I have been using my home server as wireless router for a while, and this issue has been around with me for a while too. I got the TL-WN881ND v2 from TP-Link as it was the most robust option I could find for sale in my region, and started using it as my home personal usage access point. This is a PCI-E rtl8192ee-cg device.
However, with the rtlwifi rtl8192ee driver, the AP suffers from tremendously slow performance, especially on download speeds (from the point of view of a client device). This in turn makes this driver's AP unsuitable for anything other than very light browsing. Here are some results:
(as measured from a very close device in line of sight with a clear airspace)
Link Power Save ON: download - 105 kB/s, upload - 1194kB/s
Link Power Save OFF: download - 105 kB/s, upload - 1194kB/s
reference, ethernet connection feeding the server
download - 91786 kB/s, upload - 38255 kB/s
It's always the same speeds... which is very odd. I get very similar results with the default firmware and the "_new" firmware. At this point it feels like I have tried everything under the horizon.
There is one thing that works to make this card more performant as an AP and it is to use this driver: github.com/lwfinger/rtl8192ee .
It lets the AP work at about 50Mbps download on the client device, which is half of the wired speed but still less than advertised by the device. However I would like to ditch that driver in favor of the kernel driver, because the kernel driver has features I am interested in that the alternate codebase one doesn't.
On 6/14/20 10:30 PM, yagoplx wrote:
> I have been using my home server as wireless router for a while, and this issue has been around with me for a while too. I got the TL-WN881ND v2 from TP-Link as it was the most robust option I could find for sale in my region, and started using it as my home personal usage access point. This is a PCI-E rtl8192ee-cg device.
>
>
> However, with the rtlwifi rtl8192ee driver, the AP suffers from tremendously slow performance, especially on download speeds (from the point of view of a client device). This in turn makes this driver's AP unsuitable for anything other than very light browsing. Here are some results:
> (as measured from a very close device in line of sight with a clear airspace)
>
>
> Link Power Save ON: download - 105 kB/s, upload - 1194kB/s
> Link Power Save OFF: download - 105 kB/s, upload - 1194kB/s
> reference, ethernet connection feeding the server
> download - 91786 kB/s, upload - 38255 kB/s
> It's always the same speeds... which is very odd. I get very similar results with the default firmware and the "_new" firmware. At this point it feels like I have tried everything under the horizon.
>
>
> There is one thing that works to make this card more performant as an AP and it is to use this driver: github.com/lwfinger/rtl8192ee .
>
>
>
> It lets the AP work at about 50Mbps download on the client device, which is half of the wired speed but still less than advertised by the device. However I would like to ditch that driver in favor of the kernel driver, because the kernel driver has features I am interested in that the alternate codebase one doesn't.
Half of the wired speed is probably the maximum that you should expect.
I cannot really speak for Realtek, but I would not expect a lot of current
development for a 7-year old wifi device. My role in maintaining these drivers
is to fix crashes, not to improve performance as I do not have any knowledge of
the chip internals. Your best bet would be to add the new features to the GitHub
driver. I would be happy to review any such proposed changes and merge them if
appropriate.
Larry