Realtek 10 GbE USB Adapters might be on the way?

[obovata]

I was hoping to find a RTL8261C SFP+ RJ45 adapter but they are nowhere to be found... Maybe it needed a hw revision to ship in volume?

I was browsing Amazon today and I saw a QSFPTEK brand 10Gbase-T SFP+ that claims "Upgrade Realtek Chip" in the title (ASIN: B0G5PL53V6). Interest piqued, I checked the product description but unfortunately it did not list the specific Realtek chip used. However going through the pictures in the reviews the label on the transceiver does clearly indicate RTL8261C:



Pricing at this moment is pretty reasonable at $22.99. Interestingly QSFPTEK's implementation dropped 1000base-t support. The transceiver is 10Gb/5Gb/2.5Gb only. This isn't the one universal 10Gbase-T SFP+ to rule them all.

It would be nice to see how the RTL8261C equipped QSFPTEK measures up to other 10Gb SFP+ offerings with some data. The reviews so far just give subjective impressions that aren't very useful e.g. "One thing to note (which is completely normal for this type of module): it runs very hot to the touch." and "The upgraded Realtek chipset seems to handle heat and power efficiently, as the module ran cool even during sustained high-bandwidth transfers."

 

[neophoenix]

I was browsing Amazon today and I saw a QSFPTEK brand 10Gbase-T SFP+ that claims "Upgrade Realtek Chip" in the title (ASIN: B0G5PL53V6). Interest piqued, I checked the product description but unfortunately it did not list the specific Realtek chip used. However going through the pictures in the reviews the label on the transceiver does clearly indicate RTL8261C:

View attachment 47906

Pricing at this moment is pretty reasonable at $22.99. Interestingly QSFPTEK's implementation dropped 1000base-t support. The transceiver is 10Gb/5Gb/2.5Gb only. This isn't the one universal 10Gbase-T SFP+ to rule them all.

Photo on amazon listing clearly states 1Gbe support. But these are upto 30m modules so I assume that they run hot. I am only interested in 100m modules because they run far cooler and don’t require heatsinks.

 

[obovata]

Photo on amazon listing clearly states 1Gbe support. But these are upto 30m modules so I assume that they run hot. I am only interested in 100m modules because they run far cooler and don’t require heatsinks.

Yeah it's confusing since they didn't update the product photo (nor double check their spelling). The description warns "Supprots 5Gb/2.5Gb rate over Cat 5e cable or better,up to 50 meters. NOTES: NOT Supprots 1000"

One reviewer also confirms: "Also, something important to note is that the SFP does not negotiate to 1Gig or will work 1Gig, even if you hard set the port."

 

[jmsq]

I was trawling the Microsoft Update Catalog for networking drivers that certain vendors make difficult to directly acquire via automation (Realtek) or make available outside of OEMs (Qualcomm/Mediatek). I came across a find that might make the RTL8127 (10Gbe PCIe) slightly more appealing compared to older/established equivalents: a Windows ARM64 driver. Not sure if this is due to it being the 10GbT chipset of the DGX Spark or something else, but it means that (in theory, have yet to test) you could throw either the RTL8127AT or RTL8127AF into a pcie Thunderbolt enclosure, and it should just work on the Qualcomm X Elite laptops or other Windows on ARM devices. Why not use the RTL8159 instead? In most cases you probably should, but to my knowledge the current crop of WoA devices do not support USB3 2x2 20Gbps mode and would top out below 10Gbps, if that matters to you

 

[gxcreator]

Windows 11 25H2 with the latest driver from Realtek website, and is the NetAdapterCx version 1157.21.20.1110.

the strange is that I have an 5GbE also from Realtek that I get only 2.8Gbps.
Testing the 10GbE one on my desktop that has a 20Gbps USB port, I get full line speed

Same for me, I've got 6.1 Gbps TX and 4.4 Gbps RX on 10Gbit port (AM5 B650).

 

[meego]

I have been holding out on buying the RTL8159-based Diewu/XikeStor/Lekuo USB adapters because they're still quite bulky.

My hope was for something as small and light as Ubiquiti's 5GbE USB adapters to show up at some point. That one is only 38g!



But I have been wondering if it's reasonable to expect something that small.

Given Realtek's datasheets are under NDA, I had Claude take a shot at this.

Here is what it had to say on the subject.

> Any hope of a <50g 10GbE adapter some day? Maybe something based on RTL8159?

There's very real hope, and RTL8159-based adapters are already shipping. Here's why this chip changes the game:

The RTL8159 is a huge deal because it's a pure USB chip — it uses USB 3.2 Gen 2×2 (20 Gbps) natively rather than tunneling PCIe over Thunderbolt/USB4 like every AQC113-based adapter does. That architectural difference matters enormously for size and power. The RTL8159 has a typical power consumption of just 1.95W [Amazon], which is roughly what the AQC113 draws at 10GbE but without needing the overhead of a Thunderbolt/PCIe bridge. Lower power means less heat, which means less heatsink mass needed.

The early RTL8159 adapters already appearing on AliExpress [AliExpress] and Amazon [Amazon] are mostly generic Chinese-brand units with aluminum housings and fanless designs. They advertise running cool enough for passive cooling and throttle-free sustained 10Gbps [Amazon]. If we look at the precedent set by the RTL8157 (the 5GbE sibling), Ubiquiti managed to squeeze that into a 38g stick. The RTL8159 draws more power but not dramatically so — the RTL8157 consumes under 1.5W [Amazon], so the RTL8159 is only modestly higher.

A sub-50g 10GbE adapter is plausible but will depend on a few things. The RJ45 jack itself is a physical constraint — it's bulky and heavy. The real wildcard is whether a brand like Ubiquiti or a similar design-focused OEM pushes the RTL8159 into a form factor as aggressive as their 5GbE stick. The thermal headroom is tighter with the extra ~0.5W, but it's not insurmountable in an aluminum stick design.

The other big advantage is cost: Realtek hinted these 10 Gbps chips would be less than double the price of their 5 Gbps chips, which currently sell in adapters for about $15 [CNX Software]. So we could see RTL8159 adapters in the $25–40 range, compared to $150+ for AQC113-based ones.

The main caveat: reaching full 10Gbps throughput requires a USB 3.2 Gen 2×2 host port [Club386], which is still uncommon. On a standard USB 3.2 Gen 2 (10 Gbps) port, throughput tops out around 7.2 Gbps [Lekuo]. Still much better than the 5GbE ceiling, but not the full 10G. On Thunderbolt/USB4 ports it should work fine with full bandwidth, though driver maturity on macOS is still an open question.

So yes — a compact, lightweight, cheap, fanless 10GbE USB stick is probably a 2026 reality rather than a dream.

It globally checks out. The "typical" 1.95W power consumption is from XikeStor. I'm thinking we can understand it as a maximum power consumption since it lines up pretty well with the RTL8157's 1.5W max power consumption.

I'm optimistic!

 

[SlowmoDK]

I have been holding out on buying the RTL8159-based Diewu/XikeStor/Lekuo USB adapters because they're still quite bulky.

My hope was for something as small and light as Ubiquiti's 5GbE USB adapters to show up at some point. That one is only 38g!

The Diewu is 66g and feels like a quality assembly.

Right now I have a hard time imagining a sub 50gram device without sacrificing build quality or thermals

But hopefully time proves me wrong

 

[Bulls729]

…But I have been wondering if it's reasonable to expect something that small….

iOCrest makes a USB4 10GbE Adapter that is 4x1.6x1in so not too much bigger, but not based on the RTL chipset.

 

[Bulls729]

Cable Matters has an RTL8159-based adapter up on Amazon now: Cable Matters USB C to 10Gb Ethernet Adapter

Nice to see a recognizable brand getting product to market on the 8159, but their marketing on this listing is pretty bad and honestly it's going to cause confusion for a lot of the less technical buyers who find it.

The product title leads with "Compatible with USB4 & Thunderbolt 5/4/3" like that's the headline feature, and then they quote "real-world speed" of ~7Gbps on Windows and ~5Gbps on macOS. On the Mac side they even acknowledge the OS only reports a 5Gbps link rate. They're framing those numbers like it's just normal 10GbE overhead, when what's actually happening is the adapter is bandwidth-starved because TB and USB4 ports are not the primary interface for the 8159.

The RTL8159 is USB 3.2 Gen 2x2. It needs a 2x2 host, two bonded 10Gbps superspeed lanes for 20Gbps of USB bandwidth, to actually deliver 10GbE. That ~7Gbps figure is what happens when you plug it into a port that only hands it a single Gen 2 lane, which is exactly what a USB4 or Thunderbolt port does when it sees a USB 3.x device.

For anyone following the RTL8159 and wondering why a "20Gbps USB adapter" doesn't work at full speed on a "40Gbps Thunderbolt port." The USB-IF created one of the worst branding situations in consumer tech history here, and products like this are the direct result.

USB development essentially forked into two completely different paths. Gen 2x2 was the last evolution of traditional USB 3.x take Gen 2 signaling at 10Gbps and bond two lanes together for 20Gbps. Same underlying USB superspeed protocol, just doubled up. It needed its own host and device controllers specifically designed for that lane-bonding approach.

USB4 went in a different direction. Intel donated the Thunderbolt 3 protocol to the USB-IF as a royalty-free specification, and that became USB4. So USB4 is fundamentally Thunderbolt, it tunnels multiple protocols (PCIe, DisplayPort, USB 3.x) over a shared link using what is essentially a PCIe-based fabric. That's why USB4 is backwards compatible with TB3/TB4/TB5, they're all speaking the same language at the physical layer.

So you ended up with two technologies that both use USB-C, both offer 20Gbps+, and both have "USB" in the name, but they are architecturally different. Gen 2x2 bonds traditional USB superspeed lanes. USB4/Thunderbolt tunnels protocols over a PCIe fabric. Different controllers & signaling that are not interchangeable in the way the shared branding implies. A USB4/TB controller doesn't have the lane-bonding hardware that 2x2 requires, and a 2x2 controller doesn't use the tunneling protocol that USB4/TB uses. When a Thunderbolt or USB4 port encounters a USB 3.x device like the RTL8159, it just falls back to standard USB 3.x compatibility, single Gen 2 lane, 10Gbps, done.

This is why the Aquantia/Marvell-based adapters (AQC113 etc.) deliver 10GbE over Thunderbolt and USB4. They aren't USB 3.x devices, they tunnel PCIe over the TB/USB4 link, using the native protocol of those ports.

None of this is a knock on the RTL8159 itself. For anyone with a proper Gen 2x2 port, mostly certain desktop boards with the right host controller, this is going to be a killer value for 10GbE over USB. But Gen 2x2 ports aren't common, especially on laptops, and that's the disconnect with Cable Matters marketing this thing at the Thunderbolt/USB4 crowd.

I left a review on the Amazon listing trying to explain all of this for regular buyers, but figured it was worth flagging here too since I know people in this thread are following the 8159 rollout. If Cable Matters' product team is paying any attention, they really need to rework that listing. Lead with the Gen 2x2 requirement, put TB/USB4 in the fine print where it belongs. The USB-IF already made this confusing enough without vendors making it worse.

 

[Shonk]

" Nice to see a recognizable brand getting product to market on the 8159 "

Its not really a recognizable brand they simply buy in from china direct and have customized branding
in alot of cases you are far better getting it direct

this is a perfect example they have ordered a customized version
then when they have arrived the person charged with doing the marketing material for cable matters
has no idea of the specs and such to market it so messed the whole thing up

 

[SlowmoDK]

I left a review on the Amazon listing trying to explain all of this for regular buyers, but figured it was worth flagging here too since I know people in this thread are watching the 8159 rollout. If Cable Matters' product team is paying any attention, they really need to rework that listing. Lead with the Gen 2x2 requirement, put TB/USB4 in the fine print where it belongs. The USB-IF already made this confusing enough without vendors making it worse.

There is nothing wrong with that listing, very sober and explains use cases.
Not a single wrong claim is made, even states expected speed using USB4/Thunderbolt, and for best performance use usb3.2

 

[Shonk]

There is nothing wrong with that listing, very sober and explains use cases.
Not a single wrong claim is made, even states expected speed using USB4/Thunderbolt, and for best performance use usb3.2

Thats why you should not work in marketing
your another one who doesnt know what he is talking about
everything Bulls729 said is correct

 

[SlowmoDK]

Thats why you should not work in marketing

Please point me to a single fault in the listing

 

[Shonk]

Please point me to a single fault in the listing

Its not my job to teach you how to drink milk
dont run your mouth about things you have a surface level understanding of

 

[Bulls729]

" Nice to see a recognizable brand getting product to market on the 8159 "

Its not really a recognizable brand they simply buy in from china direct and have customized branding
in alot of cases you are far better getting it direct

this is a perfect example they have ordered a customized version
then when they have arrived the person charged with doing the marketing material for cable matters
has no idea of the specs and such to market it so messed the whole thing up

Yes, they're sourcing from the same Chinese ODMs as everyone else, and yes in a lot of cases you can find the same underlying product for less if you go direct or grab a white-label version. That's true of most of the accessory brands in this space. But Cable Matters has built a presence over the years, they show up in buying guides, they get recommended on forums like this one and on Reddit, and they have a track record with reviewers. For a less technical buyer shopping on Amazon, seeing "Cable Matters" on a listing carries more weight than seeing a random string of letters with zero review history. They occupy the same tier as Monoprice, UGREEN, Anker, where the average consumer sees the name and assumes a level of vetting and quality above the no-name stuff. Whether this was a deliberate framing choice or just someone on their marketing team not fully grasping the distinction between Gen 2x2 and USB4/Thunderbolt, the end result for the buyer is the same.

EDIT: To clarify further, when these newer Realtek chips first hit the market, whether it's the RTL8127, 8159, or whatever comes next, the purchasing path is always the same. You're hunting down listings on Aliexpress, Alibaba, eBay, hoping the seller is legit, waiting on shipping from Shenzhen, and sometimes buying in bulk just to get a reasonable per-unit price. Eventually IOCrest picks them up, then the broader brands follow. Cable Matters getting an RTL8159 product onto Amazon matters for the people who aren't going to navigate that import pipeline.

There is nothing wrong with that listing, very sober and explains use cases.
Not a single wrong claim is made, even states expected speed using USB4/Thunderbolt, and for best performance use usb3.2

Your'e correct, nothing on the listing is a false claim. My point is about framing and emphasis.

When the product title leads with "Compatible with USB4 & Thunderbolt 5/4/3" and the ~7Gbps/~5Gbps figures are buried as "real-world speed" disclaimers, it presents an architectural bottleneck as if it's just normal benchmark-to-reality variance. Those aren't in the same category. One is expected overhead, the other is the port physically not having enough bandwidth for the chipset. The distinction matters to a buyer deciding whether this is the right adapter for their setup.

I get that from a support perspective, it makes sense to set expectations for the ports most buyers actually have. Most people shopping for a random USB NIC probably do have USB4 or TB and not a Gen 2x2 port, and I can see the logic in getting ahead of that. But the listing could accomplish the same thing by leading with the 2x2 requirement and noting TB/USB4 performance as secondary, rather than the other way around. Right now the hierarchy is inverted.

Anyone buying 10GbE should know what they're getting into. But this is Amazon, and the reality is most buyers don't cross-reference datasheets before clicking buy. It's the same thing that happens with "Cat7 and Cat8" cable listings. People see "Cat8 40Gbps Gaming Cable" for $15 and grab it for their 1Gbps router when their existing Cat5e patch cable running 10 feet would perform identically. But bigger numbers sell. Same dynamic here.

 

[Shonk]

Yes, they're sourcing from the same Chinese ODMs as everyone else, and yes in a lot of cases you can find the same underlying product for less if you go direct or grab a white-label version. That's true of most of the accessory brands in this space. But Cable Matters has built a presence over the years, they show up in buying guides, they get recommended on forums like this one and on Reddit, and they have a track record with reviewers. For a less technical buyer shopping on Amazon, seeing "Cable Matters" on a listing carries more weight than seeing a random string of letters with zero review history. They occupy the same tier as Monoprice, UGREEN, Anker, where the average consumer sees the name and assumes a level of vetting and quality above the no-name stuff. Whether this was a deliberate framing choice or just someone on their marketing team not fully grasping the distinction between Gen 2x2 and USB4/Thunderbolt, the end result for the buyer is the same.



Your'e correct, nothing on the listing is a false claim. My point is about framing and emphasis.

When the product title leads with "Compatible with USB4 & Thunderbolt 5/4/3" and the ~7Gbps/~5Gbps figures are buried as "real-world speed" disclaimers, it presents an architectural bottleneck as if it's just normal benchmark-to-reality variance. Those aren't in the same category. One is expected overhead, the other is the port physically not having enough bandwidth for the chipset. The distinction matters to a buyer deciding whether this is the right adapter for their setup.

I get that from a support perspective, it makes sense to set expectations for the ports most buyers actually have. Most people shopping for a random USB NIC probably do have USB4 or TB and not a Gen 2x2 port, and I can see the logic in getting ahead of that. But the listing could accomplish the same thing by leading with the 2x2 requirement and noting TB/USB4 performance as secondary, rather than the other way around. Right now the hierarchy is inverted.

Anyone buying 10GbE should know what they're getting into. But this is Amazon, and the reality is most buyers don't cross-reference datasheets before clicking buy. It's the same thing that happens with "Cat7 and Cat8" cable listings. People see "Cat8 40Gbps Gaming Cable" for $15 and grab it for their 1Gbps router when their existing Cat5e patch cable running 10 feet would perform identically. But bigger numbers sell. Same dynamic here.

That was a well thought out and explained reply
i couldnt be bothered doing it as they generally double down no matter what you say

 

[Bulls729]

Timely example of exactly what I was talking about. Jake (Jakkuh, formerly of LTT) just posted his MacBook Neo review and at around 17:31 he tests an RTL8159-based adapter. The Neo doesn’t have USB 3.2 Gen 2x2, it tops out at a single Gen 2 lane (10Gbps) on the rear USB-C port. So the adapter is already bandwidth-limited before you even get to the macOS side of things which the macOS driver stack doesn’t currently recognize these RTL8159 adapters as 10GbE controllers, so the OS itself is capping the link rate at 5GbE.

Jake seemed genuinely confused about what was happening, and he’s not someone you’d call technically illiterate. But the marketing on these adapters is poor. He hit the exact wall that we’ve been discussing in this thread, bought/received a “10GbE USB-C adapter,” plugged it into a USB-C port, and didn’t get 10GbE.

This is exactly the kind of person who would read the listing, see “Compatible with USB4 & Thunderbolt” or even just “USB-C 10GbE,” and have no reason to think twice about plugging it into a Mac with a USB-C port.

 

[WhiteNoise]

You are correct. Unfortunately not all USB4 (and thunderbolt) can do USB 3.2 2x2, actually most don't. USB4 only mandates USB 3.2 Gen 2x1.

As I said in another thread, USB 3.2 2x2 is actually pretty rare, it would have been better if Realtek made this a USB4 native device.

I put most of the blame of this mess on USB

 

[Bulls729]

You are correct. Unfortunately not all USB4 (and thunderbolt) can do USB 3.2 2x2, actually most don't. USB4 only mandates USB 3.2 Gen 2x1.

As I said in another thread, USB 3.2 2x2 is actually pretty rare, it would have been better if Realtek made this a USB4 native device.

I put most of the blame of this mess on USB

Trust me, I went in depth about this in post #209

 

[Shonk]

Gen 2 x 2 was always an optional part of the USB 4.0 i dont actually know of any chipset (e.g. PCH) that implemented it
Intel didnt the only company that may have is asmedia on an addon controller