Hacker News new | ask | show | jobs
by realxrobau 58 days ago
Are there any that actually have a SFP+ port? That's all I want. No one wants to use 10g ethernet when DACs are cheaper than cat7, and you can just change it up to a $7 multimode when you need longer runs.
9 comments

> No one wants to use 10g ethernet when DACs are cheaper than cat7,

You don't need Cat7 for 10G.

Cat6 is spec compliant up to 55mm. Cat6a to 100m, which is the same as Cat7.

If you're doing short runs like to a nearby switch, good Cat5e works fine in practice. I've run 10G over Cat5e through the walls for medium runs without errors because it's all I had. It works in many cases, but you're out of spec.

I use DAC where I can, but most people just want something they can plug into that RJ45 port in their wall that goes to the room down the hall where they put their switch.

There are several SFP+ to Thunderbolt/USB4 adapters on the market. Not cheap, though.

Yep, 10gb over copper is not power efficient so any savings you get from getting a cheap 10gb switch will just go to your power bill. Most cost effective and flexible is a used 25gb switch. Most 25gb switches can do 1/10/25gb. 10gb networking has been dead for over 10 years.
Interesting observation about power use. How close do you think we are to it being practical to wire your whole home with fiber instead of CAT6 or whatever? If you're providing all your own equipment, are willing to purchase a high-end splicer for maintenance, etc.

For laptops I assume you need USB/Thunderbolt adapters. (Still no SFP+ or SFP28 module for Framework?)

For desktops you'd use an SFP28 card (taking up a PCIe slot).

For devices like Raspberry Pi's, etc. you'd use... local RJ45 switches with optical uplink ports?

You can just do a mix.

Most of my devices only need 1G or even 100Mbps. No reason to switch to fiber. 1G/2.5G copper ports don’t use that much power.

For 10G+ things, it’s fiber or DAC first if possible then RJ45 if it’s the only option.

Then my backhaul between rooms is just single mode fiber, good up to 800G. Plug in a small switch at the end and you go back to RJ45 and PoE.

I only have 10G though (to transfer large files/RAWs between my computer and my storage). Something faster would be nice because NVMe SSDs can go 50G+ but that equipment is pricey and power hungry.

If you need 1G or 10GB over copper you can just use a SFP or SFP+ media converter in a 25GB SFP28 switch port. If you have a POE requirement, say for video cameras you either use a dedicated 1GB POE switch or power injector. A 10GBASE-T (RJ-45 copper) switch consumes 3-12 watts per port and a 24 port switch will idle at 50 to 60 watts and run hot. SFP+ and SFP28 ports use under 1 watt per port. I would never recommend a 10GBASE-T copper switch for any use case in this day and age, home or enterprise.
Wiring ports for humans to use in a flexible and future proof manner (as in a single family home, for instance) gains a lot of utility with PoE.

The convenience and flexibility of PoE would always push me towards copper wiring.

>10gb networking has been dead for over 10 years.

Not even close to being true, unless you specifically mean 10Gbps over twisted pair (Cat6/7) cable. SFP+ is the default on a ton of network gear still.

I think the point he is making is that the industry first went with a 10g single link, and then 40g over 4 links. Then they figured out how to do 25g over a single link, and 100g over 4 links. Those 25g/100g are common for enterprise switches. It might be fairer to say 40g is dead, 10g still has use cases.

Edit to add: If you want an example, these are the NVidia ConnectX nics available from FS.com, the lowest end one is 25g, then 100g, 200g etc.

https://www.fs.com/uk/c/nvidia-ethernet-nics-4014

What they mean is that the cost per bit both capex and opex/power is worse for 10G than 25G for a while now as long as you talk about new hardware.

We're at the point where 25GBaud PAM4 is being replaced by 50GBaud PAM4. That's 50 to 100 Gbit/s.

But iirc the use of PAM4 for the faster ones than "only" 25Gbit/s lanes is a hindrance to managing bottom-barrel price-per-bit. PCIe 3 was 8, PCIe4 was 16, and PCIe 5 is 32 GBaud with a line code basically like the 10+ Gbit/s Ethernet links (well, it's 66b/64b for Eth and 130b/128b for PCIe).

The SFP+ ones are all Thunderbolt or USB4 this far, i.e. not backward compatible with USB 3.x, like this QNAP one: https://www.qnap.com/en/product/qna-uc10g1sf
10G DACs are no cheaper than cat6, which is perfectly fine for 10G at most practical distances. Considering the target audience of these cards it seems pretty obvious to me that letting users "just buy a cat 6 cable" is miles more reasonable than having them buy a transceiver or DAC.

As for allowing to switch to fiber, that just seems orthogonal again to what these USB NICs are for, not to mention the SFP+ itself is probably more expensive than the NIC shown here...

DACs are very cheap (second hand and AliExpress) and they never use much W. If both machines are near each other though (which a DAC cable implies) and both run Linux and both support Thunderbolt, you might be better off with a direct ethernet over TB connection. Whether macOS supports such, I don't know.

The other side will then also need a low power NIC (of which fiber and DAC over SFP+ are less power hungry). What this article doesn't mention, is that there are also a lot of PCIe NICs on the market which aren't power hungry (RTL8127), as well as RTL8261C for switches/routers.

I've seen low power RTL NICs with SFP+ on it, too (example: [1]). With SFP+, you'll have a lot more versatility. DAC and SFP+ fiber are very cheap, btw. Especially second hand they go for virtually nothing. I have 10 SFP+ fiber lying around here doing nothing which I got for a few EUR each.

For me as European with high energy prices and solar energy gotten the beat next year (in NL), this is all very interesting.

There's a couple of good reasons why to opt for fiber in the home. You keep the energy between the different groups separated which can help. I also find fiber very easy to get through walls, allowing me to have multiple fiber connections through walls (currently I use 1x fiber + 1x ethernet for PoE possibilities from fusebox).

With all above being said, AQC100S is low power and does not get very hot. You can get these with SFP+ and PCIe/TB. They've been available for a while.

[1] https://nl.aliexpress.com/item/1005011733192115.html (no vouching for, just first hit on search)

I just wish someone would come out with a PCIE 4x1 capable card with SFP - my main desktop’s non-GPU expansion slots are all 4x1 electrically and even the one you linked is a 3x2. As far as I can tell the only 4x1 cards available are RTL8127 or AQC113 RJ45 ones :(

I suppose an NVME riser is also an option, albeit janky.

There are RTL8127 cards with SFP+, e.g. https://www.lekuo.com/product_view.php?id=659

edit: on looking closer, that still seems to be an x4 card.

Says electrically 3x2.
Right, but I don’t think a x2 slot exists so hence being physically a x4 card. If you had an open ended x1 slot you might be able to run as PCIe v4 x1.
I can also buy a roll of CAT6 and a few dozen dollars in tools and RJ-45 connectors and make my own custom length cables.
Also SFPs are always a gamble. Might work, might not, you have multiple options, meanwhile with copper RJ-45 you are guaranteed that a link will be established.
> No one wants to use 10g ethernet when DACs are cheaper than cat7,

Ethernet is media independent. Yes, yes, it was first classified for thick net, but ethernet over twisted pair (rj45 typically) is still ethernet despite the lack of vampire taps. You can run ethernet on thick or thin coax, twisted pair, dac, fiber, or even over the ether so to speak.

That said, 10g over rj45 is pretty handy when you have existing wire in walls. In my experience, it runs fine on the cat5 (not even cat5e) that's already there. Maybe it won't work on all my runs, especially if I tried all at once, but so far, I'm two for two.

The spec is for ~ 100m in dense conduit; real world runs in homes are typically shorter and with less dense cabling... and cabling often exceeds the spec it's marked for, so there's wiggle room.

I have a fairly large house (2 story 3k sqft) with all cat5e. I iperf’d every run and they could all do 10gb negotiation and TCP, most of the runs could sustain very high UDP rates with low packet loss. There’s just one run (which is the one to the internet) which had a slightly higher UDP packet loss rate. So basically every run can do 10gb fine. Been running the whole network like this for a year. It’s been great! I just need a 10 gig capable NAS. My current one can only do 3.5 or so because it’s a usb 5gb/s which isn’t really 5 gb.
Modern transceivers can do 10G on absolutely garbage twisted pair. My house was wired with absolutely dire cat5 cabling. Zero shielding and barely any copper in the pairs. I thought I'd barely be able to do 1G on them, but modern transceivers (amazon) easily do 10G over like 30M of that sort of cables.

In fact I had more trouble getting quality fiber working for that sort of distance than El Cheapo cat5. They do heat up a bit, but they work wonder.

Zero shielding may actually help. Shielding acts as an antenna when not properly grounded and continuous, which is more common than not.
Having recently terminated a nice name brand Cat6A jack (ICC brand), I’m quite unimpressed by the state of category cable shielding:

The termination comes with basically no instructions.

The connector shell contacts the shield by squishing it between the shell and the cable jacket. With a zip tie, helpfully supplied with the connector. Good luck applying much (any) compression force. Never mind that CAT6A cable jackets come in various sizes, are not manufactured to tight tolerances (why would they be?), and are quite squishy.

Compare to a high quality F connector, which is highly engineered, less expensive than a decent Cat6A jack, and makes excellent contact to the cable shield without much fiddling.

The switch that connects to it has no grounding terminal of any sort. It uses an external isolated power supply. Admittedly, the ground connection on a switch is probably doing very little at frequencies of hundreds of MHzz

The big bulky black box this little adapter replaces in Jeff's uses is actually just a PCIe/OCP card in an enclosure and you can replace that with a 10g card with SFP.
I’ve been using the qnap sfp+ thunderbolt one (I think it’s a marvel/aqantia chip) for a few years now everyday with my MacBook and it’s been solid
I would rather use Ethernet where possible. I used SFP28 for a while, but this meant an extra networking card was needed in each PC. Ethernet is universal, and now that bandwidths are catching up, I no longer see SFP as necessary in a typical home or small office network.
I think you are mixing it quite a bit. SFP+ is still Ethernet, also SFP28 gives you speed (25Gbps) RJ45 will never do, so the extra card for the extra speed is mandatory.