2.5G Ethernet Switch Question

samops03

Gawd
Joined
Oct 18, 2007
Messages
809
I currently have a PFSENSE router connected to a standard unmanaged switch. I ordered 2 pcie ethernet adapters (Realtek 8125) for my main rig and server. From what i understand i will need a new switch for these to communicate at 2.5G speed? What is the cheapest solution? Are there unmanaged switches that can do this?
 
Hello,

Yes, you would if you want to get the full 2.5Gbps speeds. Any reason why you ordered these specifically? I presume you want better through put to your server from your main rig? You could just run a direct cable from your main rig to the server using the 2.5Gbps ports and set static IP's on them, different than your main LAN subnet and then just use those static IPs for your network shares.
 
There is a qnap switch that’s affordable. But I can’t find it for sale anywhere
 
2.5gb is a waste of time and money just buy the 10gb switch. You won't save any money going with 2.5gb switch. The only thing cheaper is the NIC 2.5gbps and they aren't really supported in Pfsense. The driver kinda sucks if you run 2.5gbps RTL8125 NIC into a existing 1gb or 10gbp because of buffer size issue you will see the throughput cut in half or so. The best thing to do is either just run 1gig ethernet through out or upgrade to all 10gig ethernet 10gb nics and switch. I have been playing around with this for several days now trying to get good throughput from my cable modem CM2000 with 2.5gbp wan to a RTL8125 nic in my pfsense the throughput goes to crap like 500mbps when running 2500base-T but if i force 1000base-t it will run perfect !
 
With used 10Gb nics so cheap and dacs or sfp+ and some cheap pre-termianted fibre for a direct run, 10Gb is the way to go. The only way 2.5Gb is going to ever make sense is when it simply becomes the standard like 1Gb is today.
 
2.5gb makes a lot of sense for average users, where 10gb doesn't. Lots of users have Cat5e jumpers lying around and Cat5e structured cabling.

From a service provider standpoint, delivering internet connections above 1gb obviously require handoffs faster than gig. For that reason I welcome, at the very least, consumer routers with 2.5gb WAN and LAN ports working their way down the product stack.

2.5 just needs to become the new gig, which I foresee happening in the next 3 years in "many" new products. Right now it's locked behind higher end SKUs.
 
Actually, even 2.5Gbps is more than most average users need.

I remember the transition from 10 to 100 and 100 to 1000--the main driver from 10 to 100 was that there were applications that needed the additional bandwidth. And in most cases the existing wiring had to be changed since 100Mbps needed higher quality wiring. There was even a 100Mbps standard that used the existing 10Mbps wiring, but it never caught on even though HP was behind it--the cheaper cost of the now 'traditional' 100Mbps nics and hubs and switches just seemed to outweigh the cabling issues. And it was a good move since when gigabit came along it essentially ran on the same wiring as 100Mbps. But besides servers that needed gigabit to serve so many 100Mbps clients, the need for gigabit wasn't really there. It was the cost of gigabit that did it--I remember a gig nic costing almost the exact same price as 100Mb--and then once all the nics were gig, it probably didn't make sense to get a 100Mb switch when a gig one was 'just a little more'. But the move to gig wasn't because of a true demand.

And I think this same scenario is playing out for 2.5Gb. There are some applications that fit 2.5Gb better than 10Gb, especially when working with existing in-wall wiring. Back when people ripped out 10Mb wiring, there was an immediate known improvement that was needed everywhere--you can't say the same thing about 10Gb as it still isn't needed at the desktop level, and here 2.5Gb really helps--5Gb even more.

Until the chipset for 2.5Gb becomes only negligibly more expensive than 1Gb, I don't see it becoming standard. And without anything to push the prices down (like competition), I think the chipset makers will milk 1Gb and keep 2.5Gb a premium for as long as they can. I hope you are right about 3 years, but I think it will be more like 5 years unless something comes along to push it.
 
2.5/5GB switches are already starting to become more and more common place in the enterprise switching market. On average, the 48 port units cost about $1000 more then their 1G counterparts but I imagine this price delta will come down for the next 5 years. Given how you can run 2.5/5G on CAT6 over 100 meters without upgrading to CAT6A, they will become more "standard" in the future in a similar fashion to how 1G took the place of 10/100.

The vast majority laptops have already shed their ethernet ports. Given how most docking stations have switched to using TB3, there is room for 2.5/5 GB adapters in this application that we'll probably be seeing as costs for chips come down as well. Workstations are already using the 2.5GB spec with some even going up to 5GB. Eventually, marketing is going to take over with the old "bigger is better!" mantra we see across tech so I have no doubt mgig will become commonplace in the future. As time goes on and the offices built in the 2000's are upgraded/renovated, we'll see CAT6A become more widely utilized and go through this same exercise again in 10 years.
 
Why even bother with 2.5 / 5Gbps when a single NVMe PCIe drive can saturate a 10Gbps link already between 2 wired devices. I presume due to most "average" users using Wifi and most not moving around a lot of large files with in their LAN..
 
2.5/5GB switches are already starting to become more and more common place in the enterprise switching market. On average, the 48 port units cost about $1000 more then their 1G counterparts but I imagine this price delta will come down for the next 5 years. Given how you can run 2.5/5G on CAT6 over 100 meters without upgrading to CAT6A, they will become more "standard" in the future in a similar fashion to how 1G took the place of 10/100.

The vast majority laptops have already shed their ethernet ports. Given how most docking stations have switched to using TB3, there is room for 2.5/5 GB adapters in this application that we'll probably be seeing as costs for chips come down as well. Workstations are already using the 2.5GB spec with some even going up to 5GB. Eventually, marketing is going to take over with the old "bigger is better!" mantra we see across tech so I have no doubt mgig will become commonplace in the future. As time goes on and the offices built in the 2000's are upgraded/renovated, we'll see CAT6A become more widely utilized and go through this same exercise again in 10 years.
The problem I see for 2.5/5 is the competition from used 10Gb. With 10Gb now being established and working for so many generations, the older stuff is dirt cheap especially when using dac cables. I remember someone on reddit posting a complete 10Gb 'kit' consisting of 2x 10Gb cards and a dac for $60 shipped. That's just the price of 2x 2.5Gb nics right now.

Laptops predominantly don't need wired connections, so they're going to be focused on whatever magical wifi numbers there are out there. But the option for 2.5Gb will make sense for some scenarios as usb dongles are already available at a decent price.
 
Why even bother with 2.5 / 5Gbps when a single NVMe PCIe drive can saturate a 10Gbps link already between 2 wired devices. I presume due to most "average" users using Wifi and most not moving around a lot of large files with in their LAN..
The most used scenario I see is where people have a wired connection to their nas and do move around a lot of files. Yeah, wifi won't really be using 2.5Gb until it can consistently break 1Gb, although I believe Ruckus and a few other companies have ap units with dual gigabit even now.
 
Ya, Wifi 6 "technically" they way they advertise it can do several Gbps, but as we know, that is never true and over head and everything else, but with Wifi pushing faster speeds and more homes getting 1Gbps access, i guess the market wants to milk it for as much as it can and do 1 -> 2.5/5 -> 10Gbps and so on and spread it out.

I am working to do 10Gbps from my firewall level down for my basement renos for exactly what you noted, my NAS and for my home lab to make use of the speeds I have with those.
 
I've been eyeing 10GBaseT and multi-gig switches for a while. I'll bite eventually, maybe sometime next year. I'm not sure what I'll go with though. I might go straight 10gig, or maybe I'll get a multi-gig 10/5/2.5 model. If I go straight 10gig I'll either need a 1gig switch with a 10gig uplink too, run parallel networks with the 10gig network not connected to the Internet, or make the Linux box in my basement play router along with serving files. The big thing I'm wondering is if 2.5 or 5 is going to take off for other stuff, particularly WiFi routers and access points. WiFi might not actually hit 1gig yet at least on a single transfer, but it will eventually and probably in not too long.
 
zandor SFP+ , uses FAR less power, less heat as well if you can. But, finding switches can be so so with enough ports, pending on your budget
 
zandor SFP+ , uses FAR less power, less heat as well if you can. But, finding switches can be so so with enough ports, pending on your budget
The messy bit with SFP+ is the cabling. Big fat direct connect short copper cables or fiber unless you plug BaseT modules into it, and that's just an expensive annoyance. What I'd really like is a switch with 3-4 SFP+ ports and a few more multi-gig 1/2.5/5/10g ports (maybe not all 4 speeds), VLAN support and a port that can do 1gig and talk to a cable modem/router, but I haven't seen anything like that. Usually it's either an SFP+ 10gig switch or you get 1-2 SFP+. Basically I'd like 3-4 SFP+ for screwing around with low latency network programming stuff, then the rest just for a fast home network. 10g is probably overkill for the home network, but I would like something better than 1gig. Maybe I'll end up with multiple switches. eBay used 10-gig and some sort of consumer grade multi-gig. The fast part would be a couple machines in the basement right under my office, then a cable dropped down from my desktop in my office. It's a short enough run for SFP+ copper, though that kind of screws a nice looking wall plate. It wouldn't be any more ghetto than the CX4 Infiniband line I had running from my office to the rack in my basement in my last place though. So 3 ports, but I wouldn't mind a fourth just in case I come up with a use for it. On the other hand, I can do my programming stuff without a switch. Just direct connect. One machine would need a dual port card or two cards, so presumably dual port. So I'm still looking and thinking. Well mostly just thinking. Holiday spending always turns me into a cheap bastard for a bit. Not close to broke but looking at the Christmas present credit card bill after I finish my shopping makes me want to not spend any more money for a bit.
 
Brocade ICX6450 on ebay, 4 x 10Gbps SFP+ and 24 / 48 1Gb ports. I think the switch market is behind for consumers on 2.5/5Gbps, while we are seeing boards coming out with it, seems home networking isnt there yet.

fiber and DAC cables are not that expensive anymore, sure, perhaps a little more than CAT6, but the power and heat you save, could make up for that. fs.com

https://forums.servethehome.com/ind...s-cheap-powerful-10gbe-40gbe-switching.21107/
 
What do you all think about something like the Ubiquiti US-XG-6POE? Four 10g-BaseT, 2 10g SFP, the RJ45 ports run at 1/2.5/5/10. Seems like a great way to play around, and I can stack a cheap 8-port 1gb switch on top for low bandwidth wired devices e.g. Hue hubs and the like.
 
Back
Top