10gbe Thunderbolt 3 Slows Down when other devices are connected

Hi! Thanks for the reply!

I'm not sure it's the hub because when I connect a 10GB USB HDD to the thunderbolt 3 port it is also slow. These are the different configurations that I have tried:

10gbe connected to Thunderbolt 3

Satechi Hub connected to Thunderbolt 3

10gbe connected to Thunderbolt 3

Satechi Hub connected to USB 3 port

1 10TB External HDD connected to USB 3

1 2TB External HDD connected to USB 3

10gbe connected to Thunderbolt 3

2x 6TB HDD in Raid 0 connected to Thunderbolt 3

All of those configuration resulted in very slow transfer on my network using rsync (max 15 MB/s). If I remove the devices, then I get up to 220 MB/s.

So now, I only have the Satechi Hub connected to the USB 3 port.

I also tried using a 10gbe Thunderbolt 3 from QNAP instead of the one from Sonnet and it results in the same problem.

Hello,

Thanks again for taking the time to reply.

So basically I was trying to copy some files from my imac to my NAS (it's a Lenovo D30 running Ubuntu Server 20.04) using rsync. I wasn't really testing the speed but was just very surprised why the transfer rate was only 15 MB/s and this is when I sort of started to make a speed test.

My original configuration before attaching the 10gbe TB3 was:

Satechi Type-C Aluminium iMac Stand connected to the TB3 port

Mediasonic Duo 10G enclosure with 2 6TB Ironwolf HDD in Raid 0 connected to the TB3 port

WD My Book with 2TB HDD connected to the USB 3 port.

I was using the built-in Gigabit ethernet and rsync transfer from files in the Mediasonic to my server using rsync is around 60MB/s.

Then recently I added the 10gbe TB3 and connected it to one of the TB3 ports. I removed the Mediasonic enclosure from the imac's TB3 port and connected it to the USB-C port on the Satechi imac stand. The speed as I said was very slow - 15MB/s.

So, I transferred the Mediasonic Enclosure from the Satechi Hub and moved it to one of the USB 3 ports. Speed increased a bit, like around 28MB/s but that is still very slow.

During these tests, nothing else was connected to the Satechi Hub except for that time when I connected the Mediasonic enclosure to the hub's USB-C port.

So next I disconnected the Satechi hub from the TB3 port and connected it to one of the USB 3 ports. This was when I saw a big speed improvement up to 150MB/s.

I noticed also that if I disconnect the WD My Book 2TB from the USB 3, there's a further slight increase in the transfer speed.

So, to get the fastest possible speed, I cannot connect anything to the 2nd TB3 port while the 10gbe TB3 is connected. The Satechi imac stand is now connected to one of the USB 3, the Mediasonic is also connected to the USB 3 port and I removed the WD My Book from the USB 3 port. If I put it back, there's a slight speed decrease.

I only use the Satechi Hub for its convenience of having ports facing me. I don't connect any HDDs there and the only time I use it is when I have to charge my keyboard or trackpad or when I transfer files from SD cards so I'm fine with connecting the hub to one of the USB 3 ports instead of on the TB3 port. However, it bothers me that I cannot do so.

For the Satechi imac stand, I am using the cable that came with it. It'S attached to the hub and I cannot replace that. I can either connect it to the TB3 port or use an attachment so that I can connect it to one of the USB 3 ports. I don't think the hub is a TB3 hub but rather that its cable is USB-C so I can connect it to the TB3 port. Like I said, I don't need TB3 speeds for the hub.

For the Mediasonic enclosure, I am using a USB-C 3.1 Gen 2 cable and the WD My Book uses a USB 3 cable.

Thanks for making the analogy "TB is no different from Wi-Fi". I actually didn't think of it that way until you mentioned it. Maybe because I was and still am under the impression that TB3 is rated at 40Gb/s and adding a 10Gbe will not have that much impact since that is the only real TB3 device that I am using. What is further surprising is the fact that having the HDDs connected to the USB 3 ports are still having an impact on the speed.

I also have a mac mini 2018 and noticed this problem as well. I have an eGPU connected to TB3 and a 10Gbe to another TB3. But there I have more room to play around because the mac mini has 2 TB3 bus with a total of 4 TB3 ports so I just connect my TB3 devices to different bus, i.e., eGPU to bus 1 and 10Gbe to bus 0. If I connect a HDD to one of the TB3 ports on either bus, the speed suffers. I have to remove either the eGPU or the 10Gbe.

It seems like its the 10Gbe that is causing all the issues but it's not the unit itself because I have tried several already. So, is it that with using a 10Gbe I cannot use any other TB3 device?

Thanks so much!

Hi! Thanks again for your reply.

I am not really so concerned about the port's speed but more concerned about the difference in throughput when another device is connected to the TB3 port. I didn't expect such a big difference. I used iperf3 to check the bandwidth and it's reporting the full 10Gb, both from the server and from the imac even when there are devices connected and all. It's the actual throughput that is impacted.

I always tested transferring the same file (24 GB), to the same drive on the same server. I also used two different 10Gbe TB3, one from Sonnet and the other from QNAP and it was almost always the same results, slowest when there is another device connected to the other TB3 port, a bit faster when the devices (Satechi hub and 2x HDD enclosures) are connected to the USB 3 ports, and even faster when I remove one of the devices from the USB 3 ports, i.e., only two devices are connected instead of 3. Fastest is when there are no devices connected except the 10Gbe TB3.

I should mention that I copy the test file from one of the HDD enclosures.

But I get the same results when I use a file from the internal SDD, from 24 MB/s with all the devices I mentioned connected to 250 MB/s if there are no other devices connected except the 10Gbe TB3.

Everything in my network is 10Gb, the switch and the server. It's in the same subnet.

Like I mentioned before, I also tested the same from my macmini and also noticed a decrease in speed if I connect the 10Gbe TB3 to a port where another TB3 device is already connected. For instance, in Bus 0 of my macmini I have an eGPU connected. If I use the same TB3 port in Bus 0 the speed decreases but if I transfer it to a TB3 port in Bus 1, then it's okay. In the macmini it's not very problematic because it has 2 TB3 buses with 4 TB3 ports but sadly in the imac, you only have two TB3 ports which I think are just in one bus.

I suspect that the 10Gbe TB3 requires so much bandwidth that you really need to isolate it if you use one. I somehow regret not getting a macmini fitted with 10GBe when I bought it but then you don't have that choice with the imac, only the pro model I think can be fitted with an internal 10Gbe.

So my solution now is to just remove all the HDD enclosures and just have the Satechi hub connected to the USB 3 port. I get up to 170 MB/s write speed on the server which is faster than if I had the HDD directly connected to USB 3 with an enclosure.

I have the network share automatically mounted using auto_master and auto_smb so the shares appear as if they are part of the file system. I wish I can use NFS but that's an entirely different problem. :)

Thanks a lot for your time!