In theory, 2.5G and 5G are good for the market as they provide a lower cost competitor to 10G that should drive down prices of 10G equipment. However, my fear and what seems to be playing out in practice is that 2.5G and 5G are being used to prop up the price of current 10G equipment. Vendors are justifying their 10G pricing by pricing their half and quarter speed alternatives such that it is extremely short sighted not to spend the small additional premium to get all the way to 10G.
On the other hand, the AQN107 is selling for a reasonable price premium over the AQN108. The ability to support 2.5G, 5G, and 10G gives options for when the infrastructure is less than ideal. Switches with these extra modes will be desirable for the same reason. The AQN108 makes sense as an upgrade for systems that can't support an extra PCIe x4 card. However, while support for 2.5G and 5G at the switch is desirable, I see no reason to get a switch that doesn't also support 10G. It would be extremely short sighted. Likely, vendors will try to play with pricing, but we can already see that there isn't a large price difference. If a larger price gap develops, it will most likely be because vendors want to create the disparity and not due to a significantly greater cost of manufacturing 10G components.
The big advantage of 2.5G and 5G isn't port cost -- it's cable cost. 10Gbase-T won't run over the Cat5e/Cat6 cable that's installed pretty much everywhere -- it's even picky about Cat7. The cost of a $100 adapter is chump change relative to rewiring your building.
I feel that from a consumer POV if they need 10G it will be for the backbone, like from the main switch to NAS to the main PC most these stuff in most homes are in the same room already. the rest can be wired using 5gbps with ease.
For the way I see it, the 10G will be from NAS to main PC to main switch, all in same room. and 5G to the home theatre switch which will connect to several 1G devices like TV, Media Player, Game Console, ChromeCast, etc..
2.5G/5G will also serve AP's either point to point or chained ( like Switch-5G>------>AP>---->AP ),
2.5G will enough for single AC AP, 5G will be good for chaining multiple AP's.
For the shorter runs that are common at home lesser cabling than the officially rated sort is often still good enough. Cat6 (not 6A) is rated for 55m at 10G; runs that long in a residential network are rare. The bigger issue, especially with DIY cabling is that the quality of crimping at the cable ends needs to meet higher tolerances than with 1G.
The fun thing is that the 802.3bz standard don't refer to Cat5e or Cat6 but describe the requirements in the standard itself. I wonder if this is due to marketing.
Interesting. Sometimes that means that the standards bodies aren't getting along. According to wikipedia there are conflicting specifications for Cat6A.
"Confusion therefore arises because of the different naming conventions and performance benchmarks laid down by the International ISO/IEC and American TIA/EIA standards, which in turn are different from the regional European standard, EN 50173-1." - Source Wikipedia
I might be speaking of what I see only, but it's a very rare case when a home owner will need to run a 10G for that long.. even fiber run will cost less considering there's a lot of used gadget out there. and those new multi gigabit are still costly considering the need for not just 10G but for 5 & 2.5 as well
I don't think 10gbe would be used much for wiring a building. At least in the enterprise networks I work on, we only use them within the data center. Anything outside goes through fiber.
I would use the 5/10gb NIC at home for my ESXi host. The 1gbe ports it already has is sufficient for the 1gbps internet. The 5/10gbe would be great for connecting to the LAN access switch. I could have multiple large file transfers, while some devices are streaming HD from the media server, while the internet is being used to its max, without ever experiencing some bottlenecking on the network side.
The gaming NICs make me chuckle in this case. Perhaps there's a situation I'm not considering, but I'm struggling to find a scenario where I would need 5-10gbps on my gaming PC, considering very few likely have a WAN pipe like that?
Aside from the cable requirements, the actual phy for 5gbit is almost identical to 10gbit, so it makes sense that they're similarly priced. Aside from the extra pcie lanes, the actual chips are probably identical.
The big advantage of 5gbit is that it works with older wiring, not that it saves you a lot on IC costs (which it doesn't).
Very much looking forward to this era of Ethernet. 1G hasn't been enough to keep me happy. Link-aggregation isn't always an option, and it's less than ideal even when it is. Traditional 10G is starting to get reasonably priced, but still a but too much for the average home network. A $130 10G NBASE-T adapter is actually a reasonable price, and it can take advantage of existing Cat5e or Cat6.
10 GbE switches start to be affordable, especially used ones (ex-enterprise), since there is a push to 40G and above. I highly doubt people will want to invest in technology with half-baked implementation. I'd rather have 16+16 10G/1G rack-mounted switch bought off ebay for $100 and bunch of 10GbE cards that are around $40 each, than $130 cards that *could* possibly connect to some unannounced switches in the future... maybe.
Ex-enterprise switches tend to be loud and power hungry.
D-Link and TP-Link both have 24 1g copper + 4 10g SFP models at the $300-400 mark, Intel x520 cards are cheap on the server pulls market (Especially single port). All in all makes a for a reasonably priced home lab / small enterprise solution.
Yeah I thought suddenly a while ago..Hey let's buy a used Enterprise 10GbE switch! Then I remembered like you say, it will whine like a whiny thing. Not something I can use domestically.
i dont think used hardware can be market representative
the thing is today 1g ethernet nic costs 10 bucks and a 8 port switch 20 bucks that is 12.5 bucks per port for 8 devices solution.(8 10 bucks nics and 1 20 bucks switch 100us / 8 ports)
i think will be reasonable to pay 30 bucks per complete solution on 2.5g and 60 us for complete solution on 5g
that is 100 bucks for 2.5g 8 port switch and 20 bucks for each 2.5g nic
and 200 bucks for 5g 8 port switch and 40 bucks for each 5g nic
any cost beyond that is far from soho mass market and broad adoption of 2.5 and 5g ethernet
i think 99us per port for 5g ethernet is far from mass market, i think 60us is needed for nics and switches to make 5g ethernet ubiquitous
one advantage is pciexpress 3.0 x1 capable enough for 5g ethernet
10g ethernet can wait for most mass market
i think vendors are trying to use 2.5g and 5g ethernet to promote 10g ethernet at almost the same cost, deniying the masses the access to 2.5g and 5g at cheap prices
I haven't gone all wrecking crew in the house bc wireless is ehhh... acceptable and I cant justify drilling. As soon as they release 5GB switches, it will be time to bust out the drill and piss the wife off.
Ian: I think that may be important during tests: if that's possible please use both good quality cat 6e cables, and cheapest cat5/5e. If this tech is limited to 6/6e only (for meaningfull bumb ove 1GbE, there is little benefit for end-users.
10GbE 5,8,16 port switches need to drop below $20/port to be feasible, and realistically to ~$10/port for them to fully replace aging 1GbE switches.
I currently have a draw with a few spare 5 port and 8 port gigabit switches for when issues arise with a friend or coworkers network. Easier to troubleshoot. Until it makes financial sense to do that with 10GbE, it hasn't hit mainstream.
RDMA is great for SANs and stuff, but not useful for your normal network communication. And much of the benefit of RDMA is lost of Ethernet. Inifinband can give you 100ns RDMA latency, but 10Gb serialization takes way too long and you really need pass-through switching to appreciate any latency reductions.
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.
33 Comments
Back to Article
Xajel - Monday, May 15, 2017 - link
Nice, while still expensive, I hope now to see more multi-gigabit Switches targeted towards consumer/prosumer.And please Ubiquiti, remake the 8/16 ports PoE switches with multi-gigabit support.. and not over priced like your current ones...
chechito - Monday, May 15, 2017 - link
i agree, vendors are trying to sell 5g ethernet at almost the cost of 10g ethernetBurntMyBacon - Monday, May 15, 2017 - link
In theory, 2.5G and 5G are good for the market as they provide a lower cost competitor to 10G that should drive down prices of 10G equipment. However, my fear and what seems to be playing out in practice is that 2.5G and 5G are being used to prop up the price of current 10G equipment. Vendors are justifying their 10G pricing by pricing their half and quarter speed alternatives such that it is extremely short sighted not to spend the small additional premium to get all the way to 10G.On the other hand, the AQN107 is selling for a reasonable price premium over the AQN108. The ability to support 2.5G, 5G, and 10G gives options for when the infrastructure is less than ideal. Switches with these extra modes will be desirable for the same reason. The AQN108 makes sense as an upgrade for systems that can't support an extra PCIe x4 card. However, while support for 2.5G and 5G at the switch is desirable, I see no reason to get a switch that doesn't also support 10G. It would be extremely short sighted. Likely, vendors will try to play with pricing, but we can already see that there isn't a large price difference. If a larger price gap develops, it will most likely be because vendors want to create the disparity and not due to a significantly greater cost of manufacturing 10G components.
dtremit - Monday, May 15, 2017 - link
The big advantage of 2.5G and 5G isn't port cost -- it's cable cost. 10Gbase-T won't run over the Cat5e/Cat6 cable that's installed pretty much everywhere -- it's even picky about Cat7. The cost of a $100 adapter is chump change relative to rewiring your building.chechito - Monday, May 15, 2017 - link
i think that is nic vendors charging you a plus for avoiding change your cable??? very sad!!!5g nic costs 10 times 1g nic thats 100% overprice vs performance
Kvaern1 - Monday, May 15, 2017 - link
Highend always comes at a premium. Always.Xajel - Monday, May 15, 2017 - link
I feel that from a consumer POV if they need 10G it will be for the backbone, like from the main switch to NAS to the main PC most these stuff in most homes are in the same room already. the rest can be wired using 5gbps with ease.For the way I see it, the 10G will be from NAS to main PC to main switch, all in same room.
and 5G to the home theatre switch which will connect to several 1G devices like TV, Media Player, Game Console, ChromeCast, etc..
2.5G/5G will also serve AP's either point to point or chained ( like Switch-5G>------>AP>---->AP ),
2.5G will enough for single AC AP, 5G will be good for chaining multiple AP's.
DanNeely - Monday, May 15, 2017 - link
For the shorter runs that are common at home lesser cabling than the officially rated sort is often still good enough. Cat6 (not 6A) is rated for 55m at 10G; runs that long in a residential network are rare. The bigger issue, especially with DIY cabling is that the quality of crimping at the cable ends needs to meet higher tolerances than with 1G.yuhong - Tuesday, May 16, 2017 - link
The fun thing is that the 802.3bz standard don't refer to Cat5e or Cat6 but describe the requirements in the standard itself. I wonder if this is due to marketing.voicequal - Thursday, May 18, 2017 - link
Interesting. Sometimes that means that the standards bodies aren't getting along. According to wikipedia there are conflicting specifications for Cat6A."Confusion therefore arises because of the different naming conventions and performance benchmarks laid down by the International ISO/IEC and American TIA/EIA standards, which in turn are different from the regional European standard, EN 50173-1." - Source Wikipedia
Xajel - Tuesday, May 16, 2017 - link
I might be speaking of what I see only, but it's a very rare case when a home owner will need to run a 10G for that long.. even fiber run will cost less considering there's a lot of used gadget out there. and those new multi gigabit are still costly considering the need for not just 10G but for 5 & 2.5 as wellJorsher - Thursday, May 18, 2017 - link
I don't think 10gbe would be used much for wiring a building. At least in the enterprise networks I work on, we only use them within the data center. Anything outside goes through fiber.I would use the 5/10gb NIC at home for my ESXi host. The 1gbe ports it already has is sufficient for the 1gbps internet. The 5/10gbe would be great for connecting to the LAN access switch. I could have multiple large file transfers, while some devices are streaming HD from the media server, while the internet is being used to its max, without ever experiencing some bottlenecking on the network side.
The gaming NICs make me chuckle in this case. Perhaps there's a situation I'm not considering, but I'm struggling to find a scenario where I would need 5-10gbps on my gaming PC, considering very few likely have a WAN pipe like that?
I'm glad to see lower-priced options.
chechito - Monday, May 15, 2017 - link
i agree vendors try to boost 10g using (killing 2.5g and 5g) at almost the same pricesaratoga4 - Monday, May 15, 2017 - link
Aside from the cable requirements, the actual phy for 5gbit is almost identical to 10gbit, so it makes sense that they're similarly priced. Aside from the extra pcie lanes, the actual chips are probably identical.The big advantage of 5gbit is that it works with older wiring, not that it saves you a lot on IC costs (which it doesn't).
cdillon - Monday, May 15, 2017 - link
Very much looking forward to this era of Ethernet. 1G hasn't been enough to keep me happy. Link-aggregation isn't always an option, and it's less than ideal even when it is. Traditional 10G is starting to get reasonably priced, but still a but too much for the average home network. A $130 10G NBASE-T adapter is actually a reasonable price, and it can take advantage of existing Cat5e or Cat6.gundul - Monday, May 15, 2017 - link
This is a too long of a wait already..Vatharian - Monday, May 15, 2017 - link
10 GbE switches start to be affordable, especially used ones (ex-enterprise), since there is a push to 40G and above. I highly doubt people will want to invest in technology with half-baked implementation. I'd rather have 16+16 10G/1G rack-mounted switch bought off ebay for $100 and bunch of 10GbE cards that are around $40 each, than $130 cards that *could* possibly connect to some unannounced switches in the future... maybe.thewishy - Monday, May 15, 2017 - link
Ex-enterprise switches tend to be loud and power hungry.D-Link and TP-Link both have 24 1g copper + 4 10g SFP models at the $300-400 mark, Intel x520 cards are cheap on the server pulls market (Especially single port). All in all makes a for a reasonably priced home lab / small enterprise solution.
jabber - Monday, May 15, 2017 - link
Yeah I thought suddenly a while ago..Hey let's buy a used Enterprise 10GbE switch! Then I remembered like you say, it will whine like a whiny thing. Not something I can use domestically.chechito - Monday, May 15, 2017 - link
i dont think used hardware can be market representativethe thing is today 1g ethernet nic costs 10 bucks and a 8 port switch 20 bucks that is 12.5 bucks per port for 8 devices solution.(8 10 bucks nics and 1 20 bucks switch 100us / 8 ports)
i think will be reasonable to pay 30 bucks per complete solution on 2.5g and 60 us for complete solution on 5g
that is 100 bucks for 2.5g 8 port switch and 20 bucks for each 2.5g nic
and 200 bucks for 5g 8 port switch and 40 bucks for each 5g nic
any cost beyond that is far from soho mass market and broad adoption of 2.5 and 5g ethernet
chechito - Monday, May 15, 2017 - link
i think 99us per port for 5g ethernet is far from mass market, i think 60us is needed for nics and switches to make 5g ethernet ubiquitousone advantage is pciexpress 3.0 x1 capable enough for 5g ethernet
10g ethernet can wait for most mass market
i think vendors are trying to use 2.5g and 5g ethernet to promote 10g ethernet at almost the same cost, deniying the masses the access to 2.5g and 5g at cheap prices
thewishy - Monday, May 15, 2017 - link
I don't think 1ge is "Mass market" these days; most people use wifi... Even offices are going in that direction...It's only ever going to be an enthusiast / business solution.
chechito - Monday, May 15, 2017 - link
i think wired vs wireless topic is far from this discussionSpetsnazAntiVIP - Tuesday, May 16, 2017 - link
Cool let me know when they start offering a consumer 10 gbe switch for $100-$150 and I'll be the first to buy it.vailr - Monday, May 15, 2017 - link
Latest Aquantia drivers & discussion:station-drivers (dot) com /index.php?option=com_kunena&view=topic&catid=7&id=2520&Itemid=255&lang=en#14855
dgingeri - Monday, May 15, 2017 - link
Now if someone were to just release a reasonably priced switch for 5Gbe. That would be nice.Manch - Monday, May 15, 2017 - link
I haven't gone all wrecking crew in the house bc wireless is ehhh... acceptable and I cant justify drilling. As soon as they release 5GB switches, it will be time to bust out the drill and piss the wife off.Vatharian - Monday, May 15, 2017 - link
Ian: I think that may be important during tests: if that's possible please use both good quality cat 6e cables, and cheapest cat5/5e. If this tech is limited to 6/6e only (for meaningfull bumb ove 1GbE, there is little benefit for end-users.James5mith - Monday, May 15, 2017 - link
10GbE 5,8,16 port switches need to drop below $20/port to be feasible, and realistically to ~$10/port for them to fully replace aging 1GbE switches.I currently have a draw with a few spare 5 port and 8 port gigabit switches for when issues arise with a friend or coworkers network. Easier to troubleshoot. Until it makes financial sense to do that with 10GbE, it hasn't hit mainstream.
chechito - Monday, May 15, 2017 - link
keep in mind cheap port prices stay only on 5 and 8 ports switches, bigger switches offer more expensive cost per porti think 15 bucks per port on small switches of 5 and 8 ports for 2.5g is possible
30 bucks per port for 5.0g too
but the problem is the nics, we need 30 bucks nic en 2.5g and at least 60 bucks noc on 5.0g to make reacheable for many people
10 buck per port'?? i think not in this decade
cygnus1 - Monday, May 15, 2017 - link
Bet these don't support RDMA though... I'll stick to the used 10g gear off ebay.bcronce - Monday, May 15, 2017 - link
RDMA is great for SANs and stuff, but not useful for your normal network communication. And much of the benefit of RDMA is lost of Ethernet. Inifinband can give you 100ns RDMA latency, but 10Gb serialization takes way too long and you really need pass-through switching to appreciate any latency reductions.KabyLake - Monday, May 15, 2017 - link
I using San Francisco,California - 1 Gigabit Home Internet i wonder if my number will reach 1000 mbps.