I’ve been using the BPI R2 as a router for a while now and it works great apart from the low WiFi performance.
While in ap mode bridged with eth0, it is very unstable with hostapd. If I restart hostapd for example I have to restart the whole WiFi with:
echo 0 > /dev/wmtWifi and then echo A again.
On WiFi I get max about 34 Mbps which is very low compared to my TL-WR1043ND wifi router which has about 100 Mbps.
Maybe it’s my configuration for hostapd so here it is.
Note: The comented out line is what I’ve tried and it didn’t work at all.
When you say “unstable” you have to say which Kernel you are using and which os. Maybe you can try it with another Kernel (if using 4.14 try with 4.4) because i cannot guarantee, that i did not make a mistake while porting…i’m no professional developer…
I hope that mtk-team is working on a better wifi-driver…this android-driver is buggy
I have the same issue using kernel 4.14. As you refer to eth0 i guess you are using kernel 4.4?
I take for ever to connect and after a while the AP wifi disappear. I have to restart the whole WiFi with: echo 0 > /dev/wmtWifi and then echo A again.
The issue is in driver, what ever kernel you use. BPI should fix the driver for MT6625L.
Did you manage to have better performance using a USB wifi adapter.
I am also looking at some alternative.
I tried this adapter: Tenda w311u+ and I couldn’t get it to work because of the drivers and the lack of time I have right now. I will try it again later with the same adapter and a TP-Link TL-WN722N.
I think everybody should realize that Mediatek never intended the built-in Wifi chip to be actually used as an AP or even a client. It was meant to be a companion Wifi system to sniff the airwaves to assist an external dedicated Wifi module with channel selection,etc.
Here is the original press release when the MT7623 was announced:
Note that they talk about a “Built-in WiFi sniffer and background scan function to boost MT7615 4x4 11ac […]”
In my opinion, the MT7623 is a really good and powerful router chip. If you want fast and stable WiFi, get a decent WiFi module to go along with it. They are not expensive. That’s what the mPCIe slot is for.
I agree with your opinion, and I will suggest to buy qca9880(3x3ac), mt7612(2x2ac) or mt7603(2x2n) because you can find upstream (ath10k) or open source driver(mt76)
If you want to make dual-band ap (bgn+ac) you need 2 cards. I wanted to use internal for bgn and 7612 for ac…i dont want to use additional (3rd wifi-device) usb-dongle to realise that
The MT7615 can be configured to run in dual radio mode with 2x2bgn and 2x2ac operating simultaneously. This would give you dual band functionality with a single module. Not sure about Linux driver support though.
It is a Router-board and a home-router have to support wifi-ap the last 15 years…if today a routerboard have listed wifi onboard, EVERYBODY expects that ap is working and stable enough for home usage…
If anybody buy a routerboard like this, i guess he has a bit knowledge about local networks…no “dumb” home user will buy a routerboard where he must configure the full system and everything network related. Target-users are more or less experienced with networks and linux. User who only want to access the internet but have no knowledge about the basics will never buy such board.
I do not expect that it is usable for professional usage, but basics like wifi should work stable!
Please check if it is hardware-Problem or (only) a driver issue. And please don’t scare off customers with a Message like that. We can’t accept this “workaround”.
Private customers won’t buy this board either. They go into a store and buy DLink, TP-Link, Netgear etc. People who buy this board typically have a certain level of expertise.
These days state of the art WiFi means having multiple meshed APs around your house, each with dual radios and possibly 4x4ac. This board was never designed to do that and it was never advertised as such.
I agree, WiFi should work to some degree, but what I’m saying is that it’s not realistic to use it as an AP. As a general rule: with very few exceptions, any chip that’s NOT connected to the host CPU via PCIe, is not primarily designed to be an AP.
Those non-AP chips, if they run in AP mode, they can often only handle a very limited number of connected clients, typically 5 - 10 clients max. And that’s a hardware limitation in these chips.
I do not need currently a meshed wifi and clients are below 5 simultanously. This have to work stable. I don’t expect perfect performance and throughput near maximum. But stability is a must. Restart wifi every day for example is a no-go.
Understood. I don’t want to pick a fight and I sincerely hope that MT6625 WiFi will work for you one day. I really appreciate the work you have done so far and as I’m working on bringing VyOS to this platform I have already made good use of your Wiki page with useful links surrounding this platform and the MT7623 CPU. Thanks for putting this together!
I’m only aware of the MT7615 and another one from Broadcom (forgot which one). No open source driver support for either one. I think Moore Liu further up this thread correctly summarized your options for decent APs with proper driver support.
If you want dual radio support with this platform, IMHO you best bet is to use a RT5572 based USB dongle combined with an ath10k mPCIe module. The RT5572 has halfway decent AP support and can handle 30+ clients.
I just want to support Frank’s argument here. I’ll set my sights even lower though. AP mode ought to work, it was listed as supported. A lot of people don’t need mesh or anything fancy. I am using a Guru plug as an AP currently. It covers my office and living room just fine. I only my laptop, my cell phone, and my wife’s Ipad as usual clients. I really don’t anticipate any more.
That isn’t a “high bandwidth” application. Right now though I am not having a great deal of success getting anything stable. I also have not been working on that to hard other than just seeing that I could get a quick test connection up and running. Still got bigger fish to fry with this thing.
OK, I guess I should clarify. I never intended to give Sinovoip cover for not providing a proper AP driver (yet?). They should. Frank’s argument is valid.
However, people need to be realistic. This is not a very widespread WiFI chip, so there are only a limited number of people willing and capable of working on this. Not sure if Sinovoip has their own kernel people. Ideally Mediatek should put some effort behind it to fix this, but it’s probably not very high on their list of priorities. To complicate matters, WiFi AP driver source code is still treated as ‘proprietary’ by many vendors.
So rather than bemoaning the fact that it’s not working, it might be better to have a ‘plan b’ in place. Unfortunately this wouldn’t be the only WiFi chip that is only really usable as a client.