Why are Ethernet speeds so stagnent?

[pyrojoe34]

Why is it that Ethernet speeds have been so stagnant for years and the standard found on almost all consumer hardware is a measly 1Gbps which has been a thing since like 1999? Every other external consumer connection has blown past Ethernet (USB at 10Gbps, TB at 40Gbps, even WIFI at 3+Gbps). 10Gbps Ethernet has been a thing forever but you still never see it on consumer hardware, why? Even 100Gbps exists and 400Gbps is expected sometime next year. High end consumer motherboards will just put in two 1Gbps ports rather than bumping up to a 10Gbps port and the newest chipsets still only support 1Gbps natively. It seems crazy to me.

 

Personally, I'd really like to be able to actually max out my NAS speeds without paying the premium price of enterprise routers/switches and PCIe cards. I'd also like to stream an uncompressed 4k signal from my nice PC to my HTPC/TV but at 1Gbps I can't even stream 1080p/60 without compression and lag.

 

It just seems weird that every aspect of consumer hardware takes regular leaps and bounds but Ethernet hasn't changed in almost 20 years. I get that many consumers won't saturate 1Gbps but I can't imagine a 10Gb port being much more (if any more) expensive to produce than a 1Gb port, what's the holdup?

[BubblyCharizard]

very few consumers saturate the 1Gb port, with very few people having fast enough internet connections that warrant it

[Bananasplit_00]

because almost noone maxes out what we have today and its mainly used for internet and verry few people actiually have internet speeds that exede 1GB/s?

[pyrojoe34]

Just now, BubblyCharizard said:

very few consumers saturate the 1Gb port, with very few people having fast enough internet connections that warrant it

That doesn't seem like a valid excuse to me, especially since I don't see the higher speed ports being that much more expensive to produce. The vast majority of consumers also don't saturate PCIe speeds or SSD speeds or RAM speeds plenty of other things yet that hardware progresses rapidly and consumers usually get the newest tech at the same time or only slightly after enterprise gets the tech.

[BubblyCharizard]

1 minute ago, pyrojoe34 said:

That doesn't seem like a valid excuse to me

its harder to market a 10Gb Ethernet port than improvements in USB3.1 Gen 2 or TB3 since most don't care about ethernet

[pyrojoe34]

1 minute ago, Bananasplit_00 said:

because almost noone maxes out what we have today and its mainly used for internet and verry few people actiually have internet speeds that exede 1GB/s?

See my reply above as to why this argument doesn't make sense to me. Also, it's 1Gbps which is only 125MB/s anyone who has a NAS (which consumer NAS boxes are becoming more and more common) will be able to exceed with even a budget HDD.

 

Still, every other tech goes way beyond what most consumers use but that doesn't stop it from improving rapidly. Why wouldn't high end consumer hardware have higher speeds then?

[mariushm]

The problem is mostly the requirement for signal strength / data integrity up to 105 meters.

A network card needs a lot of power to push more than 1 gbps all the way on those 100 meters of wires.  That's why a lot of speed advancements (10gbps, 40gbps and higher) were done on fiber optics first and then on copper. 

They could have made network cards cheaply and working at higher speeds like 10 gbps or  maybe up to 40 gbps if you'd limit the cable length to let's say maximum 20 meters. You can do right now up to around 25 gbps on 3 meters of displayport cable (or something like that) so processors are quite capable of pushing data through a cable, but not at huge lengths.

The newest ethernet standard 802.3 bz allows the use of more powerful algorithms and signal processing stuff in order to make it possible to transfer 2.5gbps on regular cat5e cable and 5gbps or 10 gbps on cat6/cat6a  ... this standard was on the works for years.

 

 

 

 

[Bananasplit_00]

Just now, pyrojoe34 said:

See my reply above as to why this argument doesn't make sense to me. Also, it's 1Gbps which is only 125MB/s anyone who has a NAS (which consumer NAS boxes are becoming more and more common) will be able to exceed with even a budget HDD.

 

Still, every other tech goes way beyond what most consumers use but that doesn't stop it from improving rapidly. Why wouldn't high end consumer hardware have higher speeds then?

because almost noone uses that speed. a NAS isnt a comon thing, the avrage internet speed is way lower then what the port provides too. id love to see it going forward but for almost everyone, me included its not needed. my NAS dosent saturate my Ethernet port for example and i have one single 1TB HDD in it, the reason being that i dont move big ass files all the time

[BubblyCharizard]

1 minute ago, wrathoftheturkey said:

Dude..... Average bandwidth in the U.S. only recently exceeded 50 Mbps http://www.speedtest.net/reports/united-states/

and here I am with 175/175

 

 

 

[pyrojoe34]

5 minutes ago, wrathoftheturkey said:

Dude..... Average bandwidth in the U.S. only recently exceeded 50 Mbps http://www.speedtest.net/reports/united-states/

Since when has computer hardware been designed for what the "average" person will use? The "average" person won't even saturate a 2nd gen i3 or 4GB of RAM but that doesn't stop it from progress and innovation. The "average" flash drive won't even touch 10Gbps yet even the cheapest laptops have USB3.1. The "average" person doesn't have 4k yet YT supports 4k.

That seems like a cop out argument to me. "Average" should not be the measure of what to put into a product and lack of any improvement in the consumer PC market for 20years is insane to me.

[Blebekblebek]

The title obviously saying Ethernet and yet almost all answers mention "it's because the internet speed"

Let's assume all of you have 100GBps internet speed, and 100GBps network (Yes that's BYTE)

How are you going to fully utilize it with a single computer?

 

The answer is, you can't, there's no drive FAST enough yet, even again let's assume there is, you are not going to like the price tag that it offers.

 

"INTERNET SPEED" is not the bottleneck

 

The bottleneck is consumer grade drive that can't keep up with the speed.

Regular Hard drive still averaging at 1Gbps.

 

Until this problem is solved and being distributed around the world, we will stuck with 100/1000/100000 for a long time.

[pyrojoe34]

2 minutes ago, wrathoftheturkey said:

Barring industrial usage and an EXTREMELY tiny niche, there's just no use investing in the infrastructure to support a network bandwidth no one benefits from. And for that tiny niche, there are switches, much like there are GTX Titans for people who aren't satisfied with the general consumer market. 

Then why have 2-5Gbps Wifi on cheap consumer routers if "there's just no use investing in the infrastructure to support a network bandwidth no one benefits from". I'm not disagreeing with you that most people don't use it, I guess my problem is that everything else seems to progress beyond what 95% of people will use yet it progresses anyway. I just don't understand by Ethernet is the thing that doesn't follow this trend.

[porina]

15 minutes ago, mariushm said:

The problem is mostly the requirement for signal strength / data integrity up to 105 meters.

Basically this. Your typical cat5 or cat6 would struggle to carry the signal over distance, which is defined in the ethernet spec. Other standards that have gone faster are shorter cable length. Another trick when you want more bandwidth is simply to have more channels, but again, there's no scope to expand ethernet here as all wires are already used. Well, you could do channel bonding on more than one connection... or define a new connector and cable, but lack of backwards compatibility will hinder adoption and keep pricing high anyway.

 

Personally I'd love to have say 10gig which could offer near local levels of storage speeds with a remote NAS and in this scenario a point to point solution could be adequate and much more affordable.

[BubblyCharizard]

2 minutes ago, pyrojoe34 said:

2-5Gbps Wifi

you can't get that outside of a lab

 

real world speeds are substantially less, its just advertising

[pyrojoe34]

1 minute ago, Blebekblebek said:

The title obviously saying Ethernet and yet almost all answers mention "it's because the internet speed"

Let's assume all of you have 100GBps internet speed, and 100GBps network (Yes that's BYTE)

How are you going to fully utilize it with a single computer?

 

The answer is, you can't, there's no drive FAST enough yet, even again let's assume there is, you are not going to like the price tag that it offers.

 

"INTERNET SPEED" is not the bottleneck

 

The bottleneck is consumer grade drive that can't keep up with the speed.

Regular Hard drive still averaging at 1Gbps.

 

Until this problem is solved and being distributed around the world, we will stuck with 100/1000/100000 for a long time.

Even cheap HDDs can do at least 1.2Gbps and most laptops now come with at least a SATA SSD (~6Gbps) and higher end ones have NVME drives capable of doing upwards of 24Gbps so most drives won't be the bottleneck in the system either. I guess my problem is that everything else seems to progress beyond what 95% of people will use yet it progresses anyway. I just don't understand by Ethernet is the thing that doesn't follow this trend.

[Electronics Wizardy]

32 minutes ago, pyrojoe34 said:

That doesn't seem like a valid excuse to me, especially since I don't see the higher speed ports being that much more expensive to produce. The vast majority of consumers also don't saturate PCIe speeds or SSD speeds or RAM speeds plenty of other things yet that hardware progresses rapidly and consumers usually get the newest tech at the same time or only slightly after enterprise gets the tech.

the other problem is that the 10gbe copper stuff uses much more power you can't put it in a laptop easily. Fiber is the way to go, but require new wires. 

 

Also most homes are cat 5, and you can't put 10gbe over than without rewiring.

[NTDaws]

Because regular consumers don't need that fast of internet. Simple as that. 1. It costs more 2. Makes zero sense for consumers.

[pyrojoe34]

9 minutes ago, Being Delirious said:

1. It costs more

Is that actually true though? Does it cost more to produce or is it just because of the "enterprise tax" (I'm actually asking since I don't know). I don't actually see why a 10Gbps port would cost more to produce, it's a few more traces connecting the port to the CPU or chipset which is a negligible cost, the port is the same so there are no new molds for factories, the cables are basically the same (maybe a lower gauge wire so more copper cost but, like using a USB2.0 drive in a 3.0 port, you could just use a cheaper cat5 or 6 cable if you don't want the full 10Gbps and don't want to pay for the cable). So what costs more? 10Gbps PCIe cards are relatively cheap so it doesn't seem like it's any more costly to produce.

 

Happy Birthday btw (if your signature is correct).

[NTDaws]

Just now, pyrojoe34 said:

Is that actually true though? Does it cost more to produce or is it just because of the "enterprise tax" (I'm actually asking since I don't know). I don't actually see why a 10Gbps port would cost more to produce, it's a few more traces connecting the port to the CPU or chipset which is a negligible cost, the port is the same so there are no new molds for factories, the cables are basically the same (maybe a lower gauge wire so more copper cost but, like using a USB2.0 drive in a 3.0 port, you could just use a cheaper cat5 or 6 cable if you don't want the full 10Gbps and don't want to pay for the cable). So what costs more? 10Gbps PCIe cards are relatively cheap so it doesn't seem like it's any more costly to produce.

Faster network = more money, also keep in account in order to get these full connections, you need a good CPU for it too. A G4560 just wouldn't cut it.

[Blebekblebek]

Just now, pyrojoe34 said:

Even cheap HDDs can do at least 1.2Gbps and most laptops now come with at least a SATA SSD (~6Gbps) and higher end ones have NVME drives capable of doing upwards of 24Gbps so most drives won't be the bottleneck in the system either. I guess my problem is that everything else seems to progress beyond what 95% of people will use yet it progresses anyway. I just don't understand by Ethernet is the thing that doesn't follow this trend.

You're only as fast as your slowest team member

 

This quora thread already have some clue for your question

https://www.quora.com/Why-hasnt-Ethernet-gotten-any-faster-in-the-last-15-years

 

Now the issue is, are you willing to pay for it?

It takes like 10 years to progress from 100mbps to 1Gbps

Even new hardware (mostly SOC) still only provide 100mbps NIC, because beyond that it's useless.

[Electronics Wizardy]

6 minutes ago, pyrojoe34 said:

I don't actually see why a 10Gbps port would cost more to produce

it costs more, but not as much as it does now. 

 

You need more sheilding, a bigger chip, thats not currently built in. 

[NTDaws]

4 minutes ago, pyrojoe34 said:

Is that actually true though? Does it cost more to produce or is it just because of the "enterprise tax" (I'm actually asking since I don't know). I don't actually see why a 10Gbps port would cost more to produce, it's a few more traces connecting the port to the CPU or chipset which is a negligible cost, the port is the same so there are no new molds for factories, the cables are basically the same (maybe a lower gauge wire so more copper cost but, like using a USB2.0 drive in a 3.0 port, you could just use a cheaper cat5 or 6 cable if you don't want the full 10Gbps and don't want to pay for the cable). So what costs more? 10Gbps PCIe cards are relatively cheap so it doesn't seem like it's any more costly to produce.

 

Happy Birthday btw (if your signature is correct).

It tis thanks. Watch this video, I believe it talks about the cpu or something. It's been a while.

 

 

[mariushm]

27 minutes ago, pyrojoe34 said:

Is that actually true though? Does it cost more to produce or is it just because of the "enterprise tax" (I'm actually asking since I don't know). I don't actually see why a 10Gbps port would cost more to produce, it's a few more traces connecting the port to the CPU or chipset which is a negligible cost, the port is the same so there are no new molds for factories, the cables are basically the same (maybe a lower gauge wire so more copper cost but, like using a USB2.0 drive in a 3.0 port, you could just use a cheaper cat5 or 6 cable if you don't want the full 10Gbps and don't want to pay for the cable). So what costs more? 10Gbps PCIe cards are relatively cheap so it doesn't seem like it's any more costly to produce.

 

Happy Birthday btw (if your signature is correct).

AGAIN. It's difficult because the network card MUST be able to transmit the signals up to 105 meters through the cable.

 

In order for signals to reach the other end you need a lot of power. Also, at 10 gbps you need to be able to send up to 10 billion pulses a second through the wires, so whatever component is used to create those pulses (mosfet or transistor or circuit) must be so powerful to change the signal from 0v to let's say 2.5v within 1.25nanosecond (that's the length of time each symbol has on the wire.. so the voltage has to raise from 0v to some higher voltage like 2.5v (or the other way around, go from high voltage to low voltage) and stabilize there and all this is within 1.25 nanoseconds.

These transmission and receiver "modules" are the expensive part and are outside the actual chipset.

 

The chipset also has to include a lot of extra transistors and signal processing code in order to keep up with the rate of these symbols and also decode the weakened signals and do a lot of on-the-fly error correction because bits will inevitably be corrupted.. all this extra brains can raise the cost of a chip from let's say 1$ to at least 5$... but that's only the actual silicon cost, you still have to recover the money you spent on research and development, tens of thousands of man-hours and several spins of hardware chips (for tens of thousands of dollars each time) and testing and debugging and loads of other things. 1gbps is comparatively MUCH easier.

 

[mariushm]

21 minutes ago, wrathoftheturkey said:

Do we? Last I checked, experimental measurements cap under 100 Mbps for 2.4GHz routers, and when's the last time you got a Gigabit connection over wifi?

Some routers advertise more than 1 gbps on wireless but that's basically something like 1200 mbps on 5ghz  + 300 mbps on 2.4 gbps .... so they can advertise such router as 1.5 gbps  but no device out there will communicate with the router on BOTH wireless networks at same time.

 

And that 1200 mbps is spread across two or three "channels" and your wireless network card would also need to have 3 "channels" (or whatever they're called) in order to reach those peak speeds - cheaper wireless cards would only have 1 or 2 "channels" so even if your router can do 1200 mbps on 5ghz, your wireless card may only do 433 mbps or 866 mbps.

 

Also wireless is half duplex (kinda), you can't receive AND transmit data simultaneously, so you actually have half the bandwidth in either direction and you have to pause download to upload something and the other way around, which reduces the overall throughput.. Also, the max speed degrades with distance.

 

So in reality, you would be fine with a plain 1gbps connection to the outside world because really, even that 1200 mbps wireless on 5ghz won't top the 1 gbps link.

 

However, such use cases is why the new 802.3bz standard was ratified ... this made it possible to transmit 2.5gbps over plain standard cat5e (the cable that's rated for 1gbps) just by using better signal processing in the chips. Speeds of 5gbps and 10gbps are also possible but with cat6 cable and basically for those speeds the same logic and signal processing that's performed on high end 10gbps cards is used (so really there's little incentive to make 5gbps only network cards as any chip capable of 5gbps would also be capable of 10gbps)

 

One of the plans was to use such 2.5gbps connections  to push data into routers, and then have this split into multiple 1gbps ports and also have more than 1gbps available for wireless.

 

[Blebekblebek]

Finally there's someone understand what are they talking about, but I doubt many people would even read that wall of text.