Honest question, what do consumers use this for? Not even talking about need, but why would you want this?
Since 1Gb is normal, why would you need 2.5Gb? If 1Gb is insufficient, you'd probably want to upgrade to 10Gb. 1Gb is 125MB/s which isn't often being transferred inside a house and even if it was, would anyone care if the transfer took 2X as long unless it was a daily occurrence?
That's probably why it's taken so long for consumer grade NICs of this speed to show up, there isn't much need. There's definitely homelabbers and other enthusiasts who might find standard gigabit insufficient, but that's definitely a minority of users.
I suspect it's more of a middle grounds on cost. I could foresee someone who maybe has a home office who requires lots of bandwidth to a networked storage server for something like high definition video editing, or some other form of media editing that may require high bandwidth to reduce bandwidth. Sometimes those seconds here and there can add up.
It can also reduce the time required to do network backups/restores. Can literally shave off hours of time if it's something that you need to do frequently.
Not even middle ground, it basically wont be much more than Gigabit as a TCO, at this point the only issue is cheap Multigig home grade switches. There are billions of metres of Cat5 cable installed in the world, no doubt Realtek are going to be able to put these into motherboards for a couple of bucks, after the switches are commonplace, 2.5Gb just starts working across the world. 10Gb is something very very different.
10GBit needs new cables, 2.5GBit can just run on existing cables, which is a huge bonus. Additionally, one would hope such a card would be significantly cheaper then a 10GBit card.
2.5x speedup is still quite relevant, as that allows you to basically max out spinning discs, instead of ethernet being a limiting factor. So.. why wait for a transfer when you could be done earlier?
"You don't really need technology X" is a bad argument, it has held back advancement in so many areas. Its quite certainly feasible to assume that advancement in "needs" have been held back by commonly available technology. Ideas may have been shot down early because someone said "but ethernet is only 1GBit everywhere".
10GBase-T works fine over existing shorter (say <30 meters) Cat5E cabling in most cases. That should cover most homes if the patch panel isn't in some silly place.
I can substantiate that. I have two 10GBase-T wire runs: 1x20 meters and 1x35 meters total. The wire is CAT5E. Granted the plugs are CAT6 and receptacles are CAT6+ (purchased before CAT6A was a product). One is a server and the other is the only desktop that can really make use of that bandwidth. Tested from desktop ramdisk to server ramdisk and vice versa to remove bottlenecks. Seems to have no issue running full speed with no dropped packets.
Putting aside single event use cases, upgrading only the server/NAS to 10G can also make sense in some environments if there is frequent concurrent access to the server/NAS from multiple clients.
That is an extremely noisy EMI environment. In smaller labs and home networks I've not seen Cat5e fail to run 10GbE with no apparent bottlenecks or dropped packets in shorter runs (under 15 meters). Longer runs have generally been fine if you are talking reasonably ideal installations where its one cable not near others running to another machine or maybe a switch where you don't have a ton of stuff right there. Longest I've seen is 27m on Cat5e.
Of note I have yet to see a setup with Cat5e that has failed. I just haven't seen anything outside of those scenarios.
I have seen Cat6 at longer runs or denser environments and also not seen it fail.
Cat5e is rated to 45m for 10GbE and Cat6 is 55m. These are NOT rated for that in poor EMI environments. Cat6a is rated to 100m in challenging EMI environments (cable bundles and such forth).
For me personally, it's for transferring large raw photo files to and from my NAS. With 30MB raw files, that translates to 1-3 pictures/second for 1Gb speeds. That's just transferring and when you are editing, it is even slower.
The real question is, who gets to use this speed? I don't know of anybody that has 2.5Gb or higher switches at home. Sure, there are the few people that may have already purchased 10gb switches, but there aren't many because they are too expensive still. I haven't looked in a few months, but I haven't seen any 2.5Gb switch ports. Am I missing something?
For me personally, it's mainly the cost of the switches. NICs have been affordable for quite awhile now. However, I'm willing to pay about $10 per port for a 2.5G backbone for my network. I'm tired of my Ethernet connection being the weakest link for data transfers across the network.
Who gets to use it? People who use two cards in a point-to-point network. You may well only need it for one connection right away, and once you do get a switch the cards should still be useful.
It requires extra ports and/or extra cabling though to do that. I haven't wanted to deal with that so I haven't done it. Otherwise I am with Golgatha, other than the fact that I am fine if my core switch isn't 2.5GbE+. I don't mind stacking a small low port count switch on top of a core GbE switch. Ideally I'd love it all to be in one. But I am also not willing to pay more than about $150 or so for such a switch that has a low port count. Future proofing I wouldn't mind paying more than that for a larger switch to cover my whole network, but there I am still talking maybe no more than about $300, but that better be for a 24/26/28 port switch.
Yeah, the dearth of switches is a serious problem. Current consumer offerings are terrible, usually having only two 10GbE ports for an exorbitant price, and a handful of GbE in addition to that. Don't see the point of those at all. I'd gladly pay 2-3x the price of an 8-port GbE switch for an 8-port 2.5GbE switch, or even better one with a couple of 5/10GbE ports.
Dearth of switches? I literally saw a Netgear multigig switch (1/2.5/5/10) on sale on Amazon just yesterday for under $200. I have a home 10gbase-t network and even my switch (Netgear XS728T which is for a business) is under $1000 now. Even Asus sells a 10GBase-T switch for $100 these days... There's tons of 10gig and multi gig switches available that are affordable so I'm guessing you made your comment without doing some research first
BTW, I have symmetrical Gigabit internet (980Mbps down/ 975Mbps up) and for only $70/month, along with a 250TB RAID 50 Server/NAS running nurmerous Virtualized machines to several Thin clients around the house with three storage tiers (with the top tier being 4x HP EX950 m.2 nvme SSDs on a Highpoint Solutions SSD7103 pcie 3.0x16 RAID adapter that reaches close to 16GB/sec) that currently uses an Intel X710-Tr4 four port 10GBase-T NIC with all four ports aggregated together for a 40Gbps backbone. So yes, I need 10GBase-T networking
Netgear XS728T - Just checked amazon and this is going for $1889.90. As this switch very much matches my criteria, I'd love to know where you got it for <$1000 new.
Asus XG-U2008 (got the model from your downstream post) - $249.99 at Amazon. Only has 2x10G ports. This is frustrating as 3x10G ports would have allowed me a server, workstation, and uplink port, but only two ports requires a sacrifice I'd rather not make. Also reliability is apparently a concern.
Netgear MS510TX (first multigig switch I found) - This goes for $269 at amazon. It has 4x1G, 2x2.5G, 2x5G, 1x10GBASE-T (Copper), and 1x10GBASE-X SFP+ (fibre). This could work for some environments, but I have to wonder if it really makes sense to have so many different max port standards when all the multigig ports could have been 10G (given that it also supports 5g and 2.5G). At the very least I don't understand why they didn't make the 2.5G ports 5G given the 2.5G compatibility and marginal price difference.
It would seem prices have moved back up since you posted yesterday. Beyond that, the cheaper models have a number of quirks and trade-offs that limit their use. That said, it does appear that their are more options available than their used to be as long as you can deal with some of the trade-offs. Even if it is slow, it is good to see the market moving again.
Nope. Your aren't missing anything. The real issue are switches and most people that run >=2.5 usually use old enterprise gear that is relatively cheap to get of course with the "cost" of being huge and power hungry. I don't want a 20 port switch that's bigger than my PC...when 1gb consumer switches are tiny and fit anywhere. yeah real problem are switches and it's easy to see why: because mosz home users prefer wireless anyway and that is where consumer companies are investing in. Much more profitable to sell you new wifi routers every other year than one 5 gbps router you will keep for 2 decades.
Power hungry? You do know the average cost of a kWHr in America is just $0.125, right? I have a home 10gbase-t network and even my switch, a Netgear XS728T, which has 24 10GBase-T ports and four SFP+ ports is not big in the least at 17.3"x12. 3"x1.7", it's about the size of a cable box/direcTV receiver. Asus sells a 10GBase-T switch (the XG-U2008) that's much smaller at 9.45"x4.92"1. 06" which is small by anyone's definition. There are numerous 10GBase-T and multigig switches (that do 1/2.5/5/10 gig) on the market with numerous options under $300. With 4k and 8k streaming around the corner, everyone will need one, especially with families that have multiple streaming users at once
Some of us like to try to be friendly to the environment and power usage is a concern still.
That Netgear XS728T listed max power consumption is 134w. I assume actual use power is lower and if you've got a lot of ports in GbE mode and not 10GbE, etc. I could be reaching, but I am still guessing with just a couple of ports at 10GbE and the rest at 1GbE, you are probably still running a power budget of 30-40w utilizing a dozen or so ports active in GbE and 2 in 10GbE. That would probably be my use case with other ports connected, but not often active (my port count for my house when I am done wiring it is going to be 23 ports, including one fiber to my shed 170ft from my house where I don't want to bury Ethernet, but bury fiber).
My current 16 port GbE semi-L2 switch is listed as 15w max power consumption and uses 8w under typical use with about 8 or 9 ports typically active and all connected. I am swapping it out with same model, but 24 port as my needs have gone up with my new house. That one is listed as I think 20w max power consumption. I suspect I'll be up to around 10w or so in typical use case at a guess.
Anyway, the point is, switching to the Netgear XS728T and I'd likely be burning an extra $20-30 a year in electrical use under my guessed use case. If the actual power consumption really is closer to 130w even with most of the ports either inactive or in 100MbE/1GbE mode and just a couple in 10GbE mode...then its more like $100 more a year.
That isn't an expense just to shrug at. Over 5 years that is $500+ (typical power costs in my area are about 14 cents per kwh delivered, 9 or so for generation, about 5 in transmission fees, works out to almost exactly to 1 watt of power consumption, 24/7/265 being $1 per year).
It is certainly something to keep in mind. Heck, my whole server only uses 26w at idle and about 60w when running full out. I don't need a switch that consumes 1-2x as much power as my server.
Being "green" if you will, I absolutely look at the power consumption of networking devices. I had my eye on another switch instead of the TP-Link SG2424 I am getting (I have an SG2216 right now) because the power consumption was slightly lower, but I don't know how much real use power consumption would be lower. Looking at the Dlink 26 port model. It is listed at 15w max. But I did do the math and was able to find a used SG2424 for ~$70 or so shipped and the Dlink I can't find used (not the current version) at around $170. $100 over 5 years is about 20w higher power consumption for break even...
Wasn't worth the cost difference as I usually plan around 5 year equipment life space (either because it actually lasts that long, or that is how long I run it till I replace it).
I do want more performance, but I don't want to pay $500 over 5 years, plus the cost of the equipment, which is >>>$1000 on top of that. For me, I am looking at more like <$200 for a <8 ports of 2.5GbE or faster (it NEEDS to support 2.5GbE speeds as an option as some networking equipment like routers are starting to ship with 1/2.5GbE ports. So if it is 1/10GbE only, that will not work for me and some of the Netgear switches that support 10GbE on a port, do not support 2.5GbE on said ports). I'd be willing to pay more if it was a 24/26/28 port model where all ports were 2.5GbE or faster, but again, maybe $300 or so is my network equipment budget max (also taking in to account having to shell out another $60-160 for 2 NICs for my desktop and my server). And then on top of that, maybe another $100 in power budget over 5 years. So at most an extra 20w of power consumption for the stand alone switch in real use cases. Or else about 30 or so watts if a larger switch and I could replace my core switch with it entirely.
The power use I'd be willing to be slightly more flexible if the cost of the switch was lower. However, I don't need the faster speeds so much more that I care to add a huge power suck to my house all the time either in cost or environmental impact.
So it is a multifaceted problem for me. For now, nothing is approaching the <$200 for an 8 port switch and ~20w or less power consumption, or <$300 for a 24/26/28 (I'd even say <$300 for a 16 port as I could always add on a small 8 port very low power consumption GbE switch to get my network to have sufficient ports) and no more than 30w of power consumption. Again on both sets of numbers I'd be willing to be a little flexible, but when the MS510 is right now on sale for $269 and gets you 1 10GbE, 2x5GbE and 2x2.5GbE with 5 1GbE (26w) the cost and power consumption are just a bit more than I'd be willing to pay.
It is getting close, but emphasis on GETTING close. Don't know when it'll happen, but seems more likely it is when the 2nd (3rd?) tier guys like DLink, Trendnet and TP-Link start offering multigig switches is when we will finally at least see cost come down to what is reasonable for me to invest in it. Not sure if power consumption will have, or if it'll be yet another generation of equipment before we are there.
Apparently some of the new synology's will have 2.5Gb ports. I believe some new home routers from asus and netgear have them. So if you have a direct connection from your computer to the router, then that would act as the switch. But unfortunately that is very limited, as most will have a switch in between computers and routers.
Not a wide use case, but it's out there. I have GigE everywhere, and it's generally fine, because I am the only truly computer savvy person in my house. But if others were doing what I do, dumping large files like HD movies and ISOs to my server, I'd most definitely want something faster than 1Gb on my server. I don;t see a need to make any workstations >1Gb yet, even the 1Gb in the server is fine for multiple HD video streams to multiple TVs at the same time. Once we get well beyond 1Gb internet speeds - 1Gb Ethernet will be the bottleneck, and we'll need something faster. I just got another speed increase with no bill increase, but that's now 300Mb, so no worries with 'only' 1Gb networking.
My new 802.11ax WiFi router comes with one 2.5 Gb port. I have a Plex server with all my DVDs and Blu-ray discs ripped to it, as well as all my CDs, and a TV tuner to record live TV. There have been times when multiple people in the house have streamed from the server, and it has skipped frames. I wouldn't mind adding a 2.5G port to the server, which should stop that.
Some of us are on Fiber ISP's that offer connections faster than 1Gbps. And don't need the expense of going with 10Gbps equipment or the noise. And some people time is money so yes the 2x longer transfers can't be an issue depending on what you are doing.
It has a lot to do with existing infrastructure. Getting even CAT6 to handshake at 10Gbps is pretty tough, most installs need CAT6A if going beyond 30 feet or so. And jacks, couplers, etc, all further reduce distance.
Many of the 10Gbps installations of done only negotiate at 5Gbps due to the infrastructure, and that's usually good enough for the client who is just trying to load balance a server across a gigabit network.
2.5Gbps in a home would be desirable for anybody trying to transfer\decompress files from a torrent server to their HTPC, but that, like many beyond-gigabit applications, is pretty niche. But that doesn't mean somebody doesn't want it.
Remember, the industry has to maintain a push for interfaces that meet future demands, not current ones. The way things are going there will be a need for 2.5Gb in homes within years. My gigabit wiring is actually the bottleneck of my $70/mo AT&T Fiber internet connection (which pulls a little over 1Gbps up/down.)
Where I live, New Zealand, they are testing consumer 10Gbit connections (symmetric?) and 1Gbit is common. So I could see the ISPs giving out a 1 10Gbit + 4 2.5Gbit switch so you can use the bandwidth when they start rolling it out. As for what you can use it for? Personally faster updates/downloads for games so when I turn my PC on to a several GB update I don't have to wait as long to join my friends in game. I can already pull close to 100MB/s through steam depending on the game. With 10Gbit I could make my SSD the bottleneck! O_O (assuming the servers can keep up)
I like to take photos and edit them in Lightroom, but I only have SSDs in my main PC and use a (DIY) NAS for mass storage. Browsing through a library of 50-70MB RAW photos over GbE is quite annoying. Even moving to 2.5G would mean a massive improvement in library responsiveness for this use case. As such, cards like this would save me the hassle of keeping "in-use" photos stored locally on my SSD and then archiving them on the NAS after editing, which causes quite a lot of work with library management.
Why should you ride a bike, when you can walk? Why drive a Mercedes when you can go per pedes? Why fly when you can swim?
Gbit Ethernet was ok, as long as your storage was a single HDD: Balance!
But with SSDs or even RAIDs somewhere in your network, Gbit grew from an irritation into an intolerable bottleneck.
Where the transition from 100 to 1000 Mbit was pretty smooth, the next step went all wrong. Yet NBase-T with the 2.5 and 5 Gbit intermediate steps IMHO have finally healed that.
From experience I can tell you that 10Gbit is a bit like driving a car at 400km/h instead of 100km/h: The linear comes with exponential stress and constraints, especially downtown, with small files and synchronization.
2.5Gbit is more like 150km/h and 5Gbit like 200km/h, each a much easier job, at least here in Germany.
And when I need to go from one major metropolitan area to another here, I don't take a plane (as you would in the US), but I take my car: Going 200km/h instead of 80km/h means being back home for lunch or dinner: A worthy premium.
And it's a similar story for copying VMs updating larger Docker containers; 3x, 5x and 10x speedups are significant enough to appreciate.
When you aggregate server class hardware, 10Gbit is rock bottom, 100Gbit is becoming mid-range. But it also takes that type of iron to take advantage of it and there you tolerate the noise and power it requires.
NBase-T gives you sweet spots that even work on laptops or ordinay desktops: I have upgraded my workstations to 10Gbit low-power Aquantia 10Gbase-T two years ago and more recently put 2.5 or 5 Gbit USB adapters even into the lowlier notebooks, because it is a nice match to their SSDs. Ultrabooks don't even come with Gbit Ethernet any more, so NBase-T USB Ethernet is a natural complement to their NVMe storage.
Now that Netgear and Buffalo NBase-T switches have equal port cost as these NICs (~€30-€80), it becomes a pretty simple cost/benefit calculation with linear price points.
Beyond-Gbit is mostly about LAN connectivity today: If Internet is your only network, you may not need it for a while longer. If most of your traffic is East-West, the more the better.
1) Price, 2.5G is still plenty fast and costs much less then 5 or 10 2) Wide Compatibility, 2.5G and 5G only use PCIe x1, 10G uses PCIe x4 3) Hardware price: Routers, NAS etc with 2.5g and 5 G cost MUCH less then 10G hardware, there are plenty Switches that have 2-3 2.5G or 5G port and only single 10G 4) Main usage: Home NAS, Home Server connection, using PC in a closet/basement and feeling like you using it directly, 1G is not enough for this, especially if you use 4K resolution. If you have bunch of people using home network for gaming, work, entertainment then 2.5G still offers twice the band witch so 2-3 people can use NAS, stream 4K at the same time. 5) Its the next step, 1G should be forgotten ASAP and 2.5g will be the new minimum, most new motherboards now come with Realtek 2.5g since its cheap. 6) This cheap 2.5g Chip opens up an opportunity to release new home use routers, switches etc for same price as we paid for 1G just with 2.5g its that cheap, Aquantilla 5G and 10G has nice premium, but this 2.5G realtek and soon to come Intel 2.5G mainstream nic are dirt cheap
For transfers where even a spinning harddisk is limited by a 1 gigabit speed, but not by 2.5 Gb. Then there is also the price of switches, which is the real limiting factor. For now, I have to do with 1 GB for my home network, which means transfers for me maxes out at about 113 megabytes/s (which means that its about 0,9 Gigabit/s with the overhead in the switcher/network card that limits the speed by about 10%)
Upgrading to 10 Gbit may require upgrading the cabling that is run through walls etc. and that can be a very expensive task for long runs. This is why 2.5 Gbit has seen major interest as the common Cat 5E cabling can be re-used for the higher speeds. The 2.5 Gbit only widens the symbol library vs. 1 Gbit as the line rate and encoding scheme are the same, hence they can leverage the same cabling.
If you're cabling is mainly in-room, jumping to 10 Gbit is indeed mainly on the adapters and switch as the short patch cabling cost differences are minimal.
I mean, CAN I wait longer for transfers? Sure. However, its been annoying as heck since I moved I am down to a single GbE link between my desktop and server. I finally starting to get to wiring my new house rather than just hanging a few temp premade (and LONG) cables through my basement and up through the floor in the couple spots I needed wired access.
Anyway, I had 2xGbE previously. With SMB Multichannel that got me ~235MiB/sec transfers and down to around ~115MiB/sec now (that single link is spread on a 5 port switch to a couple of other devices right now). I am occasionally tossing big files around my network. No one will die if I have to wait 60 seconds instead of 30 seconds for a large file to transfer.
But again, it is still annoying. The every 10-16 month "opps I messed something up and now I need to transfer everything back over" data "[bring] backup" I have to do IS annoying as heck and also longer that I only have 1 good copy.
I keep everything mirrored between my desktop and server with a twice weekly backup job. I backup to an external drive when I remember. Which is usually about 4-6 times a year. So I would never lose everything, but I only have 1 most current copy, a second nearly current copy and one stale copy of my data. For anything important I make sure I don't remove the files from the source media till there are multiple copies (example, pictures from a trip off an SD card).
But if I am backing up, at 115MiB/sec that's about 7.5hrs. With dual links it takes about 4hrs (some of the smaller files it is disk based limitations and not link based).
Going to 2.5GbE would both allow me to be slightly faster than dual GbE links, but it would also allow me to get that performance with just a single network cable. Now, I am laying a crap ton in my house as I am doing the permanent wiring, but it still gives me more flexibility and performance (dual 2.5GbE!).
I'd still have disk based limits, but going to ONE 2.5GbE link would probably take my transfer times down to 3.5hrs from 4hrs at a guess. Dual 2.5GbE links would probably only squeeze maybe another 10-20 minutes faster in there as my RAID array likely can't exceed the performance of a 2.5GbE link except on the outer most tracks (about 320MiB/sec max read/write on the outer tracks with the 2x3TiB 7200rpm disks in the array, but by the time you are 20% utilized, you are down to more like 280MiB/sec max).
But it also leaves breathing room for replacing the disk arrays with either new HDDs at some point which likely would be 15-25% faster than my older 3TiB disks, or possibly some big TLC SSDs if prices come down somewhat more.
A low cost switch, especially one that is lower power is also crucial. As mentioned in the article, basically everything has at least a couple of 10GbE ports on it if it has slower ones. Even the little 10 port multigig switches I've seen are running around 30w and actively cooled. Probably in part because of those 10GbE ports. If there were some 8 and 16 port 2.5GbE switches (and maybe the same for 5GbE) that would be real nice. If they CAN get port price down to around $10 a port for 2.5GbE switches and $20 for 5GbE, I think you'd see pretty wide spread adoption.
Even if all they could do is $20 for 2.5GbE and $30 for 5GbE I think you'd still see a lot more adoption. Especially with those guys who need just a smaller switch for their 2-4 ports on their network they need as fast as possible and can then just connect it to a larger switch to handle everything else.
In my case, I see it being useful to connect my server to my desktop (minimum 2 ports there, ideally 4) and possibly 802.11ax router/APs later, which in my case is 2 total. An uplink port at 2.5GbE speed, or dual ports running at 1GbE to a larger switch is also needed. Then a main switch running all GbE ports (unless there is a single 2.5GbE uplink port) would be fine.
The vast majority of users don't even require Gigabit-Ethernet. Typical home networks don't actually transmit data between clients; by far the largest group of users only transmit data between their home-router and the clients, and by far most people have below 100 Mbit/s (the average in western countries is below 50 Mbit/s).
Then, there is the very small group actually using communication between clients in the network, like those who have NAS-boxes or other local network appliances. Those might have modern WiFi equipment supporting over 1.5 Gbit/s in MU-MIMO.
Those are limited by 1 GbE, but won't come close to requiring 5, let alone 10 GbE. 2.5 GbE is the perfect sweetspot and costs only below 50$ to upgrade the home-server that serves the multiple WiFi users.
5 or 10 GbE not only is at least twice as expensive, it very likely also requires new cables adding additional cost, for no benefit over 2.5 GbE whatsoever.
Both my Internet connection and my home file server are capable of saturating my gigabit LAN connection to my desktop, and my file server could probably saturate a 2.5 gigabit connection. A bit of a speed boost for file copies would be nice, or maybe reducing the impact of file access on Internet speeds.
It's not critical, by any means, gigabit is plenty fast, but it'd be nice.
NAS connectivity. 1Gb tops out at 125MB/s, which even a single WD black can saturate today. But 2.5 Gb would allow 250 MB/s, possibly 300MB/s, which is what most home 4 bay RAID NAS units can hypothetically hit.
Gamers would love something like this, with games hitting 150GB now, there needs to be a faster way to back all that junk up so it doest need to be redownloaded later.
10Gb would be even better, but costs. It's at least 4x as expensive to upgrade to a 10g ethernet card, and you need x4 slot, not a spare x1. More importantly, motherboards can put a 2.5Gb NIC onboard without major changes, not possible with the more demanding 10Gb, especially in Mini ITX format. 10Gb NAS units are also significantly more expensive then 2.5Gb NAS units.
I purchased a 10Gbit NIC based on the Aquantia stuff for $99 a while back. Works pretty well. Nothing affordable utilizes 10gbit though. They need to work on routers, cable modems, etc. My particular NIC also supports the 2.5 Gbit and 5 Gbit modes. I'd love to pick up a router that supports these modes along with getting some cable modem support.
The last paragraph makes little sense. This card works in any PCIe slot, just like any other PCIe card would. PCIe is backward and forward compatible by definition. The only limitation is the speed of the slot and even a PCIe 1.0 x1 slot would support the 2.5 Gbps this card needs (maybe ever so slightly below optimal speeds because of overhead). A 3.0 (or 4.0) x1 card would do that just as well.
Yeah, unclear. PCIe 2.0 x1 is exactly what this card needs. PCIe 1.0 is 2Gbps after encoding overhead.
I also don't get the statement about the price:
"not a particularly low price considering the fact that GbE ports are ‘free’ with most motherboards and Aquantia-based 5 GbE and 10 GbE cards can be bought for $70 and $90"
2.5Gbps ports are hardly common on motherboards, and $35 is half or less the price of 5-10Gbps cards.
Working at RealTek I can tell you this is the typical bottoms up approach to market share acquisition. The goal is to get the lowest common denominator at no cost adder late market advantage. Intel is introducing 2.5Gbe as a standard speed in upcoming products and most WiFi extenders are requiring 2.5Gbe as well as not needing to changing cables.
10Gbe/5/2.5/1 N-Base-T solutions came from the top down infrastructure equipment vendors like Aquantia and now merged with Marvell. Those solutions require 10.3125 Gbps SERDES so for power and area this is a cost concern as well.
The good news is that 2.5Gbe will be nearly free in upcoming designs over the next year or two for consumers. If you want higher speed other vendors already supply 10Gbps solutions like the AQR113 but draw typically 3Watts of power just for the PHY.
Why does anandtech keep saying no cheep 2.5GbE switches out? They have been for years cheap enough out. Are they copy and pasting info from years past or something?
Sure not cheap if you go 8+ ports, but plenty around $200 mark for less ports.
Honestly, I think this is a quite fairly priced product. 1 or 2 GbE ports might be free on boards, but an additional card certainly isn't. And plenty of people still buy such a card for a second port apart from their onboard GbE.
Also, while it isn't 10 GbE, it still is 150% faster than GbE, for a very affordable price. Also, the 2.5 GbE are available over longer distances and poorer cables; it is specified for 100 meters over Cat 5e which is still _very_ common. 10 GbE on the other hand requires Cat 6 and only specified up to 60 meters - and is much more heavily influenced by poor plugs, panels and workmanship.
10 GbE also heavily taxes systems, while 2.5 GbE is reasonable on most hardware. This means that for many, it is a massive performance improvement for little overall cost and few problems.
It also is quite sufficient to attach modern high-performance WiFi equipment which is hampered by GbE but far away from requiring 5 or even 10 GbE.
This makes it a perfectly reasonable upgrade for a very reasonable price.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
53 Comments
Back to Article
webdoctors - Thursday, October 24, 2019 - link
Honest question, what do consumers use this for? Not even talking about need, but why would you want this?Since 1Gb is normal, why would you need 2.5Gb? If 1Gb is insufficient, you'd probably want to upgrade to 10Gb. 1Gb is 125MB/s which isn't often being transferred inside a house and even if it was, would anyone care if the transfer took 2X as long unless it was a daily occurrence?
blackice85 - Thursday, October 24, 2019 - link
That's probably why it's taken so long for consumer grade NICs of this speed to show up, there isn't much need. There's definitely homelabbers and other enthusiasts who might find standard gigabit insufficient, but that's definitely a minority of users.inighthawki - Thursday, October 24, 2019 - link
I suspect it's more of a middle grounds on cost. I could foresee someone who maybe has a home office who requires lots of bandwidth to a networked storage server for something like high definition video editing, or some other form of media editing that may require high bandwidth to reduce bandwidth. Sometimes those seconds here and there can add up.It can also reduce the time required to do network backups/restores. Can literally shave off hours of time if it's something that you need to do frequently.
danielfranklin - Tuesday, October 29, 2019 - link
Not even middle ground, it basically wont be much more than Gigabit as a TCO, at this point the only issue is cheap Multigig home grade switches.There are billions of metres of Cat5 cable installed in the world, no doubt Realtek are going to be able to put these into motherboards for a couple of bucks, after the switches are commonplace, 2.5Gb just starts working across the world.
10Gb is something very very different.
nevcairiel - Thursday, October 24, 2019 - link
10GBit needs new cables, 2.5GBit can just run on existing cables, which is a huge bonus. Additionally, one would hope such a card would be significantly cheaper then a 10GBit card.2.5x speedup is still quite relevant, as that allows you to basically max out spinning discs, instead of ethernet being a limiting factor. So.. why wait for a transfer when you could be done earlier?
"You don't really need technology X" is a bad argument, it has held back advancement in so many areas. Its quite certainly feasible to assume that advancement in "needs" have been held back by commonly available technology. Ideas may have been shot down early because someone said "but ethernet is only 1GBit everywhere".
monsted - Thursday, October 24, 2019 - link
10GBase-T works fine over existing shorter (say <30 meters) Cat5E cabling in most cases. That should cover most homes if the patch panel isn't in some silly place.BurntMyBacon - Monday, October 28, 2019 - link
I can substantiate that. I have two 10GBase-T wire runs: 1x20 meters and 1x35 meters total. The wire is CAT5E. Granted the plugs are CAT6 and receptacles are CAT6+ (purchased before CAT6A was a product). One is a server and the other is the only desktop that can really make use of that bandwidth. Tested from desktop ramdisk to server ramdisk and vice versa to remove bottlenecks. Seems to have no issue running full speed with no dropped packets.Putting aside single event use cases, upgrading only the server/NAS to 10G can also make sense in some environments if there is frequent concurrent access to the server/NAS from multiple clients.
Dug - Monday, October 28, 2019 - link
I found that not to be true in our server room.azazel1024 - Tuesday, October 29, 2019 - link
Server room.That is an extremely noisy EMI environment. In smaller labs and home networks I've not seen Cat5e fail to run 10GbE with no apparent bottlenecks or dropped packets in shorter runs (under 15 meters). Longer runs have generally been fine if you are talking reasonably ideal installations where its one cable not near others running to another machine or maybe a switch where you don't have a ton of stuff right there. Longest I've seen is 27m on Cat5e.
Of note I have yet to see a setup with Cat5e that has failed. I just haven't seen anything outside of those scenarios.
I have seen Cat6 at longer runs or denser environments and also not seen it fail.
Cat5e is rated to 45m for 10GbE and Cat6 is 55m. These are NOT rated for that in poor EMI environments. Cat6a is rated to 100m in challenging EMI environments (cable bundles and such forth).
PedalMonk - Thursday, October 24, 2019 - link
For me personally, it's for transferring large raw photo files to and from my NAS. With 30MB raw files, that translates to 1-3 pictures/second for 1Gb speeds. That's just transferring and when you are editing, it is even slower.The real question is, who gets to use this speed? I don't know of anybody that has 2.5Gb or higher switches at home. Sure, there are the few people that may have already purchased 10gb switches, but there aren't many because they are too expensive still. I haven't looked in a few months, but I haven't seen any 2.5Gb switch ports. Am I missing something?
Golgatha777 - Thursday, October 24, 2019 - link
For me personally, it's mainly the cost of the switches. NICs have been affordable for quite awhile now. However, I'm willing to pay about $10 per port for a 2.5G backbone for my network. I'm tired of my Ethernet connection being the weakest link for data transfers across the network.GreenReaper - Thursday, October 24, 2019 - link
Who gets to use it? People who use two cards in a point-to-point network. You may well only need it for one connection right away, and once you do get a switch the cards should still be useful.azazel1024 - Tuesday, October 29, 2019 - link
It requires extra ports and/or extra cabling though to do that. I haven't wanted to deal with that so I haven't done it. Otherwise I am with Golgatha, other than the fact that I am fine if my core switch isn't 2.5GbE+. I don't mind stacking a small low port count switch on top of a core GbE switch. Ideally I'd love it all to be in one. But I am also not willing to pay more than about $150 or so for such a switch that has a low port count. Future proofing I wouldn't mind paying more than that for a larger switch to cover my whole network, but there I am still talking maybe no more than about $300, but that better be for a 24/26/28 port switch.Valantar - Friday, October 25, 2019 - link
Yeah, the dearth of switches is a serious problem. Current consumer offerings are terrible, usually having only two 10GbE ports for an exorbitant price, and a handful of GbE in addition to that. Don't see the point of those at all. I'd gladly pay 2-3x the price of an 8-port GbE switch for an 8-port 2.5GbE switch, or even better one with a couple of 5/10GbE ports.AnarchoPrimitiv - Sunday, October 27, 2019 - link
Dearth of switches? I literally saw a Netgear multigig switch (1/2.5/5/10) on sale on Amazon just yesterday for under $200. I have a home 10gbase-t network and even my switch (Netgear XS728T which is for a business) is under $1000 now. Even Asus sells a 10GBase-T switch for $100 these days... There's tons of 10gig and multi gig switches available that are affordable so I'm guessing you made your comment without doing some research firstAnarchoPrimitiv - Sunday, October 27, 2019 - link
BTW, I have symmetrical Gigabit internet (980Mbps down/ 975Mbps up) and for only $70/month, along with a 250TB RAID 50 Server/NAS running nurmerous Virtualized machines to several Thin clients around the house with three storage tiers (with the top tier being 4x HP EX950 m.2 nvme SSDs on a Highpoint Solutions SSD7103 pcie 3.0x16 RAID adapter that reaches close to 16GB/sec) that currently uses an Intel X710-Tr4 four port 10GBase-T NIC with all four ports aggregated together for a 40Gbps backbone. So yes, I need 10GBase-T networkingBurntMyBacon - Monday, October 28, 2019 - link
Netgear XS728T - Just checked amazon and this is going for $1889.90. As this switch very much matches my criteria, I'd love to know where you got it for <$1000 new.Asus XG-U2008 (got the model from your downstream post) - $249.99 at Amazon. Only has 2x10G ports. This is frustrating as 3x10G ports would have allowed me a server, workstation, and uplink port, but only two ports requires a sacrifice I'd rather not make. Also reliability is apparently a concern.
Netgear MS510TX (first multigig switch I found) - This goes for $269 at amazon. It has 4x1G, 2x2.5G, 2x5G, 1x10GBASE-T (Copper), and 1x10GBASE-X SFP+ (fibre). This could work for some environments, but I have to wonder if it really makes sense to have so many different max port standards when all the multigig ports could have been 10G (given that it also supports 5g and 2.5G). At the very least I don't understand why they didn't make the 2.5G ports 5G given the 2.5G compatibility and marginal price difference.
It would seem prices have moved back up since you posted yesterday. Beyond that, the cheaper models have a number of quirks and trade-offs that limit their use. That said, it does appear that their are more options available than their used to be as long as you can deal with some of the trade-offs. Even if it is slow, it is good to see the market moving again.
https://www.amazon.com/NETGEAR-28-Port-Ethernet-Li...
https://www.amazon.com/ASUS-XG-U2008-Unmanaged-2-P...
https://www.amazon.com/NETGEAR-Multi-Gigabit-Ether...
Dug - Monday, October 28, 2019 - link
You can still use xs708 for 8 port $550xs716 16 port $900
beginner99 - Friday, October 25, 2019 - link
Nope. Your aren't missing anything. The real issue are switches and most people that run >=2.5 usually use old enterprise gear that is relatively cheap to get of course with the "cost" of being huge and power hungry. I don't want a 20 port switch that's bigger than my PC...when 1gb consumer switches are tiny and fit anywhere. yeah real problem are switches and it's easy to see why: because mosz home users prefer wireless anyway and that is where consumer companies are investing in. Much more profitable to sell you new wifi routers every other year than one 5 gbps router you will keep for 2 decades.AnarchoPrimitiv - Sunday, October 27, 2019 - link
Power hungry? You do know the average cost of a kWHr in America is just $0.125, right? I have a home 10gbase-t network and even my switch, a Netgear XS728T, which has 24 10GBase-T ports and four SFP+ ports is not big in the least at 17.3"x12. 3"x1.7", it's about the size of a cable box/direcTV receiver. Asus sells a 10GBase-T switch (the XG-U2008) that's much smaller at 9.45"x4.92"1. 06" which is small by anyone's definition. There are numerous 10GBase-T and multigig switches (that do 1/2.5/5/10 gig) on the market with numerous options under $300. With 4k and 8k streaming around the corner, everyone will need one, especially with families that have multiple streaming users at onceazazel1024 - Tuesday, October 29, 2019 - link
Some of us like to try to be friendly to the environment and power usage is a concern still.That Netgear XS728T listed max power consumption is 134w. I assume actual use power is lower and if you've got a lot of ports in GbE mode and not 10GbE, etc. I could be reaching, but I am still guessing with just a couple of ports at 10GbE and the rest at 1GbE, you are probably still running a power budget of 30-40w utilizing a dozen or so ports active in GbE and 2 in 10GbE. That would probably be my use case with other ports connected, but not often active (my port count for my house when I am done wiring it is going to be 23 ports, including one fiber to my shed 170ft from my house where I don't want to bury Ethernet, but bury fiber).
My current 16 port GbE semi-L2 switch is listed as 15w max power consumption and uses 8w under typical use with about 8 or 9 ports typically active and all connected. I am swapping it out with same model, but 24 port as my needs have gone up with my new house. That one is listed as I think 20w max power consumption. I suspect I'll be up to around 10w or so in typical use case at a guess.
Anyway, the point is, switching to the Netgear XS728T and I'd likely be burning an extra $20-30 a year in electrical use under my guessed use case. If the actual power consumption really is closer to 130w even with most of the ports either inactive or in 100MbE/1GbE mode and just a couple in 10GbE mode...then its more like $100 more a year.
That isn't an expense just to shrug at. Over 5 years that is $500+ (typical power costs in my area are about 14 cents per kwh delivered, 9 or so for generation, about 5 in transmission fees, works out to almost exactly to 1 watt of power consumption, 24/7/265 being $1 per year).
It is certainly something to keep in mind. Heck, my whole server only uses 26w at idle and about 60w when running full out. I don't need a switch that consumes 1-2x as much power as my server.
Being "green" if you will, I absolutely look at the power consumption of networking devices. I had my eye on another switch instead of the TP-Link SG2424 I am getting (I have an SG2216 right now) because the power consumption was slightly lower, but I don't know how much real use power consumption would be lower. Looking at the Dlink 26 port model. It is listed at 15w max. But I did do the math and was able to find a used SG2424 for ~$70 or so shipped and the Dlink I can't find used (not the current version) at around $170. $100 over 5 years is about 20w higher power consumption for break even...
Wasn't worth the cost difference as I usually plan around 5 year equipment life space (either because it actually lasts that long, or that is how long I run it till I replace it).
I do want more performance, but I don't want to pay $500 over 5 years, plus the cost of the equipment, which is >>>$1000 on top of that. For me, I am looking at more like <$200 for a <8 ports of 2.5GbE or faster (it NEEDS to support 2.5GbE speeds as an option as some networking equipment like routers are starting to ship with 1/2.5GbE ports. So if it is 1/10GbE only, that will not work for me and some of the Netgear switches that support 10GbE on a port, do not support 2.5GbE on said ports). I'd be willing to pay more if it was a 24/26/28 port model where all ports were 2.5GbE or faster, but again, maybe $300 or so is my network equipment budget max (also taking in to account having to shell out another $60-160 for 2 NICs for my desktop and my server). And then on top of that, maybe another $100 in power budget over 5 years. So at most an extra 20w of power consumption for the stand alone switch in real use cases. Or else about 30 or so watts if a larger switch and I could replace my core switch with it entirely.
The power use I'd be willing to be slightly more flexible if the cost of the switch was lower. However, I don't need the faster speeds so much more that I care to add a huge power suck to my house all the time either in cost or environmental impact.
So it is a multifaceted problem for me. For now, nothing is approaching the <$200 for an 8 port switch and ~20w or less power consumption, or <$300 for a 24/26/28 (I'd even say <$300 for a 16 port as I could always add on a small 8 port very low power consumption GbE switch to get my network to have sufficient ports) and no more than 30w of power consumption. Again on both sets of numbers I'd be willing to be a little flexible, but when the MS510 is right now on sale for $269 and gets you 1 10GbE, 2x5GbE and 2x2.5GbE with 5 1GbE (26w) the cost and power consumption are just a bit more than I'd be willing to pay.
It is getting close, but emphasis on GETTING close. Don't know when it'll happen, but seems more likely it is when the 2nd (3rd?) tier guys like DLink, Trendnet and TP-Link start offering multigig switches is when we will finally at least see cost come down to what is reasonable for me to invest in it. Not sure if power consumption will have, or if it'll be yet another generation of equipment before we are there.
Dug - Monday, October 28, 2019 - link
Apparently some of the new synology's will have 2.5Gb ports. I believe some new home routers from asus and netgear have them. So if you have a direct connection from your computer to the router, then that would act as the switch.But unfortunately that is very limited, as most will have a switch in between computers and routers.
rrinker - Thursday, October 24, 2019 - link
Not a wide use case, but it's out there. I have GigE everywhere, and it's generally fine, because I am the only truly computer savvy person in my house. But if others were doing what I do, dumping large files like HD movies and ISOs to my server, I'd most definitely want something faster than 1Gb on my server. I don;t see a need to make any workstations >1Gb yet, even the 1Gb in the server is fine for multiple HD video streams to multiple TVs at the same time.Once we get well beyond 1Gb internet speeds - 1Gb Ethernet will be the bottleneck, and we'll need something faster. I just got another speed increase with no bill increase, but that's now 300Mb, so no worries with 'only' 1Gb networking.
CharonPDX - Thursday, October 24, 2019 - link
My new 802.11ax WiFi router comes with one 2.5 Gb port. I have a Plex server with all my DVDs and Blu-ray discs ripped to it, as well as all my CDs, and a TV tuner to record live TV. There have been times when multiple people in the house have streamed from the server, and it has skipped frames. I wouldn't mind adding a 2.5G port to the server, which should stop that.Makaveli - Thursday, October 24, 2019 - link
Some of us are on Fiber ISP's that offer connections faster than 1Gbps. And don't need the expense of going with 10Gbps equipment or the noise. And some people time is money so yes the 2x longer transfers can't be an issue depending on what you are doing.Samus - Thursday, October 24, 2019 - link
It has a lot to do with existing infrastructure. Getting even CAT6 to handshake at 10Gbps is pretty tough, most installs need CAT6A if going beyond 30 feet or so. And jacks, couplers, etc, all further reduce distance.Many of the 10Gbps installations of done only negotiate at 5Gbps due to the infrastructure, and that's usually good enough for the client who is just trying to load balance a server across a gigabit network.
2.5Gbps in a home would be desirable for anybody trying to transfer\decompress files from a torrent server to their HTPC, but that, like many beyond-gigabit applications, is pretty niche. But that doesn't mean somebody doesn't want it.
Remember, the industry has to maintain a push for interfaces that meet future demands, not current ones. The way things are going there will be a need for 2.5Gb in homes within years. My gigabit wiring is actually the bottleneck of my $70/mo AT&T Fiber internet connection (which pulls a little over 1Gbps up/down.)
r4tch3t - Friday, October 25, 2019 - link
Where I live, New Zealand, they are testing consumer 10Gbit connections (symmetric?) and 1Gbit is common. So I could see the ISPs giving out a 1 10Gbit + 4 2.5Gbit switch so you can use the bandwidth when they start rolling it out.As for what you can use it for? Personally faster updates/downloads for games so when I turn my PC on to a several GB update I don't have to wait as long to join my friends in game. I can already pull close to 100MB/s through steam depending on the game. With 10Gbit I could make my SSD the bottleneck! O_O (assuming the servers can keep up)
Valantar - Friday, October 25, 2019 - link
I like to take photos and edit them in Lightroom, but I only have SSDs in my main PC and use a (DIY) NAS for mass storage. Browsing through a library of 50-70MB RAW photos over GbE is quite annoying. Even moving to 2.5G would mean a massive improvement in library responsiveness for this use case. As such, cards like this would save me the hassle of keeping "in-use" photos stored locally on my SSD and then archiving them on the NAS after editing, which causes quite a lot of work with library management.abufrejoval - Friday, October 25, 2019 - link
Why should you ride a bike, when you can walk?Why drive a Mercedes when you can go per pedes?
Why fly when you can swim?
Gbit Ethernet was ok, as long as your storage was a single HDD: Balance!
But with SSDs or even RAIDs somewhere in your network, Gbit grew from an irritation into an intolerable bottleneck.
Where the transition from 100 to 1000 Mbit was pretty smooth, the next step went all wrong. Yet NBase-T with the 2.5 and 5 Gbit intermediate steps IMHO have finally healed that.
From experience I can tell you that 10Gbit is a bit like driving a car at 400km/h instead of 100km/h: The linear comes with exponential stress and constraints, especially downtown, with small files and synchronization.
2.5Gbit is more like 150km/h and 5Gbit like 200km/h, each a much easier job, at least here in Germany.
And when I need to go from one major metropolitan area to another here, I don't take a plane (as you would in the US), but I take my car: Going 200km/h instead of 80km/h means being back home for lunch or dinner: A worthy premium.
And it's a similar story for copying VMs updating larger Docker containers; 3x, 5x and 10x speedups are significant enough to appreciate.
When you aggregate server class hardware, 10Gbit is rock bottom, 100Gbit is becoming mid-range. But it also takes that type of iron to take advantage of it and there you tolerate the noise and power it requires.
NBase-T gives you sweet spots that even work on laptops or ordinay desktops: I have upgraded my workstations to 10Gbit low-power Aquantia 10Gbase-T two years ago and more recently put 2.5 or 5 Gbit USB adapters even into the lowlier notebooks, because it is a nice match to their SSDs. Ultrabooks don't even come with Gbit Ethernet any more, so NBase-T USB Ethernet is a natural complement to their NVMe storage.
Now that Netgear and Buffalo NBase-T switches have equal port cost as these NICs (~€30-€80), it becomes a pretty simple cost/benefit calculation with linear price points.
Beyond-Gbit is mostly about LAN connectivity today: If Internet is your only network, you may not need it for a while longer. If most of your traffic is East-West, the more the better.
Mr.Vegas - Sunday, October 27, 2019 - link
1) Price, 2.5G is still plenty fast and costs much less then 5 or 102) Wide Compatibility, 2.5G and 5G only use PCIe x1, 10G uses PCIe x4
3) Hardware price: Routers, NAS etc with 2.5g and 5 G cost MUCH less then 10G hardware, there are plenty Switches that have 2-3 2.5G or 5G port and only single 10G
4) Main usage: Home NAS, Home Server connection, using PC in a closet/basement and feeling like you using it directly, 1G is not enough for this, especially if you use 4K resolution.
If you have bunch of people using home network for gaming, work, entertainment then 2.5G still offers twice the band witch so 2-3 people can use NAS, stream 4K at the same time.
5) Its the next step, 1G should be forgotten ASAP and 2.5g will be the new minimum, most new motherboards now come with Realtek 2.5g since its cheap.
6) This cheap 2.5g Chip opens up an opportunity to release new home use routers, switches etc for same price as we paid for 1G just with 2.5g its that cheap, Aquantilla 5G and 10G has nice premium, but this 2.5G realtek and soon to come Intel 2.5G mainstream nic are dirt cheap
JoeyJoJo123 - Monday, October 28, 2019 - link
Middleground on cost and what's achievable on existing cable runs without having to lay down new cable.Magnus101 - Tuesday, October 29, 2019 - link
For transfers where even a spinning harddisk is limited by a 1 gigabit speed, but not by 2.5 Gb.Then there is also the price of switches, which is the real limiting factor.
For now, I have to do with 1 GB for my home network, which means transfers for me maxes out at about 113 megabytes/s (which means that its about 0,9 Gigabit/s with the overhead in the switcher/network card that limits the speed by about 10%)
Kevin G - Tuesday, October 29, 2019 - link
Upgrading to 10 Gbit may require upgrading the cabling that is run through walls etc. and that can be a very expensive task for long runs. This is why 2.5 Gbit has seen major interest as the common Cat 5E cabling can be re-used for the higher speeds. The 2.5 Gbit only widens the symbol library vs. 1 Gbit as the line rate and encoding scheme are the same, hence they can leverage the same cabling.If you're cabling is mainly in-room, jumping to 10 Gbit is indeed mainly on the adapters and switch as the short patch cabling cost differences are minimal.
azazel1024 - Tuesday, October 29, 2019 - link
Faster network access and reduced latency.I mean, CAN I wait longer for transfers? Sure. However, its been annoying as heck since I moved I am down to a single GbE link between my desktop and server. I finally starting to get to wiring my new house rather than just hanging a few temp premade (and LONG) cables through my basement and up through the floor in the couple spots I needed wired access.
Anyway, I had 2xGbE previously. With SMB Multichannel that got me ~235MiB/sec transfers and down to around ~115MiB/sec now (that single link is spread on a 5 port switch to a couple of other devices right now). I am occasionally tossing big files around my network. No one will die if I have to wait 60 seconds instead of 30 seconds for a large file to transfer.
But again, it is still annoying. The every 10-16 month "opps I messed something up and now I need to transfer everything back over" data "[bring] backup" I have to do IS annoying as heck and also longer that I only have 1 good copy.
I keep everything mirrored between my desktop and server with a twice weekly backup job. I backup to an external drive when I remember. Which is usually about 4-6 times a year. So I would never lose everything, but I only have 1 most current copy, a second nearly current copy and one stale copy of my data. For anything important I make sure I don't remove the files from the source media till there are multiple copies (example, pictures from a trip off an SD card).
But if I am backing up, at 115MiB/sec that's about 7.5hrs. With dual links it takes about 4hrs (some of the smaller files it is disk based limitations and not link based).
Going to 2.5GbE would both allow me to be slightly faster than dual GbE links, but it would also allow me to get that performance with just a single network cable. Now, I am laying a crap ton in my house as I am doing the permanent wiring, but it still gives me more flexibility and performance (dual 2.5GbE!).
I'd still have disk based limits, but going to ONE 2.5GbE link would probably take my transfer times down to 3.5hrs from 4hrs at a guess. Dual 2.5GbE links would probably only squeeze maybe another 10-20 minutes faster in there as my RAID array likely can't exceed the performance of a 2.5GbE link except on the outer most tracks (about 320MiB/sec max read/write on the outer tracks with the 2x3TiB 7200rpm disks in the array, but by the time you are 20% utilized, you are down to more like 280MiB/sec max).
But it also leaves breathing room for replacing the disk arrays with either new HDDs at some point which likely would be 15-25% faster than my older 3TiB disks, or possibly some big TLC SSDs if prices come down somewhat more.
A low cost switch, especially one that is lower power is also crucial. As mentioned in the article, basically everything has at least a couple of 10GbE ports on it if it has slower ones. Even the little 10 port multigig switches I've seen are running around 30w and actively cooled. Probably in part because of those 10GbE ports. If there were some 8 and 16 port 2.5GbE switches (and maybe the same for 5GbE) that would be real nice. If they CAN get port price down to around $10 a port for 2.5GbE switches and $20 for 5GbE, I think you'd see pretty wide spread adoption.
Even if all they could do is $20 for 2.5GbE and $30 for 5GbE I think you'd still see a lot more adoption. Especially with those guys who need just a smaller switch for their 2-4 ports on their network they need as fast as possible and can then just connect it to a larger switch to handle everything else.
In my case, I see it being useful to connect my server to my desktop (minimum 2 ports there, ideally 4) and possibly 802.11ax router/APs later, which in my case is 2 total. An uplink port at 2.5GbE speed, or dual ports running at 1GbE to a larger switch is also needed. Then a main switch running all GbE ports (unless there is a single 2.5GbE uplink port) would be fine.
thomasg - Tuesday, October 29, 2019 - link
The vast majority of users don't even require Gigabit-Ethernet.Typical home networks don't actually transmit data between clients; by far the largest group of users only transmit data between their home-router and the clients, and by far most people have below 100 Mbit/s (the average in western countries is below 50 Mbit/s).
Then, there is the very small group actually using communication between clients in the network, like those who have NAS-boxes or other local network appliances.
Those might have modern WiFi equipment supporting over 1.5 Gbit/s in MU-MIMO.
Those are limited by 1 GbE, but won't come close to requiring 5, let alone 10 GbE.
2.5 GbE is the perfect sweetspot and costs only below 50$ to upgrade the home-server that serves the multiple WiFi users.
5 or 10 GbE not only is at least twice as expensive, it very likely also requires new cables adding additional cost, for no benefit over 2.5 GbE whatsoever.
Guspaz - Wednesday, October 30, 2019 - link
Both my Internet connection and my home file server are capable of saturating my gigabit LAN connection to my desktop, and my file server could probably saturate a 2.5 gigabit connection. A bit of a speed boost for file copies would be nice, or maybe reducing the impact of file access on Internet speeds.It's not critical, by any means, gigabit is plenty fast, but it'd be nice.
Mccaula718 - Wednesday, October 30, 2019 - link
The lucky people who have an ISP who offers >1Gbps internet speeds.TheinsanegamerN - Wednesday, February 26, 2020 - link
NAS connectivity. 1Gb tops out at 125MB/s, which even a single WD black can saturate today. But 2.5 Gb would allow 250 MB/s, possibly 300MB/s, which is what most home 4 bay RAID NAS units can hypothetically hit.Gamers would love something like this, with games hitting 150GB now, there needs to be a faster way to back all that junk up so it doest need to be redownloaded later.
10Gb would be even better, but costs. It's at least 4x as expensive to upgrade to a 10g ethernet card, and you need x4 slot, not a spare x1. More importantly, motherboards can put a 2.5Gb NIC onboard without major changes, not possible with the more demanding 10Gb, especially in Mini ITX format. 10Gb NAS units are also significantly more expensive then 2.5Gb NAS units.
eek2121 - Thursday, October 24, 2019 - link
I purchased a 10Gbit NIC based on the Aquantia stuff for $99 a while back. Works pretty well. Nothing affordable utilizes 10gbit though. They need to work on routers, cable modems, etc. My particular NIC also supports the 2.5 Gbit and 5 Gbit modes. I'd love to pick up a router that supports these modes along with getting some cable modem support.otherwise - Thursday, October 24, 2019 - link
Check out the MikroTik CRS305-1G-4S+IN. Four 10G SFP+ ports, passively cooled, $150.monsted - Thursday, October 24, 2019 - link
The last paragraph makes little sense. This card works in any PCIe slot, just like any other PCIe card would. PCIe is backward and forward compatible by definition. The only limitation is the speed of the slot and even a PCIe 1.0 x1 slot would support the 2.5 Gbps this card needs (maybe ever so slightly below optimal speeds because of overhead). A 3.0 (or 4.0) x1 card would do that just as well.sheh - Thursday, October 24, 2019 - link
Yeah, unclear. PCIe 2.0 x1 is exactly what this card needs. PCIe 1.0 is 2Gbps after encoding overhead.I also don't get the statement about the price:
"not a particularly low price considering the fact that GbE ports are ‘free’ with most motherboards and Aquantia-based 5 GbE and 10 GbE cards can be bought for $70 and $90"
2.5Gbps ports are hardly common on motherboards, and $35 is half or less the price of 5-10Gbps cards.
engineer_grunt - Thursday, October 24, 2019 - link
Working at RealTek I can tell you this is the typical bottoms up approach to market share acquisition. The goal is to get the lowest common denominator at no cost adder late market advantage. Intel is introducing 2.5Gbe as a standard speed in upcoming products and most WiFi extenders are requiring 2.5Gbe as well as not needing to changing cables.10Gbe/5/2.5/1 N-Base-T solutions came from the top down infrastructure equipment vendors like Aquantia and now merged with Marvell. Those solutions require 10.3125 Gbps SERDES so for power and area this is a cost concern as well.
The good news is that 2.5Gbe will be nearly free in upcoming designs over the next year or two for consumers. If you want higher speed other vendors already supply 10Gbps solutions like the AQR113 but draw typically 3Watts of power just for the PHY.
-Engineering Grunt
GreenReaper - Thursday, October 24, 2019 - link
Linux 5.4-rc1 has support for the RTL8125 controller in the r8169 (RTL8169) driver.8181244b-24ac-73e2-bac7-d01f644ebb3f@gmail.com/" target="_blank" rel="nofollow">https://lwn.net/ml/netdev/8181244b-24ac-73e2-bac7-...
https://www.spinics.net/lists/netdev/msg602045.htm...
https://github.com/torvalds/linux/commits/master/d...
There also appears to be a driver for 2.4 and 2.6 kernels (used by OpenWRT?) here:
https://github.com/heri16/r8125
GreenReaper - Thursday, October 24, 2019 - link
Looks like Ars Technica's URL filter combines unfavourably with its email address filter...GreenReaper - Thursday, October 24, 2019 - link
Or Anandtech, even. (What I get for opening the tabs next to each other at 4AM! :-)yetanotherhuman - Friday, October 25, 2019 - link
Great. But why aren't there tons of cheap 5/2.5Gbit switches around alreadyimaheadcase - Friday, October 25, 2019 - link
Why does anandtech keep saying no cheep 2.5GbE switches out? They have been for years cheap enough out. Are they copy and pasting info from years past or something?Sure not cheap if you go 8+ ports, but plenty around $200 mark for less ports.
RSAUser - Friday, October 25, 2019 - link
A $200 switch is not cheap, all 3 of my 1Gbps switches + access points are about $180 in total.Yes, I get 1Gbps on everything, including WiFi if ac support on the device (so pretty much all my devices besides the TV).
Most of my stuff is the Tenda line, seem to be the best budget option here.
R3MF - Friday, October 25, 2019 - link
Q - Is there any indication yet whether it is 2.5Gb or 5Gb that will win out as the mass-market standard for consumer gear?Because it will be one or the other, probably depending on whether 5Gb gear can be marketed at a price only marginally higher than 2.5Gb...
Mr.Vegas - Sunday, October 27, 2019 - link
This web site is 3 months late, I seen these on eBay 3-4 months agothomasg - Tuesday, October 29, 2019 - link
Honestly, I think this is a quite fairly priced product.1 or 2 GbE ports might be free on boards, but an additional card certainly isn't. And plenty of people still buy such a card for a second port apart from their onboard GbE.
Also, while it isn't 10 GbE, it still is 150% faster than GbE, for a very affordable price.
Also, the 2.5 GbE are available over longer distances and poorer cables; it is specified for 100 meters over Cat 5e which is still _very_ common.
10 GbE on the other hand requires Cat 6 and only specified up to 60 meters - and is much more heavily influenced by poor plugs, panels and workmanship.
10 GbE also heavily taxes systems, while 2.5 GbE is reasonable on most hardware.
This means that for many, it is a massive performance improvement for little overall cost and few problems.
It also is quite sufficient to attach modern high-performance WiFi equipment which is hampered by GbE but far away from requiring 5 or even 10 GbE.
This makes it a perfectly reasonable upgrade for a very reasonable price.
regsEx - Tuesday, October 29, 2019 - link
They've been on Ali since few months already