Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

M1 USB Ethernet issues

I

iandk

macrumors regular
Original poster
Aug 10, 2019
162
65
Hey there!

So I just received my new M1 Air in addition to my M1 Mini and I'm supper happy so far.

But since I'm working quite a lot on my local NAS and the Air doesn't have an Ethernet port, I bought an UNI USB C 8 Port hub with an integrated Gigabit NIC.
Since the transfer speed was varying I tried installing the Realtek driver.

However I had to notice, that the driver seems to be outdated (only up to macOS 10.15) and the installer shows that it's specifically for the Intel Macs.

Does anyone know if there's new driver in the works or maybe an other USB NIC that's updated already?

My M1 Mini and Workstation doesn't have the transfer speed issues, therefore it must be related to the driver.



Thanks!
 
iandk said:
Hey there!

So I just received my new M1 Air in addition to my M1 Mini and I'm supper happy so far.

But since I'm working quite a lot on my local NAS and the Air doesn't have an Ethernet port, I bought an UNI USB C 8 Port hub with an integrated Gigabit NIC.
Since the transfer speed was varying I tried installing the Realtek driver.

However I had to notice, that the driver seems to be outdated (only up to macOS 10.15) and the installer shows that it's specifically for the Intel Macs.

Does anyone know if there's new driver in the works or maybe an other USB NIC that's updated already?

My M1 Mini and Workstation doesn't have the transfer speed issues, therefore it must be related to the driver.



Thanks!
Click to expand...
I would contact the manufacturer about the possibility of a new driver but be aware that driver support in macOS Big Sur is different than in the past. Apple now requires that most drivers run in user-space through their DriverKit framework. This includes ethernet adapters. Most hardware doesn't have DriverKit support yet.
 
Apple_Robert said:
You would be better off getting a single Ethernet to USB-C connection (Belkin), in my opinion. No driver needed.
Click to expand...
I have one of those and it has the same issues.

However it seems to only affect the upload speed, download is fine.
 
You mean upload speeds like to your web space hosted somewhere? Depending on the internet connection you have this might be something which comes with your contract - often (if not usually at least here in Europe) download and upload speeds are asymmetrical resulting in much lower speeds for the latter.
 
The upload speed to my local NAS.
Downloading files is fine and reaches around 940MBit/s so pretty much the limit.

But Uploading files often sits around 600-700Mbit/s.

This issue doesn't occur on my M1 Mini (with the integrated Ethernet port) or Windows Workstation.

So probably some sort of driver issue with the M1 and external USB Ethernet adapters.


However I can't find any official Apple USB Ethernet adapter with USB3 or Thunderbolt 3
 
I use a non rtl USB c hub and this Worms even in recovery mode wirhout driver
 
Yes, but I tested it with iperf3 so there should be no bottleneck.

Btw: My M1 Mini and Workstation do not have this issue.
So it must be related to the UNI USB Ethernet adapter or driver issue.

I'll see tomorrow, if the official NIC has the same issue
 
  • Like
Reactions: pldelisle
iandk said:
Yes, but I tested it with iperf3 so there should be no bottleneck.

Btw: My M1 Mini and Workstation do not have this issue.
So it must be related to the UNI USB Ethernet adapter or driver issue.

I'll see tomorrow, if the official NIC has the same issue
Click to expand...
I just did the same thing. I have a Amazon Basics USB3 ethernet 10/100/1000 & USB3 3 port dongle that is connecting on the 5 Gb/s USB3.0 hub. With iperf3 I'm seeing:

2013 MacPro Server
M1 MacBook Air Client

iperf3 --client macpro.local --time 30 --interval 10
  • [ 7] 0.00-30.00 sec 2.42 GBytes 692 Mbits/sec sender
  • [ 7] 0.00-30.00 sec 2.41 GBytes 691 Mbits/sec receiver
iperf3 --client macpro.local --time 30 --interval 10 --reverse
  • [ 7] 0.00-30.00 sec 2.96 GBytes 847 Mbits/sec sender
  • [ 7] 0.00-30.00 sec 2.96 GBytes 847 Mbits/sec receiver
That is a pretty significant difference. I see the same in reverse if I make the Mac Pro the client and the M1 MBA the server. I'm never using more than about 50% of a single performance core on the M1 MBA and 850 Mbits/sec shouldn't be much bandwidth anyway. I'm not sure why using the M1 as the sender is so much slower than if the M1 is the receiver.
 
Wow, thank you so much for your testing!

Glad you came to the same result, that means there's definitely some sort of driver issue
 
iandk said:
Wow, thank you so much for your testing!

Glad you came to the same result, that means there's definitely some sort of driver issue
Click to expand...
According to System Information, the hub uses a Realtek ethernet chip. There is no way that the M1 should be maxed out at a mere 690 Mbps sending. I might be able to connect a Thunderbolt 3 cable to my Thunderbolt 2 Mac Pro and use the TCP bridge to see if it is a ethernet driver problem or a network problem.
 
I use a TP link USB ethernet (Realtek RTL8193)
iperf3 reports 900+ mbps lan speeds in both directions if connected directly to a USB port (for testing)
It does drop to 800-850 in normal usage which is Mac -> Apple multiport adapter -> TP Link USB hub -> Ethernet Adapter
 
abhi182 said:
I use a TP link USB ethernet (Realtek RTL8193)
iperf3 reports 900+ mbps lan speeds in both directions if connected directly to a USB port (for testing)
It does drop to 800-850 in normal usage which is Mac -> Apple multiport adapter -> TP Link USB hub -> Ethernet Adapter
Click to expand...
My adapter is using a Realtek Chipset as well, it seems to be the RTL8153(B).

Not sure why you don't have this issue. Are you using a M1 Mac as well?
 
jdb8167 said:
According to System Information, the hub uses a Realtek ethernet chip. There is no way that the M1 should be maxed out at a mere 690 Mbps sending. I might be able to connect a Thunderbolt 3 cable to my Thunderbolt 2 Mac Pro and use the TCP bridge to see if it is a ethernet driver problem or a network problem.
Click to expand...
So using a TCP Bridge over Thunderbolt from the M1 MacBook Air with Thunderbolt 3 to the 2013 Mac Pro with Thunderbolt 2 using the Apple Thunderbolt 2<->Thunderbolt 3 adapter I get pretty much the same in both directions. This leads me to agree that it is an 1 Gbps ethernet driver issue.

2013 MacPro Server
M1 MacBook Air Client

iperf3 --client macpro.local --time 30 --interval 10
  • [ 5] 0.00-30.00 sec 36.4 GBytes 10.4 Gbits/sec sender
  • [ 5] 0.00-30.00 sec 36.4 GBytes 10.4 Gbits/sec receiver
iperf3 --client macpro.local --time 30 --interval 10 --reverse
  • [ 5] 0.00-30.00 sec 32.7 GBytes 9.37 Gbits/sec sender
  • [ 5] 0.00-30.00 sec 32.7 GBytes 9.37 Gbits/sec receiver
So in this case, sending from the Mac Pro was just under 10% slower where using the 1 Gbps ethernet adapter sending from the MacPro was about 22% faster. Not sure exactly what this means but you can sure tell that bandwidth isn't the issue with the Thunderbolt connection over 10 times faster.
 
Yea I guess it's a driver issue.
Will test it tomorrow with the "official" Belkin adapter and report back
 
iandk said:
My adapter is using a Realtek Chipset as well, it seems to be the RTL8153(B).

Not sure why you don't have this issue. Are you using a M1 Mac as well?
Click to expand...
Yes, its a M1 MBA
I am getting pretty much the same speed on the M1 as on a windows i5 10th gen system with the same USB setup
(i.e. ethernet adapter via Multiport -> Hub)
A direct connection between the eth adapter and MBA (via a USB C-> USB A cable) yields slightly faster results i.e. around 925-950 mbps . I am guessing the 50-80 mbps loss with the chained connection is on account of overheads at either the Apple adapter (which is also driving a 4k60 monitor) or the TP Link USB hub (which has several other peripherals)

Code: 
Connecting to host 192.168.5.155, port 5201


[  4] local 192.168.5.185 port 53623 connected to 192.168.5.155 port 5201


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-10.00  sec  1015 MBytes   851 Mbits/sec               


[  4]  10.00-20.00  sec  1.01 GBytes   868 Mbits/sec               


[  4]  20.00-30.00  sec  1009 MBytes   846 Mbits/sec               


- - - - - - - - - - - - - - - - - - - - - - - - -


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-30.00  sec  2.99 GBytes   855 Mbits/sec                  sender


[  4]   0.00-30.00  sec  2.99 GBytes   855 Mbits/sec                  receiver

Code: 
abhinav@M1-Air ~ % /Users/abhinav/Downloads/iperf3 -c 192.168.5.155 --time 30 --interval 10 -- reverse


Connecting to host 192.168.5.155, port 5201


[  4] local 192.168.5.185 port 53685 connected to 192.168.5.155 port 5201


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-10.00  sec   997 MBytes   836 Mbits/sec               


[  4]  10.00-20.00  sec  1.02 GBytes   879 Mbits/sec               


[  4]  20.00-30.00  sec  1.02 GBytes   874 Mbits/sec               


- - - - - - - - - - - - - - - - - - - - - - - - -


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-30.00  sec  3.01 GBytes   863 Mbits/sec                  sender


[  4]   0.00-30.00  sec  3.01 GBytes   863 Mbits/sec                  receiver





iperf Done.
 
Last edited:
  • Like
Reactions: jdb8167
abhi182 said:
Yes, its a M1 MBA
I am getting pretty much the same speed on the M1 as on a windows i5 10th gen system with the same USB setup
(i.e. ethernet adapter via Multiport -> Hub)
A direct connection between the eth adapter and MBA (via a USB C-> USB A cable) yields slightly faster results i.e. around 925-950 mbps

Code: 
Connecting to host 192.168.5.155, port 5201


[  4] local 192.168.5.185 port 53623 connected to 192.168.5.155 port 5201


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-10.00  sec  1015 MBytes   851 Mbits/sec              


[  4]  10.00-20.00  sec  1.01 GBytes   868 Mbits/sec              


[  4]  20.00-30.00  sec  1009 MBytes   846 Mbits/sec              


- - - - - - - - - - - - - - - - - - - - - - - - -


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-30.00  sec  2.99 GBytes   855 Mbits/sec                  sender


[  4]   0.00-30.00  sec  2.99 GBytes   855 Mbits/sec                  receiver

Code: 
abhinav@M1-Air ~ % /Users/abhinav/Downloads/iperf3 -c 192.168.5.155 --time 30 --interval 10 -- reverse


Connecting to host 192.168.5.155, port 5201


[  4] local 192.168.5.185 port 53685 connected to 192.168.5.155 port 5201


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-10.00  sec   997 MBytes   836 Mbits/sec              


[  4]  10.00-20.00  sec  1.02 GBytes   879 Mbits/sec              


[  4]  20.00-30.00  sec  1.02 GBytes   874 Mbits/sec              


- - - - - - - - - - - - - - - - - - - - - - - - -


[ ID] Interval           Transfer     Bandwidth


[  4]   0.00-30.00  sec  3.01 GBytes   863 Mbits/sec                  sender


[  4]   0.00-30.00  sec  3.01 GBytes   863 Mbits/sec                  receiver





iperf Done.
Click to expand...
Looks like you have not gotten a reverse connection. There is a space between -- and reverse in your command line.
 
Last edited:
Right - thanks for spotting the typo ..
The adapter seems to be behaving the same regardless of M1 or windows x64..
(It's Connected via the same chain in both scenarios)

See screenshots below

1611733757434.png


1611733867867.png
 
  • Like
Reactions: jdb8167
Just received my Belkin adapter and it basically has the same issue
 
Well, I've just tested both adapters with my M1 mini and it's working fine there!

So it must be somehow related to the Macbook drivers
 
  • Like
Reactions: jdb8167
iandk said:
Well, I've just tested both adapters with my M1 mini and it's working fine there!

So it must be somehow related to the Macbook drivers
Click to expand...
One thing I noticed is that the results vary a lot depending on what else is running. I ran the test while composing a bug report using Apple’s Feedback Assistant for example and got 1/2 the total bandwidth. This reduction in bandwidth was much larger than expected based on Activity Monitor showing plenty of high performance core time availability. I think it might be interesting to compare I/O limitations in the M1 with my Xeon Mac Pro. This might be another limitation of a relatively low-end SoC vs a workstation CPU.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back