Getting 25 Gbps Thunderbolt Ethernet on My Mac Studio
Posted by speckx 3 days ago
Comments
Comment by Neywiny 3 days ago
The upside, though, was that I plugged it in, got my throughput, and never had to worry ever again. Remember that engineering time is mucho $$. So troubleshooting overheating, divising a solution, manufacturing it, etc, that all eats into the savings.
The $400 dock would be cool though. It seems smaller than the sonnet and fixes my charging problem. Weirdly I didn't think I'd care about not including transceivers but they did actually come in handy, so that is a downside. Might try to get one of these if I need another one in the future.
Comment by itopaloglu83 2 days ago
Comment by Neywiny 2 days ago
Comment by voxadam 2 days ago
I initially read this as SONET[1] and was seriously questioning your technical choices.
[1] https://en.wikipedia.org/wiki/Synchronous_optical_networking
Comment by danw1979 2 days ago
what a man does in his own homelab, even if it does involve ATM, is none of our business.
Comment by pyvpx 2 days ago
Comment by wolvoleo 1 day ago
Comment by danw1979 1 day ago
It was 53-byte “cell” switched, but used virtual circuits as a path through the network. The data plane had no globally routable address in the cell header. I don’t think most people considered it a “packet” network, and being an ITU standard it definitely belonged on that incumbent, connection-oriented, telco side side of the protocol wars.
Whatever anyway, it was weird.
Comment by wolvoleo 20 hours ago
However when I worked with SDH it was particularly popular because some customers like banks vetoed ATM because they were incredibly worried one of their packets would make its way to a third party. At that time encryption was not yet commonplace. I always thought that risk was a bit overblown but they were adamant.
SDH was more like a software switched multiplexing (where the previous generation PDH had this in hardware). And it was not exposed to the customer, they just got a symmetric feed of 2mbps or a multiple thereof.
Still, technically the same could happen if we configured it wrong. But for some reason the banks were less worried about it.
But yes it was not packet by packet switched no. I have to admit my knowledge of ATM was never the best as I worked exclusively on the SDH network.
Comment by icedchai 2 days ago
Comment by shrubble 2 days ago
Comment by walrus01 2 days ago
Comment by LargoLasskhyfv 1 day ago
That rang some bells, working in a communal data-centre, serving all of town, and smaller neighbouring towns. At 100Mb/s. Funny thing was the town still used many terminals in its offices. Serial Siemens terminals, speaking to Siemens RM400, RM600 and something called BS2000, also by Siemens. Of course with baudrates, parity, even/odd, start&stop bits nobody else would have used. Because Siemens...
So how do you connect them? By using adapters not much bigger than a matchbox, somehow powered by the serial port! At insane prices per piece, more than the terminal itself.
Shrug? Vee arr government, vee don't need to care about zat!1!!
(Town is broke by now, btw.)
Similar things in all police stations of that state.
(Many more towns, counties, cities broke there, too.)
Reasoning for using FDDI was that there existed equipment (by Allied Telesis IIRC) which could detect someone messing with the fibre somewhere, out of sight.
Not that much later, in the German HQ of some larger insurance company, I had the privilige of playing with FDDI at 1Gb/s. And witness the oh so unbreakable IBM S390(or was it still 370 at the time?) go down hard, several times...
Hrrm. What else did I witness? Oh, some Compaq thing made for a rack, 4 units high, sitting on a desk without cover, making you feel like you'd be in a vacuumer, or right next to an airport. Whooooouuuuiiiiieeeehissssss!
That thing ran Novell Netware, responsible for every login from any client across all offices in town. Handled by a full-length ISA-Card, full eight bits wide! Not 16. Not using the EISA-slots...(facepalm)
Comment by PhageGenerator 2 days ago
> I have SMB multichannel enabled, but this real-world speed limitation could also be related to my use of my lower-power Arm NAS (running on Ampere Altra, with 32 fairly-slow CPU cores).
Based on my own testing with a 25 Gb NIC setup on macOS I strongly suspect the issue is not on the NAS side, and the hardware I was working with is by no means underpowered.
The Mac I was working with is one of those rack mount M2 Mac Pros with a 25 Gb PCI-e card. On the NAS side it was a several hundred terabyte all NVMe flash multi-node cluster, each node with Xeon CPUs and 2x 25 Gb bonded NICs. Doing some rough throughput testing from macOS over multi-channel SMB yields throughput rates that are barely faster than the built-in 10 Gb copper NICs on the Mac.
Using iperf I can confirm that I can get near enough to 25 Gb/s raw network throughput on the Mac. And other non-macOS systems see significantly faster throughput to the same NAS. I am not sure exactly what the issue is that causes 25 Gb NICs to underperform so significantly for file sharing tasks, but I am reasonably certain that the issue is in macOS.
Comment by randusername 2 days ago
Comment by CobaltFire 2 days ago
ConnectX-4 and newer are natively supported. I use a ConnectX-4 LX in a $50 Razer CoreX off of craigslist. Works fine, hit similar limits to what Jeff did.
Using an OCP is the neat trick here. MUCH smaller setup.
As far as the other posters comment about Tx's... that is an entirely different topic. I've never had an issue with FS.com optics personally.
Comment by monster_truck 2 days ago
If you are unfamiliar, the reality of the situation is quite bad. Slight hardware revisions of adapters/switches, which if you are lucky can be differentiated by manufacture date or serial number sequence, will not recognize known good modules, and if they do you then have to actually test if they can establish link, then let them burn for 24+ hours to see if they can maintain it without dropping (often, because of forward error correction, you will not start to see flapping until enough correctable errors have accumulated and the buffer overflows). Then there are firmware updates where this process must then be repeated.
You would think with how expensive all of this shit is you would simply be able to buy the officially listed as compatible components but my experience with them has been that a shotgun approach is best for your sanity and deadlines, buying sample kits of each module and trying them all, and then once you find the ones that play nice, buying as many of them as is reasonable and testing all of them.
If I had a dollar every time two modules from the same brand, that came in the same box, from the same production run, did not work with one another, I would be able to test your theory for free. Network Engineering is Fun
Comment by simoncion 2 days ago
It also sounds like you're not saying that OP cannot "...just put a real PCIe NIC in an eGPU enclosure with a fan", but rather that regardless of where that NIC is situated, it's your experience that SFP quality is poor. Do I correctly understand what you're saying?
Edit: FWIW, I've had no problems with the ~two foot PCI-E riser cable that connects my graphics card to my motherboard. [0] I expect this OCuLink thing [1] will work fine, too.
[0] Taking up several PCI-E slots just for a damn graphics card is really dumb. IMO, mid-grade and better cases should have off-mobo mount points for oversized devices and offer riser cables to connect said devices to the installed motherboard.
Comment by ccssowa 2 days ago
Atto’s new 100G Thunderlinks work well. I’ve seen speeds of just over 5 gigabytes per second. Atto definitely does a lot of kernel extension wizardry to get macOS to achieve these speeds.
A bit OT: I hope Blackmagic really shakes up the 100G connectivity market, and that Apple realizes their platform is extremely well-suited for IP Video 2110 workflows — if they can ever get networking under control.
Comment by pzmarzly 3 days ago
Comment by gorkish 3 days ago
His iperf3 results show that it's able to go full beans on a single 25g port. A cross platform SMB file copy is kind of a meaningless benchmark in this context.
IMO a pretty cool product would be a network switch providing native thunderbolt interfaces to multiple upstream hosts. That wouldn't magically fix this RDMA issue, though. This is just a separate thought. Seems like the kind of thing Mikrotik should make.
Comment by CobaltFire 2 days ago
Speeds aren't any better, but at least it was cheap and fiber.
Comment by cyberax 3 days ago
I found that SMB on macOS is just bad. If I try to heavily load it (e.g. by doing mass conversions in my Audible library), it can cause stalls and disconnections.
I managed to find a set of parameters (in /etc/nsmbd.conf) that makes it work more-or-less reliably by trial-and-error: dir_cache_off=yes, file_ids_off=yes, notify_off=yes, mc_on=no.
It's still terrible, though. It's so bad that Linux in a Parallels VM works _better_.
Comment by monster_truck 2 days ago
It's not just software companies or home enthusiasts either. There are a lot of ie medical practices or accounting/benefits agencies out there who have sworn off Apple for life after having to spend extreme amounts of money on data recovery to hopefully get their years of records back, usually in a do or die situation like an IRS audit which causes them to discover that their basement was flooding and all of the paper records are destroyed.
Comment by areoform 3 days ago
MacOS has supported RDMA over thunderbolt since March 19th.
Comment by wrs 3 days ago
Comment by throw0101d 3 days ago
* https://learn.microsoft.com/en-us/windows-server/storage/fil...
Comment by geerlingguy 3 days ago
I do wonder what Apple's custom-built servers are running for networking. I wish they still made Mac Pro-level chassis specifically for better IO. Thunderbolt 5 is nice, but far inferior to a PCIe x16 slot (which can hit 400 Gbps networking) nowadays.
Comment by xp84 2 days ago
If they were focused on the needs of the kind of person who moves enough terabytes of files around to need this kind of speed, we'd see them still making either computers with slots, or first-party 25Gbps networking.
I'm not really trying to criticize them for it -- Tim Cook's bean counters clearly did the math and they're certainly free to optimize their Mac R&D spend for things that'll sell college students MacBook Neos and Airs, rather than worrying about the MKBHDs, Linus Sebastians and Jeff Geerlings of the world (this is just a list of people I've seen using extremely fast networking).
Comment by tomjakubowski 2 days ago
Comment by xp84 3 minutes ago
So, it seems that Tim Cook decided at some point that while it's nice to make $5,000 in profit from a Mac Pro sale every day or two (okay, I kid, I'm sure they sell a bit more than 365 a year), even if they hang those "pros" out to dry completely and they defect to Windows, Apple would be better off spending their time focusing on ways to make the MacBook Air and Neo more attractive or profitable because they move orders of magnitude more units of those.
Note: I don't like this, but my feelings have nothing to do with it, I think it's just a classic Tim Cook Numbers-first business decision
Comment by geerlingguy 2 days ago
They kinda hit a standstill on the networking side as they moved to the trash can Mac, and I think that had killed momentum towards any high speed IO besides the Thunderbolt ports moving forward.
I was hopeful they'd keep the Mac Pro going into the M-series era but after one version... it's gone.
Comment by kridsdale1 2 days ago
Comment by dlcarrier 3 days ago
Comment by olavgg 3 days ago
Comment by gorkish 3 days ago
The perf is amazing
Comment by dlcarrier 3 days ago
Now, I'm off to determine if an eMMC M.2 module is a native PCIe device…
Comment by olavgg 2 days ago
If you buy CX3 and Mellanox SX6036(easy fan adjustment and low power usage) switches on Ebay for cheap. Which are great up to 56Gbps. You can run infiniband which is way easier to setup, just remember you still need to enable IP over IB to start the handshake for the RDMA connection.
Comment by lmz 2 days ago
Comment by esseph 3 days ago
Comment by teaearlgraycold 3 days ago
Comment by AnthonBerg 3 days ago
Comment by m463 3 days ago
Comment by teaearlgraycold 3 days ago
Comment by NooneAtAll3 3 days ago
Comment by AnthonBerg 2 days ago
Comment by kalleboo 2 days ago
Comment by throwaway219450 3 days ago
See also NVMe drives? They draw enough idle power to drain a laptop battery overnight, and will easily hit 60+C in a bad enclosure. That’s only a 10Gbps module. The USB4 enclosures are all heatsink.
Comment by jauntywundrkind 3 days ago
And your computer already speaks the PCIe and thunderbolt. It has that probably on CPU. Here we have to receive USB, convert it to Ethernet. On much less up to date process nodes than your CPU.
Third, until very recently for even 10Gbit it meant using a usb4<->pcie bridge chip which itself was hot, to provide a generic pcie connection, which then talked to a conventional network chip, so it was a double hop: USB <-> pcie <-> nic. I don't know if 25Gbit to USB direct convert is available yet but we've seen a wave of vastly cheaper 10Gbit adapters come out that are single chip USB<->nic, probably not pcie at all, that are much much cooler. 25Gbit is probably still not here yet, maybe?
Comment by colechristensen 3 days ago
With ethernet it's for lots of reasons but a big one is not creating ground loops between distant equipment. If you didn't isolate the network cable would be sinking potentially amps of current between distant devices and that just isn't sustainable. You don't have this problem attaching USB between two devices.
Comment by wmf 3 days ago
Comment by ssl-3 3 days ago
Comment by wmf 2 days ago
Comment by StillBored 2 days ago
(but its also displayed in the pcie PowerLimit capabilities you can see with lspci)
Comment by altairprime 2 days ago
Comment by frollogaston 3 days ago
Comment by teaearlgraycold 2 days ago
Comment by fragmede 2 days ago
Comment by frollogaston 1 day ago
Comment by Aurornis 2 days ago
The NIC in this setup was released 12 years ago. It's built on a very old process.
Comment by monster_truck 2 days ago
Comment by Palomides 3 days ago
you can look up the rated power use of a connectx4 card, they do a lot of offloaded computation, too
Comment by selectodude 3 days ago
Comment by alsetmusic 3 days ago
Comment by GeekyBear 3 days ago
> this real-world speed limitation could also be related to my use of my lower-power Arm NAS (running on Ampere Altra, with 32 fairly-slow CPU cores). I was testing writing to an array of fast enterprise NVMe SSDs, so they shouldn't be the bottleneck here.
He was only getting 1 Gigabyte of throughput out of the built in 10 Gigabit Ethernet connection and moving to the 25 Gigabit connection only made a marginal impact on throughput, after all.
Comment by dcrazy 2 days ago
Comment by wmf 2 days ago
Comment by geerlingguy 2 days ago
Comment by nutjob2 3 days ago
So for instance you have two or more machines that have USB 4 ports at 20 to 80 Gbps and you plug in the appropriate USB C cables to connect them.
What do you need to make the connections appear as a (point to point) part of a regular TCP/IP network? Seems much cheaper than Ethernet if you already have the ports.
Comment by imoverclocked 3 days ago
I believe MacOS supports it via the usual network control panel. I’ve used it for transferring data from an older laptop to a newer one. The cable is more expensive than fiber, last time I checked.
Comment by QuantumNomad_ 3 days ago
Comment by davkan 3 days ago
Latency was an order of magnitude higher than twisted pair but the throughput was greater than 10Gbps and I didn’t have to spend on 10GbE capable servers and network equipment. I vaguely remember that they had to be proper Thunderbolt not just USB-C but that could have changed it’s been quite a while.
Comment by warpflyght 3 days ago
Comment by ssl-3 3 days ago
But for all certified Thunderbolt 4 and 5 ports, this kind of networking support is a requirement. (It is optional with Thunderbolt 3.)
The rest is up to the operating system.
Comment by robin_reala 3 days ago
Comment by ianburrell 2 days ago
Thunderbolt 4 is profile with USB4 tunneling, 40Gbps speed, and 100W power. Thunderbolt 5 is 80Gbps and 240W power.
Comment by undersuit 2 days ago
Comment by kccqzy 2 days ago
Comment by rconti 2 days ago
Plugable 2.5G USB-C to ethernet adapter (bought November 2025), SABRENT USB C to 2.5Gbps ethernet adapter (bought October 2024), UGREEN USB C 2.5G ethernet adapter (also October 2024). In my experience they immediately fall on their face in a speedtest and drop to a couple hundred Mbps, experience packet loss, etc. This is testing across a few modern Apple Silicon Macs.
I now have a WavLink WL-NWU341G 5Gbps dongle with the RTL-8157 that seems to work well, but I've only run a couple of tests-- I am away from my home setup and don't have wired ethernet run to all the rooms where I'm living.
If you do some Googling, there's some good information out there about it, but when I bought the bad ones, I couldn't find any information corroborating my experience.
Comment by a_carbon_rod 2 days ago
Comment by kalleboo 2 days ago
Comment by myelin 2 days ago
On a CalDigit TS3 dock, I've had one working off the rear Thunderbolt socket for over a year, but it fails in minutes if plugged into the USB-C one. I've never had issues with a CalDigit Elements (TB4) hub. I've had success plugging them directly into various Chromebooks. On one Windows laptop, they seemed to work on one USB-C port but not the other.
So... reliable once working, but getting there can be a pain.
Comment by kalleboo 2 days ago
These were my notes
* Front 5Gbps USB-C Fresco Logic
* Front 5Gbps USB-A Fresco Logic
* Rear 10Gbps USB-C ASMedia
* Rear USB-A, Right-hand 3 Fresco Logic
* Rear USB-A, Left-hand 1 ASMedia
Comment by rconti 2 days ago
My logic was that I didn't want to spend $400+ on a TS4 for my WORK computer, just to get a faster ethernet for the lulz. I was happy to spend $25 for bragging rights.
Of course, now that I've spent that same $25-30 4 times, it's beginning to feel like less of a bargain.
Comment by ridiculous_fish 2 days ago
For my Macs, the IOCrest USB4 adapters with the AQC113 chipset have been rock-solid and can saturate my 10Gbe network. They do run hot but no stability issues. That's what I use and recommend.
The newer RTL8159 based dongles are smaller, cheaper, cooler, and can do 10 Gbe on paper. But they top out at 7 Gbps on Macs, which don't implement the USBx2 spec.
Comment by thrdbndndn 2 days ago
Comment by justincormack 2 days ago
Comment by egorfine 2 days ago
Comment by lostlogin 2 days ago
Buying a Mac with built in 10gbe is so nice compared to dongle hell.
But it’s not 25gbe
Comment by mgerdts 2 days ago
Comment by rowanG077 2 days ago
Comment by qurren 2 days ago
Comment by dotwaffle 2 days ago
Comment by geerlingguy 2 days ago
Comment by nodesocket 2 days ago
Comment by brcmthrowaway 2 days ago
Comment by washbasin 2 days ago
Comment by airspresso 2 days ago
Comment by geerlingguy 2 days ago
The 4 Mac Studio cluster I tested with Exo and Thunderbolt RDMA was a loaner cluster from Apple. There's no universe in which I could justify buying that.
I always mention any equipment that's provided / loaned / etc. and the nature of my relationships with vendors—I try to also keep this document updated for transparency's sake: https://github.com/geerlingguy/youtube/blob/master/Sponsors....
Comment by brcmthrowaway 2 days ago
Comment by TacticalCoder 2 days ago
Nothing against Jeff Geerling and I love his YouTube channel and I'm glad he's one of us HNer but this crop of circular "I make YouTube vids about big tx speeds and big NAS because I make vids hence I need big tx speeds and big NAS so I can speak about big tx speeds and big NAS because I make YouTube vids about big tx speeds and big NAS" videos is a bit tiring IMO.
I'm not into cryptocurrencies but, I don't know, host a full Ethereum node, host a Plex/Jellyfin server serving thousands of movies, host LLMs, etc.
I mean: talk about something else than editing vids because you make vids about vids.
Comment by geerlingguy 2 days ago
LLMs: https://www.jeffgeerling.com/tags/llm/
Etc... (I don't do any crypto)
Comment by wmf 2 days ago
Comment by SlavikCA 3 days ago
Author talks about 4k video, but AI tells me that for "4K production mastering (ProRes 422 HQ, 4:2:2, 10-bit)" the rate is under 2Gbps.
I'm not saying it's not needed, - just curious.
Comment by andy99 3 days ago
Edit: https://www.jeffgeerling.com/blog/2025/15-tb-vram-on-mac-stu...
Comment by logicallee 2 days ago
[1] https://www.apple.com/shop/product/mdw94am/a/thunderbolt-5-u...
Comment by gregsadetsky 3 days ago
Obviously it's possible to move this to a NAS, not use a mac, etc. etc. - but just wanted to chime in with my little use case of moving multi-gig-files around. :-)
I was just looking for a thunderbolt-sfp solution. Super timely!
Comment by kllrnohj 3 days ago
That'd be the rate for 1 stream at 1x playback. Multiple streams because of overlays/crossfades/etc... along with playback at >1x (such as during rendering in particular) will change that significantly.
Comment by hephaes7us 2 days ago
A typical NVMe SSD can easily do sequential reads at 7GB/s, which outpaces a 25Gbps connection by more than 2x. NAS is more typically backed by SATA HDDs, but there's some relatively easy way to put an NVMe disk in front of them as a cache.
Do you ever _need_ it? Perhaps not, but there's plenty of common activities just stemming from NAS usage which see a linear speedup, and would continue to get faster up to 100Gbps even.
Comment by fragmede 2 days ago
Comment by SlavikCA 2 days ago
Comment by urams 2 days ago
Comment by QuantumNomad_ 3 days ago
Comment by geerlingguy 3 days ago
I'm in a never-ending pursuit of cutting time when transferring large files though — AI models is one thing, but I have 100-200 GB projects I transfer across for backup or import (using an external TB5 drive capable of 3-5 GB/sec), and if I can do that in 1 or 2 minutes per project instead of 3 or 4, that's a noticeable speedup.
Definitely overkill for 98% of my workflow though. 10 Gbps is fine.
Comment by colechristensen 3 days ago
Comment by aegisora_ai 2 days ago