Is now the time to go to 5 Gbps or 10 Gbps home networking?

[tiddlywink]

..

 

[sfx2000]

Well - one can consider a migration approach...

Find the gear that supports things on the end-points first - and then fill in the middle with a switch...

Place to look would be on the unmanaged switches as compared to similar number of ports for 1Gbe - we're almost there with Realtek on the switching layer...

@TheLostSwede - comments here?

 

[coxhaus]

Why bother with 5G and 10G when you can get 100G? If you question the need for 100G, you should also be questioning the need for 5G/10G.

I think 10gig is a much better target for a copper solution.

2.5gig to me is ho-hum, not much of an upgrade. Not worth it to me.

 

[TheLostSwede]

Well - one can consider a migration approach...

Find the gear that supports things on the end-points first - and then fill in the middle with a switch...

Place to look would be on the unmanaged switches as compared to similar number of ports for 1Gbe - we're almost there with Realtek on the switching layer...

@TheLostSwede - comments here?

Realtek has had a "cheap" 10 Gbps switching IC for a few years now and it's what is already used in a lot of the low cost SFP+ switches. The issue is the PHY side of things, as there isn't a low cost 10 Gbps PHY as yet, but based on my chats with Realtek at Computex, it's something they're working on, but it's not likely to happen for another year or two.

 

[sfx2000]

Realtek has had a "cheap" 10 Gbps switching IC for a few years now and it's what is already used in a lot of the low cost SFP+ switches.

True - but at the moment, their 2.5Gbe solutions enable some very attractive switches - both managed and unmanaged - unmanaged we're at the point where the cost delta with 1Gbe is basically zero...

 

[TheLostSwede]

True - but at the moment, their 2.5Gbe solutions enable some very attractive switches - both managed and unmanaged - unmanaged we're at the point where the cost delta with 1Gbe is basically zero...

5 Gbps switches will likely use their 10 Gbps switch IC combined with their 5 Gbps PHY's, which means 5 Gbps switches will still cost less than 10 Gbps switches, but won't be anywhere near as cheap as 2.5 Gbps switches.
No-one else appears to be working on 5 Gbps switching though, making 5 Gbps the least attractive option overall, despite offering a good compromise on speed vs. cost vs. thermals.

 

[Digilog]

No-one else appears to be working on 5 Gbps switching though, making 5 Gbps the least attractive option overall, despite offering a good compromise on speed vs. cost vs. thermals.

Well we had originally 1Gb then 10 Gb then 40Gb in base-T. Its nice to see 10Gb finally getting mass produced since its been around for years. 40Gb, they didn't develop much outside some xenon setups and a lot skip to 100G. Finally 100G cards have came down in price since PCIe4 was released.

 

[heysoundude]

long story short, if you're considering 1Gbps symmetrical service from your ISP, your network should be 2.5G capable at the bare minimum.
2.5G is the emerging standard that we should all shoot for, even if we only pay for slower connections. Further, in building our networks with greater capabilities, we have to let our ISPs know what our networks are capable of so that they look into offering appropriate speeds and pricing tiers. My ISP CSR sighed when I asked if I could have their cable modem with the 2.5G port rather than the old one with the 1G port. I'd already established that I'm not your "Joe Average" customer when it comes to this stuff, and he put me on hold to work some magic on his end to make my request happen. (I suspect he may have to "make a mistake" to get my line over provisioned for 1Gbps speeds, and let the company audit realize I'm only paying for 120Mbps to get the hardware I requested).

 

[Digilog]

long story short, if you're considering 1Gbps symmetrical service from your ISP, your network should be 2.5G capable at the bare minimum.
2.5G is the emerging standard that we should all shoot for, even if we only pay for slower connections. Further, in building our networks with greater capabilities, we have to let our ISPs know what our networks are capable of so that they look into offering appropriate speeds and pricing tiers. My ISP CSR sighed when I asked if I could have their cable modem with the 2.5G port rather than the old one with the 1G port. I'd already established that I'm not your "Joe Average" customer when it comes to this stuff, and he put me on hold to work some magic on his end to make my request happen. (I suspect he may have to "make a mistake" to get my line over provisioned for 1Gbps speeds, and let the company audit realize I'm only paying for 120Mbps to get the hardware I requested).

Cable Internet and their pricing tiers has always been a ripoff. Because most of the time the connection doesn't use the bandwidth. In the US the internet connections was purposely delayed more. This was the outcome of a "bandwidth equality" garbage policy because everyone's internet was faster and bandwidth size isn't a factor. But even testing bandwidth on a friend's connection that he paid maximum bandwidth still doesn't give me a 5ms ping from google that I had on my 200Mb connection in 2003. Same connection today is 15ms. or worse.

 

[rborth]

Consider that in the early 1980s the state of art was Sneaker net. That was when you copied files to a floppy and walked to another computer and inserted it in a drive. Throughput then depended somewhat on who was walking an how far. If the entire floppy was full I figure the through put was in the neighborhood of 20 bytes per second. 1 Gigabit per second Ethernet was a vast improvement but still sufficient in a lot of cases.

Santa is bring me an Asustor NAS that will house a RAID 5 HDD array and A RAID 5 of PCIe 3.0x1 M.2 SSDs. I have the expectation that the two 2.5Gb/S for a total of 5Gb/S Ethernet could be saturated at times.

My point as an old guy is the definition of what fast enough is a moving target and usually in an upwards direction.

 

[GHammer]

I have a few servers and a NAS on my LAN.
When it became possible to go 2.5, I went.
I'm not going to spend the cash, power, or cooling for 10 Gb internally.
If I do anything video related, I find my 4 Tb SSD both cheaper and faster than any LAN I'm going to afford.
The 2.5 has been great for streaming to the TV, moving files around the LAN and for capable devices, better throughput on WiFi.

 

[Digilog]

I'm not going to spend the cash, power, or cooling for 10 Gb internally.

Since the prices of 10G networking went down, I started updating my network to 10G. I have 2 computers that I had for the past ten years plus I just bought a router motherboard for $60 and a assembled 1U w/o drives for under $200 from the server recyclers that has many 10G ports. But there are others 10G solutions that are reasonable, like the Banana Pi boards.

I don't see the need to go to the current fast speed commercial/industial connections (25 GBb , 40Gb, or 100Gb) at my house.

 

[GHammer]

Since the prices of 10G networking went down, I started updating my network to 10G. I have 2 computers that I had for the past ten years plus I just bought a router motherboard for $60 and a assembled 1U w/o drives for under $200 from the server recyclers that has many 10G ports. But there are others 10G solutions that are reasonable, like the Banana Pi boards.

I don't see the need to go to the current fast speed commercial/industial connections (25 GBb , 40Gb, or 100Gb) at my house.

I'm not going to have used (or new) commercial equipment in my house. Noisy, power hungry, ugly (IMO).
I also learned that homebrew equipment isn't always worth the cost in time for me.
I have the equipment in my signature. Runs great, no heat issues, no big power draw, wife loves the look of it. Serves my actual needs well.

 

[Digilog]

I'm not going to have used (or new) commercial equipment in my house. Noisy, power hungry, ugly (IMO).
I also learned that homebrew equipment isn't always worth the cost in time for me.
I have the equipment in my signature. Runs great, no heat issues, no big power draw, wife loves the look of it. Serves my actual needs well.

I do different things than a average consumer with a computer. But there are several different options for 10G that don't include a datacenter server, but if you want to know, the average wattage of one of those servers is only pulling 35W. Its when you stick a raid stack of dives in it they start to pull wattage.

As far as the noise, my workstation tower PCs are quiet. So I guess you are thinking about server tower or a rack mount like the 1U with load fans that sound like a hair dryer.

 

[Digilog]

5 Gbps switches will likely use their 10 Gbps switch IC combined with their 5 Gbps PHY's, which means 5 Gbps switches will still cost less than 10 Gbps switches, but won't be anywhere near as cheap as 2.5 Gbps switches.
No-one else appears to be working on 5 Gbps switching though, making 5 Gbps the least attractive option overall, despite offering a good compromise on speed vs. cost vs. thermals.

The thing about that, there is little to no differences in prices of the parts that make a 10Gb vs 2.5 Gb switch.

 

[sfx2000]

Again - I'll say that 2.5Gb is essentially free these days compared to 1Gb...

Stepping up to 10Gb is still an investment that most folks do not need...

 

[Digilog]

Stepping up to 10Gb is still an investment that most folks do not need...

A lot of people said that about 1Gb networking when it emerged into the consumer market.

Its all in application of the network and the use demands involved. Residential networks don't normally have a lot of devices requiring a lot of constant bandwidth at the same time.. I do like how the industry went to spf and spf+ uplink ports on switches so the network segment can be upgraded or customized for the application (like a fiber run to a different building for electrical isolation) The price of SPF+ ports have gone down along with base-T and fiber modules to where the price differences are not very much.



When looking for networking accessories and wiring for a network design people should look at the "future proof" aspect of it. I future proof a network a long time ago when I was a network contractor and performed an install at a data center and used my cat8 cable scraps instead of recycling them. When I sold that house, the real estate agent was impressed that I future proof the house network. At that time I didn't even realize that putting a security system and camera system in was a selling perk.

 

[TheLostSwede]

The thing about that, there is little to no differences in prices of the parts that make a 10Gb vs 2.5 Gb switch.

Of course there is, it's a huge difference, as all 10 Gbps switches today use dedicated PHY's and you need one per port, whereas recent 2.5 Gbps switches integrate the PHY side into the switching IC.

This is MaxLinear's single chip switching solution for up to 10 2.5 Gbps ports. Even older 2.5 Gbps switches with individual PHY's would be a lot cheaper, as a 2.5 Gbps PHY is around $1-2 at the most, whereas a 10 Gbps PHY is at least around $20-30.

MxL86282C - MaxLinear

The MxL86282C is a highly integrated 2.5G Ethernet switch with eight integrated 2.5GBASE-T PHYs. The MxL86282C also supports two 10G/ 2.5G / 1G uplink por

www.maxlinear.com

This is the PCB in the TP-Link TL-SX105 10 Gbps switch. It might not look that different to you, but there's a huge difference.

Pic borrowed from https://einfoldtech.com/

Next time, please read up on the topic before making stupid comments.

 

[TheLostSwede]

Again - I'll say that 2.5Gb is essentially free these days compared to 1Gb...

Stepping up to 10Gb is still an investment that most folks do not need...

Close, but not quite, at least not on the switch side.

Intel lists their recommended pricing for their Ethernet chips.

Latest 2.5 Gbps

Intel® Ethernet Controller I226-V - Product Specifications | Intel

Intel® Ethernet Controller I226-V quick reference with specifications, features, and technologies.

www.intel.com

Latest 1 Gbps (which is from 2015)

Intel® Ethernet Connection I219-V - Product Specifications | Intel

Intel® Ethernet Connection I219-V quick reference with specifications, features, and technologies.

www.intel.com

The difference is about 80 cents.

Things are more complicated on the switch side, as there is a handful of companies making switches, vs. only two making 2.5 Gbps Ethernet controllers, so far.

And yes, most people "don't need" 10 Gbps Ethernet, until they try it and find out how much faster it is, then they'll want it. At least if they do any kind of actual networking at home, outside of accessing the internet. Is it for all consumers, no, far from it, but this is where 2.5 Gbps can make a difference, since as you say, it's getting very cost competitive with 1 Gbps Ethernet these days, and it will still cut the time a lot of things take in half.

 

[sfx2000]

Things are more complicated on the switch side, as there is a handful of companies making switches, vs. only two making 2.5 Gbps Ethernet controllers, so far.

And yes, most people "don't need" 10 Gbps Ethernet, until they try it and find out how much faster it is, then they'll want it. At least if they do any kind of actual networking at home, outside of accessing the internet. Is it for all consumers, no, far from it, but this is where 2.5 Gbps can make a difference, since as you say, it's getting very cost competitive with 1 Gbps Ethernet these days, and it will still cut the time a lot of things take in half.

We'll get to a point where residential broadband will force the issue for the higher rate NIC's and switches, and 2.5 wont be enough....

When we get there, prices will align..

Interesting that STH reviewed a 10G-BaseT switch that was $200USD the other day - $25/port is actually not bad, and it's showing where the trend is heading...

XikeStor SKS1200-8XGT Review A Cheap 8-port 10Gbase-T Switch

At $199, the 8-port 10GbE XikeStor SKS1200-8XGT switch is one of the cheapest unmanaged 10Gbase-T switches on the market, with caveats

www.servethehome.com

Here's the link...

https://www.amazon.com/gp/product/B0CRGTQYPX?tag=snbforums-20