10G Ethernet issue on Mac mini 2024 with M4

Question

Level 1

50 points

10G Ethernet issue on Mac mini 2024 with M4

This was posted as an answer to an existing thread, but It is probably better to create a separate thread for it.

I have a problem with the M4 equipped with a 10G interface which can not operate at 10G speed, it works very well at 1G, but can't seem to work more than few seconds when speed is set to 10G (we are network experts, with +20y of expertise in Linux / BSD networking).

Problem can be described as follow:

1. Context:

MacMini 2024 with 10G Ethernet option - 15.2 (24C101)
Connected to a 10G fiber switch - FS switch: S3900-24T4S-R - latest firmware
Proper 10G transceiver (Ethernet >> fiber)
Switch port known to work with 10G fiber

2. Problem:

speed auto-select does not work (fall back to 1G)
forcing speed to 10GbaseT allows interface to switch to 10G for few seconds (interface is up and can ping with no pb).
after a "few seconds" (random) interface disconnects / reconnects randomly and finally stays down
changing flow control or other settings does not solve the issue
only way to "solve" the issue is to lower port speed which is "not ok at all"

Cable is brand new cat. 6A (have bought one specially for this) and of course compatible with 10G speed.

I suspect a firmware problem with the 10G Apple interface -- not sure which manufacturer is behind this, probably some low end manufacturer…

Does anyone have any clue on how to solve this ?

I have seen couple of threads related to this, so I don't think I am the only one with this pb.

[Re-Titled by Moderator]

Mac mini

Posted on Jan 28, 2025 09:33 AM

Reply

Answer 1

gregober Author

Level 1

50 points

Jan 29, 2025 01:23 AM in response to gregober

Upgrading to 15.3 didn't solve my issue.

Interface is now properly recognized as 10G (some data obfuscated):

en0: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
	options=467<RXCSUM,TXCSUM,VLAN_MTU,TSO4,TSO6,CHANNEL_IO>
	ether d0:xx:e5:xx:ab:bb
	inet6 fe80::1895:xxxx:xxxx:28d%en0 prefixlen 64 secured scopeid 0xf
	inet 10.xx.xx.x5 netmask 0xffffff00 broadcast 10.xx.xx.255
	nd6 options=201<PERFORMNUD,DAD>
	media: autoselect (10Gbase-T <full-duplex>)
	status: active

But unfortunately after successful ping for 200s interface stops working (it does not go down, simply stops communicating).

Lowering the speed to 1000baseT allows interface to be more steady again.

So upgrade didn't fix anything.

Reply

Answer 2

hcsitas

Level 6

13,958 points

Jan 29, 2025 03:07 AM in response to gregober

You don’t need a transceiver - just a router with 10G ports and 2 machines with 10G cards one of which you already have. No internet/WAN either - just a local network test with no conversion boxes, not even walls getting in the way.

What works with SFP+ outside of the Apple environment is n/a.

Reply

Answer 3

Jan 29, 2025 03:23 AM in response to gregober

This would appear to be the switch you are talking about:

FS – S3900-24T4S-R, 24-Port Gigabit Ethernet L2+ Switch, 24 x Gigabit RJ45, with 4 x 10Gb SFP+ Uplinks, Stackable Switch, Fanless, Support ETL Certification

All of the 24 RJ-45 ports are Gigabit Ethernet ports that support 10 Mb/s, 100 Mb/s, and 1 Gb/s speeds. The only speed they have in common with the 10 Gb/s Ethernet port on the Mac mini is 1 Gb/s.

I assume that the SFP+ port that you were using was not configured as a stacking port.

Reply

Answer 4

hcsitas

Level 6

13,958 points

Jan 28, 2025 10:50 PM in response to gregober

Make a test as follows:

bypass all wall wiring and connect the Mac directly to a router with 10G ports like this one https://rog.asus.com/ch-en/networking/rog-rapture-gt-be98-model/spec/ with a cable not longer than 1m
connect a Linux machine with 10G ports to the same router, 1m direct
WAN port can be left disconnected i.e. internet not needed
test large file transfers using ftp in both directions from both machines

The goal would be to eliminate your switch as the culprit i.e. any test that does that is good too. Otherwise it’s suspect #1, no blaming the Mac.

Reply

Answer 5

gregober Author

Level 1

50 points

Jan 29, 2025 01:42 AM in response to gregober

Update >> Upgrading to 15.3 didn't [completely] solve my issue.

[see previsou post]

What's even stranger is that It seems to be working using 2.5G and 5G:

➜  ~ ping -c 100  10.100.100.1
PING 10.100.100.1 (10.100.100.1): 56 data bytes
64 bytes from 10.100.100.1: icmp_seq=0 ttl=64 time=0.465 ms
64 bytes from 10.100.100.1: icmp_seq=1 ttl=64 time=0.490 ms
64 bytes from 10.100.100.1: icmp_seq=2 ttl=64 time=0.609 ms
64 bytes from 10.100.100.1: icmp_seq=3 ttl=64 time=0.420 ms
64 bytes from 10.100.100.1: icmp_seq=4 ttl=64 time=0.548 ms
[…]
64 bytes from 10.100.100.1: icmp_seq=97 ttl=64 time=0.634 ms
64 bytes from 10.100.100.1: icmp_seq=98 ttl=64 time=0.636 ms
64 bytes from 10.100.100.1: icmp_seq=99 ttl=64 time=0.651 ms

--- 10.100.100.1 ping statistics ---
100 packets transmitted, 100 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 0.420/0.590/1.011/0.075 ms

and config of interface:

en0: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
	options=467<RXCSUM,TXCSUM,VLAN_MTU,TSO4,TSO6,CHANNEL_IO>
	ether d0:xx:e5:xx:ab:bb
	inet6 fe80::1895:xxxx:xxxx:28d%en0 prefixlen 64 secured scopeid 0xf
	inet 10.xx.xx.x5 netmask 0xffffff00 broadcast 10.xx.xx.255
	nd6 options=201<PERFORMNUD,DAD>
	media: 5000Base-T <full-duplex>
	status: active

Lowering the speed to 5000baseT allows interface to be steady again.

BUT STILL NO 10G upgrade to this speed target does not work.

As a synthesis upgrade to 15.3 solved:

detection of speed in "autoselect" >> now properly identified as 10Gbase-T interface
working at speed of up to 5G
but still no luck with full 10G though - stops working after couple of seconds

Reply

Answer 6

gregober Author

Level 1

50 points

Jan 28, 2025 10:41 PM in response to hcsitas

5m

far far away from the limit of 100m and it is not going through any patch panel - direct connection.

Reply

Answer 7

gregober Author

Level 1

50 points

Jan 29, 2025 02:00 AM in response to hcsitas

Testing with 1m cable is "not compliant" with our target.

My desk is 4m from switching cabinet, therefor I am using 5m cable length. Max length for 10G ethernet is 100m.

Cable is brand new and has been specifically bought for this, it is compliant with the target standard (6A) and is shielded cable, no problem with cables.

Reply

Answer 8

hcsitas

Level 6

13,958 points

Jan 29, 2025 02:14 AM in response to gregober

Sure, test with 5m.

As a reminder, target here is to confirm your Apple theory, otherwise it's just a theory. Additionally, Sequoia 15.3 makes no mention of 10G fixes - so don't get your hopes up. Relaying telephone conversations by anonymous users to these forums is unreliable as is what Apple Support may or not speculate during, or mentions of those posts by other users no matter what their level. As of now the ball is firmly in your court, you know what to do next :)...good luck!

Reply

Answer 9

gregober Author

Level 1

50 points

Jan 29, 2025 02:52 AM in response to hcsitas

Yes, we must confirm the "Apple theory", unfortunately I only have one 10G transceiver for 10G Ethernet, so I can't test that on other HW for the time being.

I can confirm that 10G with SFP+ is working on the same interface.

Reply