F5F Stay Refreshed Power Users Networks Adapter 3D Club CAC-1520, USB 3.2 Gen1 Type-C to 2.5G NBASE-T

Adapter 3D Club CAC-1520, USB 3.2 Gen1 Type-C to 2.5G NBASE-T

Adapter 3D Club CAC-1520, USB 3.2 Gen1 Type-C to 2.5G NBASE-T

A
AlongCameBen
Member
68
05-18-2024, 09:08 PM
#1
I just acquired one of these adapters based on the Realtek r8152 for my ITX setup since the GPU is occupying the sole PCIe lane. I ran iperf3 to a server connected via 10G Ethernet with standard 1500 frames. Here’s what the results showed:

- Connecting to lcars, port 5201 in reverse mode, remote host lcars sent [5] local 192.168.1.94 port 35086 linked to 192.168.1.253 port 5201.
- Interval Transfer Bitrate ranged from 0.00–1.00 sec: 263 MBytes at 2.21 Gbits/sec, 264 MBytes at 2.22 Gbits/sec, and so on up to 265 MBytes at 2.22 Gbits/sec.
- Receiver speeds varied between 2.21 and 2.44 Gbits/sec depending on the interval.
- Overall, the transfer consistently hit around 264 MBytes per second with stable throughput.

I also noticed that changing the MTU to 9000 didn’t significantly improve performance; it slightly increased to 2.23Gbit but caused more retries and a brief spike to 2.44Gbit before settling. Real-world NFS transfers from the server seemed slower, peaking near 250MB/s with noticeable fluctuations, likely due to other clients using the same 1500 MTU.

I’m relieved the adapter worked well and am thinking about connecting the onboard Intel Gigabit directly to a spare NIC on the server for NFS backups. This could help balance speeds and improve response times.
A
AlongCameBen
05-18-2024, 09:08 PM #1

I just acquired one of these adapters based on the Realtek r8152 for my ITX setup since the GPU is occupying the sole PCIe lane. I ran iperf3 to a server connected via 10G Ethernet with standard 1500 frames. Here’s what the results showed:

- Connecting to lcars, port 5201 in reverse mode, remote host lcars sent [5] local 192.168.1.94 port 35086 linked to 192.168.1.253 port 5201.
- Interval Transfer Bitrate ranged from 0.00–1.00 sec: 263 MBytes at 2.21 Gbits/sec, 264 MBytes at 2.22 Gbits/sec, and so on up to 265 MBytes at 2.22 Gbits/sec.
- Receiver speeds varied between 2.21 and 2.44 Gbits/sec depending on the interval.
- Overall, the transfer consistently hit around 264 MBytes per second with stable throughput.

I also noticed that changing the MTU to 9000 didn’t significantly improve performance; it slightly increased to 2.23Gbit but caused more retries and a brief spike to 2.44Gbit before settling. Real-world NFS transfers from the server seemed slower, peaking near 250MB/s with noticeable fluctuations, likely due to other clients using the same 1500 MTU.

I’m relieved the adapter worked well and am thinking about connecting the onboard Intel Gigabit directly to a spare NIC on the server for NFS backups. This could help balance speeds and improve response times.

B
BooyaLuver19
Member
77
05-18-2024, 09:08 PM
#2
You can connect your GPU via a virtual link to an actual USB C port.
B
BooyaLuver19
05-18-2024, 09:08 PM #2

You can connect your GPU via a virtual link to an actual USB C port.

L
Lalesita
Member
50
05-18-2024, 09:08 PM
#3
It offers a full USB-C connection plus DisplayPort. The main concern is how the GPU would handle the PCIe interface under heavy load, though it seems the design accounts for high bandwidth needs typical of VR headsets.
L
Lalesita
05-18-2024, 09:08 PM #3

It offers a full USB-C connection plus DisplayPort. The main concern is how the GPU would handle the PCIe interface under heavy load, though it seems the design accounts for high bandwidth needs typical of VR headsets.

C
CaptanJim
Member
160
05-18-2024, 09:08 PM
#4
The 2080 Ti works well with just a little saturation in an 8x port.
C
CaptanJim
05-18-2024, 09:08 PM #4

The 2080 Ti works well with just a little saturation in an 8x port.