Sanity check before I return another Realtek based 2.5G USB adapter

[DHSteeze]

TLDR version - I know the Realtek 8156/8156B adapters can kinda suck, but due to 2 of my systems in question being ITX-based and only having a single x16 PCIe slot (which is currently filled with a GPU in both), USB adapters are more or less my only choice. How likely is it that an adapter can receive at 2.3-2.4G, but can only send at far less?

 

Non-TLDR version - I thought I'd get some input before I lose my mind or get stuck in a buy/return pattern of hardware here. I recently decided it would be a "great" idea to upgrade my small 1G ethernet network in my house to 2.5G, and it's been somewhat frustrating to say the least. I've now purchased 4 Realtek 8156/8156B USB-C based 2.5G adapters and I've already returned 2, with a third about to be returned as well unless I can figure out what's going on.

 

To give a quick overview of my setup, most of the devices on my network are wireless and handled by a separate TP Link WiFi6 router. The 1G LAN I'm upgrading is only 3 devices, and all are in the same room, so CAT5e runs to my 2.5G 5 port switch are very short (sub-5 ft). I did all testing through the 2.5G switch, as well as direct connections from PC to PC to take the switch out of the equation. It was all done bi-directionally between all machines with all power saving options disabled, with/without jumbo frames enabled at 9k on all 3, and all set to 2.5G Full Duplex instead of Auto negotiation. All built-in ethernet ports were disabled in Windows when using the USB 2.5G adapters to ensure the correct port was used during testing. I used iPerf and then confirmed those results with large file transfers between machines. I had also read a fair amount about the Realtek 8156/8156B adapters beforehand to know the included drivers were probably outdated, so I also updated all machines with the latest drivers from the Realtek website.

 

My devices are:

1. Alienware M15 R7 laptop (i7-12700H, 16GB RAM, 3x USB 3.2 gen1 type-A ports, 1x Thunderbolt4/USB4 type-C port, built-in Intel Killer E3100 2.5G ethernet)

2. ITX NAS (Gigabyte A520, AMD R5 5500, 16GB RAM, 6x USB 3.2 gen1 type-A ports)

3. ITX VR Gaming PC (ASRock A520, AMD R5 5500, 32GB RAM, 5x USB 3.2 gen1 type-A, 1x USB 3.2 gen1 type-C ports)

I initially bought 2 very cheap adapters ($15 each) to try the switchover. Once installed and configured, 1 of the adapters would send at 2.3-2.4G, but would only receive at 1-1.2G. The second adapter would neither send nor receive at anything over 800ish Mbps. I swapped both adapters back and forth between all 3 machines, and I tried all USB ports on all machines, as well as the Thunderbolt port on the Alienware. Both USB adapters yielded roughly the same speed results regardless of which port or machine they were plugged in to. I even tried an older Macbook Air I had laying around with similar results.

 

So for those 2 cheapest adapters, I assumed I could rule out driver issues, since now 4 machines exhibited similar results. I returned both of those adapters (Edimax, the cheapest on Amazon) and ordered another adapter, but from another brand (Wavlink). The Wavlink functioned perfectly at 2.3-2.4G in both directions with a direct connection to my Alienware and also through the switch. Thanks to that, I ordered another Wavlink only to have to have it function similar to one of the Edimax adapters, it would send at 2.3-2.4G but receive at around 1G. Again, I confirmed the initial results by moving the good and bad adapter around to the same 4 machines, and got roughly the same results.

 

So before I return the second Wavlink adapter in hopes I get a better one or try yet another brand, is there any logical reason the adapter would function unidirectionally like that? Other than the Russian roulette option of buying and returning Realtek USB adapters, I suppose I do have 2 other options I could try if I get desperate: I could try a 5G USB adapter since they aren't Realtek 8156/8156B based; or the Gigabyte board in the NAS supports PCIe bifurcation so I could split the PCIe slot with an adapter to allow the use of the GPU and a PCIe based 2.5G card.

[mariushm]

If the gigabyte a520 board has a m.2 slot with pci-e lanes, you could get a m.2 to pci-e x1/x2/x4 adapter board and then use a riser cable to add a pci-e x1 network card to the system.

 

Example : https://www.ebay.com/itm/284048215450 or https://www.ebay.com/itm/254730098594

 

or version with molex power connector : https://www.ebay.com/itm/266040997576

 

The sata (or molex) power cable is only needed if you plan to use a riser cable directly to the ethernet card.

PCI-e slots have 3.3v and 12v, but the m.2 connector has only 3.3v, so the adapter board gets 12v from the sata power cable and puts it into the slot

 

For example if you use a cable like this https://www.ebay.com/itm/265508233633 or a cable like this https://www.ebay.com/itm/284915179693 then you need to power the m.2 adapter with that power cable, to inject 12v into the pci-e slot on the m.2 card.

 

There are cables that have separate 12v input, in which case you probably don't need to power the m.2 slot, but you must power the cable. here's an example of such cable : https://www.ebay.com/itm/266071935002

 

And there's riser cables that use USB cables to connect just the data wires from a pci-e x1 slot (or your pci-e x4 slot on the m.2 card) to a small adapter board and that adapter board needs some input power to produce the 12v and 3.3v that go into the slot.

For example see this : https://www.ebay.com/itm/124508160736

The tiny bit with the pci-e x1 slot you could plug into the m.2 to pci-e adapter (and you don't need to give extra power to the m,2 adapter) and the usb cable is used to connect the data wires of the pci-e x1 connection to the adapter board, which has a voltage regulator on it that produces 3.3v from the 5v coming in through that SATA power connector. SATA also has 12v so that goes into the pci-e x4 slot as well.

 

[DHSteeze]

Thanks for the reply and the idea! Both the A520 boards have an m.2, but each are occupied with their respective systems boot drive.

 

I've setup a return and replacement for the third faulty USB adapter, so I'll see how that goes once I get it. If it's still a no-go when I get the replacement though, I'll have to start looking at other options like splitting my PCIe slot or using the m.2.

[Alex Atkin UK]

2 hours ago, mariushm said:

If the gigabyte a520 board has a m.2 slot with pci-e lanes, you could get a m.2 to pci-e x1/x2/x4 adapter board and then use a riser cable to add a pci-e x1 network card to the system.

You don't even need to do that. https://www.aliexpress.com/item/1005003533018920.html

 

I'm using one of these right now in my gaming PC.  It might be ugly, but as a last resort you could find some way to route the cables out of the case and it loose outside, if there is no spare PCIe slot to put the shield in.  I find there is usually space on most cases above the PCIe slots.

[Alex Atkin UK]

1 hour ago, DHSteeze said:

Thanks for the reply and the idea! Both the A520 boards have an m.2, but each are occupied with their respective systems boot drive.

 

I've setup a return and replacement for the third faulty USB adapter, so I'll see how that goes once I get it. If it's still a no-go when I get the replacement though, I'll have to start looking at other options like splitting my PCIe slot or using the m.2.

Are you sure its not a cable issue?  I have a cheap Plugable brand Realtek adapter in my Mac Mini although running Asahi Linux:

Connecting to host server.lan, port 5201
[  5] local 192.168.10.45 port 34940 connected to 192.168.10.253 port 5201
[ ID] Interval           Transfer     Bitrate         Retr  Cwnd
[  5]   0.00-1.00   sec   275 MBytes  2.31 Gbits/sec    0    404 KBytes       
[  5]   1.00-2.00   sec   275 MBytes  2.31 Gbits/sec    0    399 KBytes       
[  5]   2.00-3.00   sec   276 MBytes  2.32 Gbits/sec    0    385 KBytes       
[  5]   3.00-4.00   sec   277 MBytes  2.32 Gbits/sec    0    393 KBytes       
[  5]   4.00-5.00   sec   275 MBytes  2.31 Gbits/sec    0    385 KBytes       
[  5]   5.00-6.00   sec   276 MBytes  2.32 Gbits/sec    0    379 KBytes       
[  5]   6.00-7.00   sec   278 MBytes  2.33 Gbits/sec    0    427 KBytes       
[  5]   7.00-8.00   sec   278 MBytes  2.33 Gbits/sec    0    430 KBytes       
[  5]   8.00-9.00   sec   276 MBytes  2.32 Gbits/sec    0    407 KBytes       
[  5]   9.00-10.00  sec   276 MBytes  2.32 Gbits/sec    0    396 KBytes       
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-10.00  sec  2.70 GBytes  2.32 Gbits/sec    0             sender
[  5]   0.00-10.00  sec  2.70 GBytes  2.32 Gbits/sec                  receiver

 

Connecting to host server.lan, port 5201
Reverse mode, remote host server.lan is sending
[  5] local 192.168.10.45 port 46232 connected to 192.168.10.253 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-1.00   sec   278 MBytes  2.33 Gbits/sec                  
[  5]   1.00-2.00   sec   279 MBytes  2.34 Gbits/sec                  
[  5]   2.00-3.00   sec   279 MBytes  2.34 Gbits/sec                  
[  5]   3.00-4.00   sec   280 MBytes  2.35 Gbits/sec                  
[  5]   4.00-5.00   sec   280 MBytes  2.35 Gbits/sec                  
[  5]   5.00-6.00   sec   279 MBytes  2.34 Gbits/sec                  
[  5]   6.00-7.00   sec   279 MBytes  2.34 Gbits/sec                  
[  5]   7.00-8.00   sec   280 MBytes  2.35 Gbits/sec                  
[  5]   8.00-9.00   sec   279 MBytes  2.34 Gbits/sec                  
[  5]   9.00-10.00  sec   277 MBytes  2.32 Gbits/sec                  
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-10.00  sec  2.73 GBytes  2.34 Gbits/sec  4685             sender
[  5]   0.00-10.00  sec  2.72 GBytes  2.34 Gbits/sec                  receiver

The server there is on 10Gbit which is probably why it got retries when sending to the 2.5Gbit client.

 

My Club3D one seems to work fine on Windows 11:

[mariushm]

2 hours ago, Alex Atkin UK said:

You don't even need to do that. https://www.aliexpress.com/item/1005003533018920.html

 

I'm using one of these right now in my gaming PC.  It might be ugly, but as a last resort you could find some way to route the cables out of the case and it loose outside, if there is no spare PCIe slot to put the shield in.  I find there is usually space on most cases above the PCIe slots.

Jesus, they make Cat6a cables where each pair of wires has a specific number of twists per inch for signal integrity, and they strongly push for less than 1cm of untwisted wires when you crimp an ethernet jack, and then these guys fuck it up with those unshielded untwisted wires between the board headers and the port on the bracket.

 

But yeah, it would be a good solution...

 

2 hours ago, DHSteeze said:

Thanks for the reply and the idea! Both the A520 boards have an m.2, but each are occupied with their respective systems boot drive.

A 64 GB sata drive is like 10-15$

 

[DHSteeze]

I would like to say definitely not a cable issue, but the whole situation just seems so bizarre to me that I don't feel confident ruling out anything 100%. I have several cat5e cables on hand and have tried them all, some are off the shelf and some are ones I've made, tested and use regularly. I've also been using a cable I know works at the 2.5G speeds between the Alienware and the original Wavlink adapter I bought, which works perfectly in both directions on all the machines I tried it in.

 

The fact I just got my replacement for the "bad" Wavlink and it functions the same as the one it's meant to replace, means I think I'm done wasting my time trying these 2.5G adapters. I've read a fair amount about them being a little spotty and being finicky about drivers, but I've now tried 2 different brands and 5 total adapters to only get 1 that works 100% as expected. I'll try to first source out a 5G USB version since they don't use the Realtek chipset, and then go from there.

[Alex Atkin UK]

On 3/31/2023 at 11:55 PM, mariushm said:

Jesus, they make Cat6a cables where each pair of wires has a specific number of twists per inch for signal integrity, and they strongly push for less than 1cm of untwisted wires when you crimp an ethernet jack, and then these guys fuck it up with those unshielded untwisted wires between the board headers and the port on the bracket.

I know what you mean, my first thought was "wtf, is that even going to work?" but it certainly seems to, at least on my switch.  I probably wouldn't expect it to work well at 100m though.