due to lack of time, and Measy U2C is not included into the chart because it could not reliably transfer the file, and I only managed to fully transfer the file once, over 5 attempts. This also happened once with T428, but the 3 other tests worked. As I previously assumed, the connection speed, be it 65 Mbps, 135 Mbps or 150 Mbps, does not actually matter. The conclusion of those tests is a bit disappointing, as the only one I can make is that you can’t really make conclusions. If I had only use the original router firmware, I would have said CS868 Wi-Fi is a disaster, but after firmware upgrade of the router, CS868 is the best ever mini PC I’ve ever seen when it comes to Wi-Fi throughput, and is clearly ahead of any other devices. Bear in mind CS868 only has a chip antenna, and can still easily outperform other device with an standard internal antenna.
With regards to Wi-Fi modules, Realtek RTL8188EUS is both at the top of the chart and the bottom with the new router firmware, so the overall performance seems to depend on overall system design, and the interoperability with the router. One of the advantage of devices based RTL8188 is that the drivers source code is available, Mediatek MT5931 source has recently been “leaked”, but I don’t think Broadcom IC source is available.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress
Personally I had such difficulties with my sticks, but all changed after first moving from the included power supply to take power out of my TV, and after seeing the both Wifi signal speeds and quality skyrocket (and never lose contact after that) I moved to use high-quality USB high-current hub as my “extension cord” for all my 5V devices, there has never been any problems since, even the overheating from overclocking doesn’t seem to freeze the sticks anymore… All my sticks are registered as an 150Mbps devices, but my router indicated they are really only 65Mbps standards compliant… works… Read more »
great comparison
@anon
I’ve had not had any issue with mini PCs with regards to lost Wi-Fi signal for nearly one year of use, and I only started to see the problem many other people experienced, when I started to use T428 and CS868.
A router firmware upgrade fixes it for some, and makes it worse for others. But trying it use a better power supply might be a good idea.
Power supply and taking antenna out of box. It’s easier than it sounds. 1st open the case using screwdriver or knife starting from HDMI plug and moving clockwise. Using a knife or drill make a hole on the side of case. Get antenna out of case. Get much better Wifi reception. Also – try testing 10 or 15 meters away from router with walls on the way – that will be more realistic.
@bravomail
I already have walls on the way, but my router is a bit closer than the 10 or15 meters. I’m quite convinced getting the antenna out of the case with my setup will bring zero benefit. Maybe if you are in an area with many signals it can help, but where I live there’s only one signal. The best Wi-Fi device from the test above only has a chip antenna, and it easily beats all other devices with a proper internal antenna. E-Great H5 already has an external antenna, but Wi-Fi is weak.
I think that the problem is the test, SMB is the worst bandwide(or throughput) test posible, you must isolate latency (rtt) and throughput, also layer 2 error is important. in all measure avg, min, max and standard deviation are useful to understand where is the problem.
Also try isolate components in the transmition system to find measure change, if you change many components at same time(soc,wifi module, OS kernel version, driver version, network stack configuration) you dont find the responsability of individual component.
Thanks, your articles are always very useful.
@Klein Yes, I’ve been thinking about SAMBA being an issue. But after I thought the throughput will just be lower than HTTP or NFS, and it’s probably the main way people stream videos from the local network (TBC), so it may actually be more relevant than other methods. For some system the throughput is pretty good e.g. 2.5MB/s for a while, then it suddenly collapses to 200 KB/s (but the Wi-Fi signal still seems OK), and sometimes it goes back up to complete the transfer. I don’t think it’s due to SAMBA. Maybe I’ll try some wget to /dev/null transfers… Read more »
Are you sure, that those AP6[4|3|2|1]xx WiFi modules are made by Broadcom?
Poking around the web, I couldn’t find any reference linking those to BC.
On the other hand however, there are now indications, that AP6493, AP6441, AP6476, AP6383, AP6330, AP6376, AP6210 and AP6181 are manufactured by ampak.com.tw, a subsidiary of gemtek.com.tw, even though, none of those products can be found on either company’s website.
@LinuxMint
No I’m not sure, but that’s what the guys’ at GeekBuying told me. Other resellers also write the same stuff, but they often just copy/paste the specs, so if one makes a mistake it’s spread all over the internet.
One reader also mentioned ampak, but at the time I looked I could not find AP6330 nor AP6210. The company however does manufacture Wi-Fi modules, so it could be possible ampak makes the modules based on Broadcom chipsets. Not quite sure.
@LinuxMint
The document below (from AMLogic) seems to confirm the module is AMPAK AP6330
https://www.dropbox.com/s/1gywbkqdx8repx6/AMLOGIC.pdf
Many things are hidden the log and wlan.ko (renamed to rk31xx or rockchip), but I can find this in T428 log:
dhd_bus_select_firmware_name_by_chip: firmware_path=/system/etc/firmware/fw_RK903_ag.bin
<4>[20236.166795] Dongle Host Driver, version 1.28.13.1.1 (r)
<4>[20236.166805] Compiled in drivers/net/wireless/bcmdhd on Mar 18 2013 at 10:57:53
bcmdhd looks like broadcom and a search for RK903 gives:
RK903 (BCM40183 – Broadcom 4330 Based Chipset)
So this is should be our chipset: http://www.broadcom.com/products/Wireless-LAN/802.11-Wireless-LAN-Solutions/BCM4330
@Jean-Luc Aufranc (CNXSoft)
This also means Bluetooth 4.0 in AP6330 should support High Speed, but not “Low Energy”.
thanks for this test, although i am more confused now than before. 🙂 i would love to see another test like this with more full size boxes (like the Geniatech, Mele, etc) and maybe using a greater distance to the router, and also using something like speedtest.net APK for the testing (for upload/download tests)… i have noticed huge improvement when boxes have an external antenna, and my house is prone to these issues with my router being at the opposite end of the house with many walls between..
@avirex The router is basically in the middle of the house, so I’d have to move the TV and all accessories outside to test greater distances 🙂 I would not use application such as speedtest.net to test Wi-Fi performance, unless I have stellar Internet access, and in my case it would just test unreliable Internet speed instead of Wi-Fi performance. I had one dongle with an external antenna once, and it performed poorly. Actually the best device I have, with regard to Wi-Fi performance, does not even have an actual antenna as you may think of it, but just a… Read more »
@Jean-Luc Aufranc (CNXSoft)
Thank you for getting to the bottom of it.
That was some awesome detective work on your part 🙂