Test Edimax EW-7822UAC and NETGEAR A6200
#1
Hi guys I have bought these two wireless adapters and tested them with Openelec and OSMC:

Edimax EW-7822UAC Wireless AC1200 (Realtek RTL8812AU chipset)
http://www.amazon.com/gp/product/B00BVRQ...ge_o00_s00

NETGEAR A6200 Wireless AC1200 (Broadcom BCM43526 chipset)
http://www.amazon.com/gp/product/B00UA98...ge_o00_s00

These are the speeds I managed to achieve:
Openelec 5.95.5

Netgear A6200 - Not recognized

Edimax EW-7822UAC
5GHz 6-6,3 MB/s
2.4GHz 3,8-4,1 MB/s

OSMC 2015.08-1
Netgear A6200 - Not recognized

Edimax EW-7822UAC
5GHz 5-5,8
2.4GHz 1,8-2,1

Unfortunately NETGEAR A6200 was not recognized, I guess no drivers at this point?

The transfer rate with Edimax EW-7822UAC is still pretty slow in comparison to the transfer rate I get when I use it with my Windows 7 laptop.

The most unfortunate thing is that at this point I can almost safely say that there are no wireless adapters that can achieve speeds of 10+MB/s on Openelec or OSMC.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#2
(2015-09-17, 15:20)vdb86 Wrote: Unfortunately Edimax EW-7822UAC was not recognized, I guess no drivers at this point?

Presumably you meant the Netgear - unless this situation has improved there's unlikely to be any progress.
Texture Cache Maintenance Utility: Preload your texture cache for optimal UI performance. Remotely manage media libraries. Purge unused artwork to free up space. Find missing media. Configurable QA check to highlight metadata issues. Aid in diagnosis of library and cache related problems.
Reply
#3
You are correct, I'll fix that.
That is quite unfortunate, I though Broadcom was a bit better than Realtek.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#4
I've done additional tests using iperf.
These are the results using iperf on Raspberry Pi 2 (60 sec, TCP window size: 43.8 KByte (default))

Ethernet: 673 MBytes 94.1 Mbits/sec

Edimax EW-7822UAC 2.4GHz: 431 MBytes 60.2 Mbits/sec

Edimax EW-7822UAC 5GHz: 887 MBytes 124 Mbits/sec

Edimax EW-7811UTC 2.4GHz: 308 MBytes 43.1 Mbits/sec

Edimax EW-7811UTC 5GHz: 829 MBytes 116 Mbits/sec

These are the results using iperf on slow Windows 7 laptop (Intel atom CPU N450, 2GB Ram)(60 sec, TCP window size: 63.0 KByte (default))

Edimax EW-7811UTC 2.4GHz: 266 MBytes 37.1 Mbits/sec

Edimax EW-7811UTC 5GHz: 402 MBytes 56.1 Mbits/sec

Edimax EW-7822UAC 2.4GHz: 364 MBytes 50.9 Mbits/sec

Edimax EW-7822UAC 5GHz: 502 MBytes 70.2 Mbits/sec

This tells me that the drivers are not the ones to blame (at least not 100%), speed depends on the rest of the hardware too, I guess some USB controller and whatever else.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#5
Sam Nazarko managed to fix up the RTL8812au driver a bit, transfer speeds with Edimax EW-7822UAC are considerably better than with OE.
Ethernet transfer speed is also somewhat better.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#6
disappointing, seems like the built in 100Mbit or a USB Gbit adapter (up to the max USB2 bus speed) is the way to go where possible.
I have mine using the built-in ethernet, glad to know that the only thing I'd be missing by going to 802.11ac is a wire and $30.
Thank you for doing the tests and posting the results.

Here are my tests for ethernet and Gbit USB adapter for those who may be wondering about that route. I used your same IPERF2 settings, (60 sec, TCP window size: 43.8 KByte (default))
This is from a RPI2 to a local Core I5 Win8.1 file server. Results were consistent across 3 runs

RPi2 - Openelec - Millhouse Jarvis build #0919
Built in Ethernet: 674 MBytes 94.2 Mbits/sec
Plugable USB3-E1000: 1.31 GBytes 187 Mbits/sec

PC with same USB3 adapter in a USB3 port to file server for reference
Plugable USB3-E1000: 7.86 GBytes 884 Mbits/sec
(Edit) To quell my own curiosity, the same adapter plugged into the USB2 port of the same PC
Plugable USB3-E1000: 1.57 GBytes 225 Mbits/sec

Overclocking the RPi2 sdram made no difference, so it appears to be a bus speed limit not a memory speed limit.
After the reboot, speeds increased from 187 to 192 Mbits/sec over consecutive runs with sdram at both 450 and 500mhz.

I'd have expected the 802.11ac tests to be close to the RPi2 USB limit I appear to be hitting, at 187 Mbits/sec, even with a single antenna (433 Mbit/s.)
Reply
#7
Considering what most wireless N adapters used to give us, these are very nice results for wifi on a Pi. Enough for 1080i MPEG2 live TV or a raw bluray ISO (1080).
Reply
#8
@vdb86 can you check if you can play a raw BluRay without buffering in each of the cases?
You may want to try with/without buffermode=1 if the bandwidth achieved varies.

Playing raw BluRays reliably over wifi is a bit of a holy grail (on any Kodi platform). We get occasional reports of success but they seem hard to reproduce.
It would be nice to confirm if a specific adapter can achieve it.

Obviously this just says it is possible in some circumstances. Actual router used, thickness of walls, wifi congestion etc will affect results for others.
Reply
#9
@popcornmix
Sure thing, but I'd have to test it using OSMC, since Sam updated RTL8812au driver.
Transfer speeds with OE are noticeably slower.
Do you have any raw BluRay in mind?
I have a Avatar.2009.EXTENDED.1080p.3D.BluRay.Half-OU.x264.DTS-HD.MA.5.1-RARBG which is 26,7GB, would that work?

@J_E_F_F
Yesterday I used the 1TB NTFS HDD to test transfer speeds and then afterwards formatted it to ext4, of course I used Edimax EW-7822UAC.
These are the results (didn't do Iperf, just simple testing):
Openelec NTFS
Ethernet 124s-7,92MB/s
2.4GHz 268s-3,66MB/s
5GHz 158s-6,22MB/s

Openelec EXT4
Ethernet 101s-9,73MB/s
2.4GHz 292s-3,36MB/s
5GHz 180s-5,46MB/s

OSMC EXT4
Ethernet 95s-10,34MB/s
2.4GHz 133s-7,39MB/s
5GHz 131s-7,5MB/s

So basically moving from NTFS to ext4 helped ethernet to perform better, but screwed up wireless transfer speeds.
With OSMC everything performs faster, I don't know what Sam and the rest of the team did but it's pretty good.
Hopefully Openelec guys will be able to do something about it.

In the end I do believe that customized drivers (for Raspberry Pi 2) might improve the performans of some wireless adapters, but I doubt anyone's going to do something like that.
I snooped around linux drivers and it's overwhelming amount of work and testing, especially if they're kinda crap right from the start.
The main obstacle seems to be Raspberry Pi's hardware. Some hardware will perform better if you connect it to another usb port).
I haven't seen this documented anywhere but I tested it myself.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#10
(2015-09-21, 12:54)vdb86 Wrote: @popcornmix
Sure thing, but I'd have to test it using OSMC, since Sam updated RTL8812au driver.
Transfer speeds with OE are noticeably slower.
Do you have any raw BluRay in mind?
I have a Avatar.2009.EXTENDED.1080p.3D.BluRay.Half-OU.x264.DTS-HD.MA.5.1-RARBG which is 26,7GB, would that work?

It would be interesting although it looks like a (fairly high bitrate) re-encode.
Either a raw bluray iso file, or a remux (without reencode) of a bluray to mkv would be slightly more interesting as the bitrate is probably a little higher.

Quote:Some hardware will perform better if you connect it to another usb port).
I haven't seen this documented anywhere but I tested it myself.

As far as I'm aware, this is purely an RF issue, rather than a hardware/software issue.

The USB socket furthest from the centre of the PCB (e.g. top socket of top-right connector) is likely to give the best signal,
as the RF is less likely to be blocked by the sockets/PCB. This may vary with orientation of Pi and direction towards router.
Reply
#11
Quote:It would be interesting although it looks like a (fairly high bitrate) re-encode.
Either a raw bluray iso file, or a remux (without reencode) of a bluray to mkv would be slightly more interesting as the bitrate is probably a little higher.

I'll get Interstellar 2014 1080p BluRay REMUX AVC DTS-HD MA 5 1-RARBG, it's 42,71GB.
When I test this out I'll let you know.

Quote:The USB socket furthest from the centre of the PCB (e.g. top socket of top-right connector) is likely to give the best signal,
as the RF is less likely to be blocked by the sockets/PCB. This may vary with orientation of Pi and direction towards router.

I'm not talking about RF signal, I literally witnessed worse performance if I used lower left usb for hdd and upper right usb for wireless, there were other combinations but this was the most extreme one.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#12
(2015-09-21, 15:01)vdb86 Wrote: I'm not talking about RF signal, I literally witnessed worse performance if I used lower left usb for hdd and upper right usb for wireless, there were other combinations but this was the most extreme one.

How are you determining it is not an RF issue?
I'd say it's either RF interference (being further from processor reduces noise)
Or it's RF attenuation (if wifi has to pass through other metal sockets then it will be attenuated, so "middle" sockets are worse than "edge" sockets.
Reply
#13
Quote:How are you determining it is not an RF issue?
I'd say it's either RF interference (being further from processor reduces noise)
Or it's RF attenuation (if wifi has to pass through other metal sockets then it will be attenuated, so "middle" sockets are worse than "edge" sockets.
Because I use usb extension cable (15cm) to connect the wireless adapter which is always at the same position with a clear line of sight in regards to router.
Physically how ever I plug-in that usb cable should not create any interference.
Digital Clock Screensaver with a lot of options - check it out!
Reply
#14
(2015-09-21, 15:13)vdb86 Wrote: Because I use usb extension cable (15cm) to connect the wireless adapter which is always at the same position with a clear line of sight in regards to router.
Physically how ever I plug-in that usb cable should not create any interference.

I've talked to the experts and there is the four sockets are all identical in terms of how they are connected to the hub chip and the software will treat them identically.

How reproducible is this? Can you make sure the wifi dongle is in the same position, and repeat the experiment. Try to move as little as possible between tests.
Report the iperf number for each socket. Reboot between each test. Test the sockets first in order 1,2,3,4 and then in order 4,3,2,1.
Reply
#15
(2015-09-21, 16:16)popcornmix Wrote: I've talked to the experts and there is the four sockets are all identical in terms of how they are connected to the hub chip and the software will treat them identically.

How reproducible is this? Can you make sure the wifi dongle is in the same position, and repeat the experiment. Try to move as little as possible between tests.
Report the iperf number for each socket. Reboot between each test. Test the sockets first in order 1,2,3,4 and then in order 4,3,2,1.

Well I played with it yesterday, I'm pretty sure it's for real, although it didn't make sense to me either...
I'll test it out, but I don't know if iperf will show anything special since I feel that the problem is internal usb communication.
I'll have the test results if not today than tomorrow!
Sure thing
Digital Clock Screensaver with a lot of options - check it out!
Reply
 
Thread Rating:
  • 0 Vote(s) - 0 Average



Logout Mark Read Team Forum Stats Members Help
Test Edimax EW-7822UAC and NETGEAR A620000