I know, I know…we won’t know until Framework tells us. But I thought I’ll start a thread and see if there’s any / much interest in this. e.g. Show / vote our interests in having this.
(Just saw this video…cheap 5Gbps interface, which prompted the question in my head)
2.5 Gbps is already fast enough for most applications using ethernet, so there doesn’t seem to be much benefit switching to 5 Gbps.
The Framework Ethernet expansion module is already pricey at 39$ so there isn’t a really big point in making a 5Gbps one that is even more expensive if not a lot of people are going to buy it.
To few / some / most → unknown. Having this thread may trigger that discussion. What use cases are out there among FL users?
I would gladly buy 2 of them. Main use case is transferring virtual machines between machines and NAS. It would almost half the time I spend transferring over the network (When in a hurry I’ll just dump them onto an external SSD and sneakernet them to other machines).
I’m an early adopter of 2.5Gbps. Key was the advent of low cost switches.
I would like to upgrade to 5Gbps for my NAS, and it won’t cost much more than 2.5Gbps. The low cost 5Gbps PCIe chipset is the Realtek RTL8126. The low cost USB 3 chipset here would be the Realtek RTL8157. These are low cost adapters - the PCIe adapters are now widely available on AliExpress, in both PCIe and M.2 formats. The 2.5Gbps RTL8125/RTL8156 were pin-compatible with earlier gigabit controllers, so they were easy to implement and didn’t cost that much more. The 5Gbps controllers should be the same, although they run a little hotter.
There are still no affordable 5Gbps switches but they’re coming, reportedly this month: Realtek Previews Platform for Sub-$100 5GbE Network Switches
Everyone asks - why not just go to 10Gbps? Old enterprise NICs are very affordable. It’s the size, power and heat of these NICs that turns me off. Some of them are 10-15W! And the switches aren’t that affordable, even used.
These NICs will be about 2W. Much more power efficient, much cooler.
I just got a 5Gbps RTL8126 PCIe NIC for my NAS, and it cost the same as the 2.5Gbps NIC it’s replacing. I’m eagerly awaiting a switch. I would gladly add one to my Framework and it seems like the 5Gbps controller can be swapped into the 2.5Gbps design without many other changes. But this is a niche case for sure.
I can’t say I’ve used my 2.5Gbps Framework expansion card much, but it is nice to have it, and if there is a 5Gbps equivalent I would get one. Just waiting on that switch!
Oh and before anyone says “eww, Realtek”, google the Intel I225-V. By contrast, the RTL8125 never had such problems. I’ve had no problems with the RTL8125 on two desktop motherboards and a PCIe NIC, the RTL8156 on a USB NIC on a Mac and I’ve had a Realtek-based 2.5Gbps switch running fine for a year now, which replaced a Broadcom-based 2.5Gbps switch which got fried by a bad power supply.
I’m not saying that there isn’t a use case for 5 Gbps networking to be added to the Framework Laptops, it’s just that the value proposition isn’t really there both for the end user and for them as a company. I do see the benefit of 5 Gbps networking for a NAS, but the same benefit can easily be achieved with just a third party adapter. Framework doesn’t have to spend money on designing a new 5Gbps switch expansion card, while the end user can still get the same transfer speed.
Framework’s expansion cards are already quite expensive for adapters and while it would stick out a bit, it is much cheaper to just purchase a USB C expansion card and a separate USB C to ethernet adapter.
Framework has already produced one Ethernet expansion card and if they make a 5 Gbps one, it is likely to be even more expensive than this one. You can easily get the same speed from a decent quality adapter that is only 70% the price. Here’s an example of a decent 2.5G adapter: https://www.amazon.com/UGREEN-Ethernet-Aluminum-Thunderbolt-Compatible/dp/B0CD1FDKT1.
Those RTL8126 with their 1,7W power use ¹ seem to be awesome.
I have a 10G switch at my desk and a 10G connection down to my NAS so I could put 5G or 10G to good use in my Framework 13. I would however probably prefer an external dongle at least for 10G since all my 10G capable RJ45 hardware runs HOT. I fear we might be a couple of years away from low power 10G ethernet, if at all possible.
An USB-4 dock with one of those 5G chips, two DisplayPort and a few USB-C/A would be an instant buy if at all reasonably priced.
As for the Framework module, I think I would like a more compact one than the current one more. Even if that means going back to 1G. 90% of my usecase for ethernet is debugging and configuring network equipment and not pushing massive amounts of data around.
You can easily get the same speed from a decent quality adapter that is only 70% the price. Here’s an example of a decent 2.5G adapter: https://www.amazon.com/UGREEN-Ethernet-Aluminum-Thunderbolt-Compatible/dp/B0CD1FDKT1
That may be true, but a 30% premium (in this case $10) to have a module rather than another dongle could well be worth it for many people. Framework laptops have never been the thing you buy if you’re looking for the absolute cheapest option, after all.
The real problem with 10G Ethernet is that doing it as RJ-45 generates a ton of heat because of the signalling properties of 10G-BaseT, and SFP+ is a huge cage that can’t be compatibly made smaller. 5G-BaseT doesn’t have those properties, certainly not as badly.
When you say that a 5G-BaseT part is not worth it for the company, what assumptions are you making about the cost, viable price, and market size for the part? A lot of people thought that making modular, repairable laptops wasn’t financially viable either, so I’m curious as to what your reasoning is here. It seems to me that it really comes down to the price premium at Framework’s acquisition scale for the 5Gbit switch instead of the 2.5Gbit one, and I don’t really know enough about the BOM and procurement stuff to estimate. They don’t offer a 1Gbit RJ-45 adapter, so the 2.5Gbit premium must have gotten small enough to make it not worth having a second SKU. There’s a price delta between 2.5 and 5 that makes it true again, I’m sure
There are already RTL8126-based PCIe ethernet cards for less than the Framework 2,5G expansion. It would surprise me if it would not be possible to replace the 2,5G SKU with a 5G one for a pretty similar price. 1.7W does seem manageable inside an extension card? Idle power should be a good bit lower.
My WisdPi RTL8157 5GbE adapter just arrived yesterday. Based on my testing I don’t think that chip can be put into the expansion module form factor without substancially improving the cooling capacity and a bit of PCB redesign. It runs pretty hot, especially after sustained full speed transfer, albeit not nearly as hot as something like the 10 gig AQC107/113 but still a bit uncomfortable to the touch. Also the chip itself seems to be a little larger than RTL8156B, at 8*8mm.
We are excited to see everyone’s enthusiasm for the 5Gbps network cards and the support for the Wisdpi WP-UT5 USB 5GbE network card. We are interested in developing this dedicated product for Framework and would like to know if we can count on Framework’s support for this initiative.
Contact Framework directly.
frame.work/contact-us
This is a community forum, and while Framework staff does sometimes read and answer questions, this forum is not an official channel to ask them direct questions.
Does anybody like this ? Preliminary assessment suggests that we should be able to design a dedicated USB 5Gbps network expansion module for Framework.
For people mentioning high power usage and things getting hot for 5 an 10g devices.
It is maybe worth pointing out that single mode optic fibre instead of copper uses far fewer watts.
It think we just need a fibre connector that can self clean the fibre ends and be small enough to use on a laptop to be invented.
I mentiom single mode fibre because its thinner and more bendy and less likely to be damaged.
I currently use 10g thunderbolt adapters with my FW16 with single mode fibre cables, and it stays pretty cool.
Looks great!
If we attempt to launch a crowdfunding campaign, I wonder how interested everyone would be in the product? What price do you think would be appropriate?
I’d be interested for ≤ 60 EUR. [1]
no idea what the manufacturing cost of this module could be, so ignore me if I’m totally out of line with the price point ↩︎
Definitely less than that.
