High Sierra native support for 10Gb ethernet

[PTuser]

Hm, both error counters are 0, which means that the connection is good (so in my opinion the cable and shielding are fine).
Since testing the localhost with iperf yielded troughput far above 10gbit/s, the CPU shouldn't be the problem either.

Is it possible that the cards run too hot? The AQC107 chipset tends to throttle at high temperatures.
Or do you happen to have some firewall software running that might be slowing the connection down?

I dont think it is running hot as temperatures are monitored in my Master Mac 1 (the Hackintosh) and I cant see any difference. Did you see the article that Syba sent me? no firewalls are active....
It feels like I should persevere as they are actually being seen by the OS.....I can return them as a last option.

 

[Kris Kelvin]

I dont think it is running hot as temperatures are monitored in my Master Mac 1 (the Hackintosh) and I cant see any difference.

Unfortunately, that doesn't have to mean anything. Hot spots can be very localized. It would be worth a try shutting them down, letting them cool for a few hours and then run iperf first thing after starting up again – just to rule that out.

Did you see the article that Syba sent me?

Yes; it's not really helpful in this case.

• A higher MTU helps when CPU-bound, but that doesn't seem to be the case here
• SMB settings are irrelevant when using iperf
• You're not sharing a Thunderbolt bus, so that doesn't apply
• Kernel tuning isn't necessary for 10G in recent versions of macOS
• A cabling problem should have resulted in errors in Network Utility.app
• No transceivers or switches are used

Do you happen to have a third machine to test with?
Or you could put both cards in the same computer to rule out some weird PCI problems.

 

[PTuser]

Thank you for explaining this....I dont have a third machine with PCIe slots - Ill switch off - Im in the UK and its 11pm here anyway - Will try it tomorrow and report.

Would you try different cards? The chances are they will still be the ACQ107 chipset, so Im not sure if it would help, although theres a card with Broadcom 57810S . Do you happen to know if that chipset is supported natively?

 

[Squuiid]

Thank you for explaining this....I dont have a third machine with PCIe slots - Ill switch off - Im in the UK and its 11pm here anyway - Will try it tomorrow and report.

Would you try different cards? The chances are they will still be the ACQ107 chipset, so Im not sure if it would help, although theres a card with Broadcom 57810S . Do you happen to know if that chipset is supported natively?

What is the share? AFP, SMB or NFS?

 

[PTuser]

What is the share? AFP, SMB or NFS?

Hi - not sure what you mean?

 

[Kris Kelvin]

Would you try different cards?

The more things you can test, the higher the chance you'll find the problem.
Unfortunately, 10GbE devices are few and far between in non-enterprise environments.

theres a card with Broadcom 57810S . Do you happen to know if that chipset is supported natively?

It isn't. Only AQC107 works out of the box.
While there are some other cheap options (e.g. Tehuti-based cards), AQC107 is currently the best option for your use-case, I think.

 

[Squuiid]

Hi - not sure what you mean?

Where are you copying files from and to? Using which protocol?
If SMB, and not running Mojave make sure you do this:
https://forums.macrumors.com/thread...or-10gb-ethernet.2092875/page-4#post-26276246

 

[PTuser]

Where are you copying files from and to? Using which protocol?
If SMB, and not running Mojave make sure you do this:
https://forums.macrumors.com/thread...or-10gb-ethernet.2092875/page-4#post-26276246

The tests are between the system drives SSDs of each machine that are attached by a single cable between each card. No NAS... How do I check if im running SMB?

 

[PTuser]

Unfortunately, that doesn't have to mean anything. Hot spots can be very localized. It would be worth a try shutting them down, letting them cool for a few hours and then run iperf first thing after starting up again – just to rule that out.



Yes; it's not really helpful in this case.

• A higher MTU helps when CPU-bound, but that doesn't seem to be the case here
• SMB settings are irrelevant when using iperf
• You're not sharing a Thunderbolt bus, so that doesn't apply
• Kernel tuning isn't necessary for 10G in recent versions of macOS
• A cabling problem should have resulted in errors in Network Utility.app
• No transceivers or switches are used

Do you happen to have a third machine to test with?
Or you could put both cards in the same computer to rule out some weird PCI problems.

Hi Kris - Running from cold - Mac 2 remote to local is 3.14gb/s Local to remote 6.33 gb/s so its no different really.

I have just remembered that someone was helping me sort out some issues on my Hackintosh (Mac1). The previous card was an intel 1gb in the same slot. I remember he messed with the DSDT because the card wasn't native to the OS. Would IO reg show up anything? Im not versed in this kind of editing at all. My thoughts are as the Chipset on the 10gb card is native it shouldn't matter. Is it worth uploading my ioreg?

 

[PTuser]

Ok - Moved the Pcie Card for the third time this afternoon Mac 2 - The Test from iperfutil on Mac 2 Local to remote starts at 9.8gb/s and then reduces pretty quickly to 7.15gb/s - Remote to Local 6.84 gb/s - odd how each slot is slightly different

BY further messing with the network hardware settings on each comp Local to remote from Mac 2 is now stabilised at 9.80gb/s for about 5 mins and is down to 8.5 gb/s now. Longer I test for the slower it gets - is that normal? (just gone under 8 as I type. )

 

[Kris Kelvin]

I remember he messed with the DSDT because the card wasn't native to the OS.

I'm afraid that's out of my league.

Local to remote from Mac 2 is now stabilised at 9.80gb/s for about 5 mins and is down to 8.5 gb/s now

9.80Gbit/s sounds about right, so the PCI slot or its configuration seems to have been the problem.
The throtting really could be because of thermal constraints.

I think it's time to do some file transfer tests again.

 

[PTuser]

I'm afraid that's out of my league.



9.80Gbit/s sounds about right, so the PCI slot or its configuration seems to have been the problem.
The throtting really could be because of thermal constraints.

I think it's time to do some file transfer tests again.

Hi - you gave me the definitive way to use the iprefutil which was really helpful so thank you. I have a 14 gig ziped file - will that do? which method do you suggest?

 

[Kris Kelvin]

I have a 14 gig ziped file - will that do? which method do you suggest?

I'd suggest just mounting a file share over AFP or SMB and copy the file like you usually would. You can then use Activity Monitor.app to see the exact transfer speed. If your drives are quick enough, the file should be copied in less than 20 seconds.

 

[PTuser]

I'd suggest just mounting a file share over AFP or SMB and copy the file like you usually would. You can then use Activity Monitor.app to see the exact transfer speed. If your drives are quick enough, the file should be copied in less than 20 seconds.

Host to slave 14.18 Gb in 1 file took 2.5 mins
Slave to Host 14.18 Gb in 1 file took 3 mins
System drive to system drive both ssds

 

[Kris Kelvin]

Host to slave 14.18 Gb in 1 file took 2.5 mins
Slave to Host 14.18 Gb in 1 file took 3 mins
System drive to system drive both ssds

That's almost exactly 1Gbit/s. Are you sure it's using the 10G connection?
Also, what kind of SSD? A 2.5" SSD connected directly to a Mac Pro's SATA port won't exceed 250MB/s and might be well below that.
Did you connect via the AFP or SMB protocol?

 

[PTuser]

Hi Both SATA SSDs - The 1g connection was turned off before doing the test. Im not sure about AFP or SMB - how would I check that? apologies for my limited knowledge

 

[Kris Kelvin]

Hi Both SATA SSDs

So the 10G connection is 5x faster than the SATA drive in your Mac Pro. The transfer will be limited by that and will take at least a minute for the 14GB file.

Im not sure about AFP or SMB - how would I check that?

When you select the shared volume in Finder and choose "Get Info" from the Contextual Menu, the info window will tell you this.
If you connect manually, you can specify the protocol to use (see https://support.apple.com/en-us/HT204445).

 

[Skyfokker]

With that, you're testing a long chain of things at the same time (drives, RAID software, filesystem, SMB client, SMB server, Blackmagic Disk Speed Test App).

To find out more about just your networking situation: What are your iperf results?

So I have some interesting updates: the cards were in pcie slot3 and Iperf showed 5.8gb/s. Slot 2 showed worse: about 4 gb/s. In slot 4 it went up to 7.8 gb/s!

Then I switched the Ethernet card in the server to a sonnet presto in slot 4 and it went up to 9gb/s! So slot 4 worked! In both client and server Mac pros. Server sonnet presto client Aquentia 107

Strange thing is black magic disk test showed max of around 500-600 mb/s in AFP which is higher than SMB. They are running on 10.13.6 so SMB writing is not an issue. What gives?

 

[PTuser]

So I have some interesting updates: the cards were in pcie slot3 and Iperf showed 5.8gb/s. Slot 2 showed worse: about 4 gb/s. In slot 4 it went up to 7.8 gb/s!

Then I switched the Ethernet card in the server to a sonnet presto in slot 4 and it went up to 9gb/s! So slot 4 worked! In both client and server Mac pros. Server sonnet presto client Aquentia 107

Strange thing is black magic disk test showed max of around 500-600 mb/s in AFP which is higher than SMB. They are running on 10.13.6 so SMB writing is not an issue. What gives?

Hi what was your original card?

 

[Skyfokker]

Hi what was your original card?

I’ve been testing with the aquantia 107 for clients and sonnet presto for server. Decided to switch the server to 107 for testing

 

[PTuser]

Ive tested my Mac 2 with the Black Magic Speed test - Got read 720mb/s write 198mb/s . Tried a 4th cable that arrived today - pref util with that cable is stable at 6.78mb/s both ways. AFP and SMB make little difference

 

[Kris Kelvin]

Ive tested my Mac 2 with the Black Magic Speed test - Got read 720mb/s write 198mb/s.

That's probably as good as it gets, considering the speed limit of the Mac Pro's SATA port. The much higher read rate is most likely because the data has been cached by the operating system.

 

[PTuser]

Hi Kris
I purchased a couple of Sonnet 10G cards to see if there was any improvement. Ill return the worse performing pair . Im getting test error on Mac 2 with the iperf ulil app on the slave using the same settings you suggested on each. Quit a few times but can't get it going - the connections active - Ive transferred files - so far its proving troublesome

Ive sorted this - it was a stupid mistake on my part. results are so far a tiny bit slower on the Sonnets. Tried SMB and AFP - not really any noticeable difference between the protocols. Its much more finicky regarding cables. One cable produced a read write speed of under 100 mb/s on Speed disk. Change cable - back to 700 mb/s read.

 

[Dandu]

Hi,

I have a Sonnet Card, and a big problem : i can't use AirPlay 2 with my HomePod. When i try to stream from iTunes, there is no sound on the HomePod. It works with the Ethernet from my display (Thunderbolt Display) and from a Tehuti card (but the driver are unstable)

 

[overshoot]

Hi,

I have purchased that card off eBay https://www.ebay.com/itm/Carte-Rése...e=STRK:MEBIDX:IT&_trksid=p2060353.m2749.l2649

It's an AQC107 chipset.
Card is detected, drivers are loaded but the cable will never be detected as plugged.
I've read other post saying there is either a firmware modification to be done or a kext patch using Clover (on Hackintosh).

Is anyone successfully using this card in a Mac Pro 5.1 or 4.1?