Trying to figure out why Internet Speed is so bad through Ethernet Over Power

Hi All,

Seeking some assistance here, as I am really not sure what is going on.

I have a 1gb connection through Fibre - my Router has a built in speed-test and this is the result at the Router.

I run a Virtual Machine connected to the Router via an SFP+ port and on that virtual machine I run Speedtest Tracker and this checks the speed every hour. These are the results.

I run an Ethernet Over Power line using the this device:

I also tested:

Both of these devices have Gigabit Ethernet, both make bold claims, the TP link is 1,300 mbps and the Devolo is 2,400 mbps.

Of course this is nonsense, we all know it is - but this is what I get at my Desktop:

image

I was expecting some degradation, but this seems extreme - 1/10th the speed?

Oddly, despite this limitation, the Devolo is still routinely better in my testing - only by 10Mbps but it is consistently 10Mbps faster.

Anyone have any thoughts on this? Any experience, some settings to tweak, things to try?

Doesn’t seem that bad looking at NETGEAR PL1200 Powerline 1200 Reviewed - SmallNetBuilder which is rated as “good”. There’s usually a lot of noise generated by devices on your circuit and wiring may also affect performance.

Interesting development…I moved the devices to a different plug socket.

The TP-Link now reads:

image

Devolo top result was:

image

So, I think it is either the socket I was using was not the best - whether that is due to it being in an extension, and there an extension of the existing electrics - or because I have like 90 devices plugged into the secondary socket next to the device I am not sure - I will test it out.

This is of course, still far less than the 2,400 Mbps advertised…

Depending on distance you might want to consider using a wireless bridge setup instead.

I think the TP Link is a wireless extender - but my Win 10 has decided that despite my device clearly being visible having used it for the last 3 months, my desktop doesn’t have wireless anymore…I will get around to fixing it at some point.

Another test I just ran.

I had the Magic 2 Lan 1 : 1 - which contains a single port, I tested that using a modern laptop near to the room where the Fibre comes in.

80 mbps down, about 140mbps up.

Which was odd, I would have expected it to be better than the results above.

So I swapped out the 1: 1 for a Triple - just like the one for the Desktop.

240Mbps down 240 Mbps up.

So there is some advice - don’t buy the 1:1 version it is crap.

Still does point to the electrics getting worse as you move further from the room where the fibre comes in.

In the grand scheme of things I shouldn’t complain…I was getting 3.4mbps at the top end - now I am getting 180Mbps at my desktop…51 fold increase…and that 180 is almost triple what I was getting simply by moving to another plug socket - I do now have a wire going around a corner which isn’t the best…but I can fix that at some point too.

I also am getting almost 1000Mbps at the NAS, so not an issue really I suppose.

So, I looked at my motherboard and realised it had 2 ports 2.5gbs and 10gbps (it made no difference to the testing btw I didn’t just discover this I knew it had two ports all along)…and thought - hey, this motherboard does teaming.

So I teamed them together for link aggregation…and absolutely destroyed the internet speeds

I have no idea why, but this is my current speed…

image

I have these two:

image

Connected together to create:

image

The details of which are:

image

Given it is a 2.5gbps and a 10gbps port, I was expecting 12.5gbps but even 2 should be fine, given the cables are only Cat 5 and Cat 6 cables.

Seems like Link Aggregation doesn’t want to work for me…

Fixed whatever was going on there - not sure what exactly - but even with proper link aggregation not actually get any faster speeds than I was getting earlier, bit disappointing.

honestly if you got working better than 100baseT that’s a win

powerline is the worst

3 Likes

Powerline is bad for the reason twisted-pair is part of the cabling engineered standards. The twists prevent crosstalk. Normal ol molex is a massive antenna. Just because 550MHz can ride along side 60Hz doesn’t mean it should lol. Like GigaBuster said, beating anything over 100baseT is a huge win.

So much shit can interfere over powerline.

Ethernet Over Power line

I made that mistake before. At best it’s going to be bad, and it’s much more likely to be awful with constant disruptions whenever your fridge kicks on or something.

This. Many of the “low end” units don’t even bother having ethernet ports that can handle more than 100baseT to begin with because the performance cannot be anything but that shitty. The absolute best speeds I’ve ever seen anyone ever report was just over 300mbps, and they may have well have been a unicorn.

Your options are basically:

  • Ethernet over coax (Need a MOCA capable router/modem, a filter to keep your signals in your house, potentially another one for the router/modem (I don’t need one), a compatible coax splitter, and converter units at the coax jacks). There may be changes coming to future Docsis versions in modems that may cause compatibility issues in the future, but I’m not clear on that.
  • Get better wireless coverage
  • Learn how to run ethernet/fiber cable

I make use of ethernet over coax for my place, though I also have a 10G fiber lan for the things that can use it (so I can work from my NAS at speeds greater than a SATA HDD)

1 Like

If you have time, you can try unplugging everything in your house from the wall sockets that you possibly can, except whatever is on two sides of your powerline and turn off all the lights, and use your laptop of phone to test the powerline performance in a few different places in your home on a few different sockets.

If you’re lucky, your wiring could be good and there could be e.g. a radio alarm clock or another crappy device with a crappy power supply making a ton of electrical noise that you can chuck in the bin. … or maybe one of the cables for e.g. a TV can be replaced with one with a ferrite loop.

If you’re unlucky, … chuck powerline in the bin and do ethernet over cat6/cat6a

2 Likes

Thanks for sharing your ideas. I’ll try this method. Lets see :slightly_smiling_face:

1 Like

https://www.tp-link.com/uk/support/faq/2928/

Explanation about the Powerline Rate and Actual speed of the Powerline Adapters

Includes such gems as:

Let’s assume the Powerline speed between the TL-WPA8630 and TL-PA8010 is 1000Mbps in an ideal environment with few interferes, normally it will make an actual throughput of 300Mbps-350Mbps. This rule basically applies to all Home Plug AV products.

Basically, 1300Mbps is little more than criminally misleading marketing.

4 Likes

There are a few things to watch out for when using Powerline:

  1. if the outlet you are placing one endpoint on is wired past a GFCI outlet then performance will not only be terrible, but that GFCI could occasionally trip randomly because of the network traffic going through it. While normal Ethernet networking is point-to-point (your computer to a switch, other computer to the same switch, wifi AP to switch), electrical wiring is connected via daisy chain. So from the breaker it hits the first wall outlet, goes to the next outlet, on and on. Normally, each room has its own breaker, or sometimes two rooms next to each other share a breaker.

  2. speeds are best if the start and end points are on the same side of the transformer power phase. US homes (well, 95% of them) have two phases of power coming in, each of them are ~120v to neutral or going between both phases is 240v. This lets you run normal 120v wall outlets and have a larger 240v circuit for the clothes dryer, kitchen oven, etc. So if one powerline adapter is on “phase A” and the other is on “phase B” then the communication between the two must go outside your house and all the way back to the transformer on the street to make the connection. Obviously, this will result in very bad performance. To check whether both devices are on the same power phase or not you have to go to your main circuit panel and count things out yourself. I was going to type it all, but the wording got too confusing so I made a very quick little picture to show you how to find the power phase each breaker is on:

  1. Your speeds might drop a lot whenever someone turns on the microwave, blender, hair dryer, stuff like that. Any appliance that has either a transformer in it or a motor. This can be disruptive when trying to stream a movie or play a game.
3 Likes

I’ll just add my $0.05 (inflation) here as well: Powerline doesn’t just suffer from interference, especially if you are crossing circuit breakers, but it CAUSES interference all throughout it’s operating frequency range.

I took powerline out of my parents’ house for this exact reason, as every time I drove up, my CB radio went wild, and my father reported issues with both AM and FM radio. Fortunately, the interference didn’t extend beyond about 100 yards, but it did peg the signal meter on my radio near the house.

And of course I had the same problems with speed, so there’s that.

2 Likes

you could try increasing the mtu rate for that network.
currently it will likely be set to 1500 or there abouts… you should be able to bump that substantially and should increase your datarates.

I’m sure it’s “working”, but Windows implements only basic 802.3ad with no layer 3+4 hashing and no MAC balancing, so all traffic between a pair of hosts with the same MAC address will go over only one link which you don’t get to choose (hence the recommendation that all the links are the same throughput). You’d need multiple routers and multiple TCP streams to see any benefit.

Most other OS’s implement MAC balancing, like balance-xor mode in Linux which does use multiple links for a single TCP stream (or any protocol).

1 Like

I think above, I mentioned that this is a dual plug socket, and in the secondary socket, I had an extension lead with about 90 other things plugged into it.

Here is an interesting thing, in the second plug, with the extension lead - I also had plugged in one of those remote control plugs - one that is connected to wifi that you can turn on and off from an App.

It also measures consumption of electricity.

I moved everything off of that secondary plug, bar 1 item, which was still plugged into the remote plug.

These are the results of that move.

image

A 40mbps jump is nothing to sneeze at.

Then I was looking at it thinking - hmmm, what if that Remote Control plug is the problem? - so I took that out…

image

Damn, that is a big jump from 70Mbps…

So, I think I narrowed in on at least one issue that I should probably have thought about earlier - that thing connecting to the Wifi - seems to have messed with the EoP.

Still far short of the rated amounts, but this is a significant improvement over what it was before.

1 Like

I tried this, but variance in testing is soo wide that it actually becomes difficult to tell whether any difference has been made.

For example, I was getting between 152-155Mbps after doing this - but then I just tested again and got 131-138Mbps.

I will say that I think upload has been positively impacted, as my uploads jumped by a good deal - currently about 195Mbps.

Also think that Intel 2.5gb is better than 10gb Marvell - but that is probably because I am not getting enough bandwidth to really gain any benefits of the 10gb port

1 Like

Umm… Romex?

I’d be happy with internet backbone speeds…

1 Like