Weird cabling issue (Solved)

[TheLostSwede]

I was hoping to get some help here.

We've just done a bunch of work on the house and changed some wiring around.
I have an un-marked Ethernet cable (seems to be Cat5e) from the first to the second floor, which was installed by a guy from the cable company that we have our internet service with.
It has been a bit finicky, as it's a stiff cable and he just terminated it in two RJ45 ports.

I just changed the port on the first floor to a socket and I'm now having performance issues, for apparently not reason. The socket I used it rated for Cat5e and I followed the T568B wiring standard and the RJ45 connector on the other end is also T568B. I will change this to a socket as well at some point.
I've lost 30-40Mbps in download speed and 10Mbps in upload speed.

In the same wall mount box, there's another RJ45 socket and a Coaxial socket for the cable modem. I would assume neither of these should cause any interference?
I should point out that the router sees the connection as 1000Mbps, but for some reason I don't get the full speed out on the internet.

Any suggestions on what I might've done wrong or what might cause this issue?

 

[ColinTaylor]

Connect a PC directly to one end of the cable (presumably your router is connected to the other) and then monitor the NIC statistics as follows:

Windows:

C:\Users\Colin>netstat -e
Interface Statistics

                           Received            Sent

Bytes                      40787128         4209980
Unicast packets               46500           30200
Non-unicast packets           34324            1740
Discards                          0               0
Errors                            0               0
Unknown protocols                 0

Linux:

# netstat -i
Kernel Interface table
Iface      MTU    RX-OK RX-ERR RX-DRP RX-OVR    TX-OK TX-ERR TX-DRP TX-OVR Flg
eno1      1500 15205642      0   6908 0      11447656      0      0      0 BMRU
lo       65536      205      0      0 0           205      0      0      0 LRU

If you're seeing send or receive errors there's a problem with the link. Maybe the UTP cable has been laid next to a mains power cable.

 

[TheLostSwede]

Thanks for that, no errors, all looking good in that sense.
I really can't figure out why this happened. I mean, sure, the RJ45 socket could be bad, but this bad?
Going to hook up my laptop at the other end and do some proper tests tomorrow.

Interface Statistics

                           Received            Sent

Bytes                    1169125608      1337598384
Unicast packets            10410088         5236872
Non-unicast packets         3854208          217864
Discards                          0               0
Errors                            0               0
Unknown protocols                 0

 

[ColinTaylor]

I think that pretty much rules out a cabling problem then. It was a bit of a long shot. Ethernet either works or it doesn't, it doesn't "slow down".

 

[TheLostSwede]

That's what I thought as well, until today...
Running speed test against various servers, I'm as I said, 30-40Mbps slower in terms of download and 5-10Mbps in terms of upload.
If I connect at the first floor, I have the same speeds as I had before, i.e. 200-210Mbps down, 28-29Mbps up.
However, on the second floor, the speeds now never exceeds 177MBps down and 22Mbps up.
Same cable, the only thing I changed was the connector at the ground floor...
I even re-did it, to make sure I hadn't messed something up.

 

[degrub]

set up a couple PCs with iperf. Run one as server the other as client. Test directly between the two connections only ie PC to PC with no equipment in between. That will confirm your link rate up to the speed of the nics.

You probably have a bad termination in the new connector. Not bad enough to cause packet loss, but an issue with link rate. i would re-terminate with a different connector.

Otherwise, first power cycle your router/switches and see if it changes.
You might have a weak cable between the outlet and the PC. Unless you happened to be going over AC wireless and didn't realize ?

 

[coxhaus]

If you are getting zero errors as stated above then look elsewhere.

 

[TheLostSwede]

set up a couple PCs with iperf. Run one as server the other as client. Test directly between the two connections only ie PC to PC with no equipment in between. That will confirm your link rate up to the speed of the nics.

Otherwise, first power cycle your router/switches and see if it changes.
You might have a weak cable between the outlet and the PC. Unless you happened to be going over AC wireless and didn't realize ?

That was the plan for tomorrow.

Tried that, changed the cable between the router and wall jack as well, twice...

I have no idea what's inside the walls where the cable runs, but the cable itself hasn't changed in any way, except the RJ45 connector on one end, so I don't think that AC power lines is the issue...

 

[Samir]

If all things remain the same except the one change you made, then I hate to say it, the problem was with the one change you made.

The stiffer wire is solid versus stranded and is designed for structural installs like you have. Just don't bend it too much as it can break inside the jacket if there is too much stress.

 

[Smokindog]

Probably me but I'm still not clear exactly what you have. I THINK you had a cable run that had RJ45 connectors making it a patch cable from one floor to the other. IF you connected it to your PC on one floor and your switch/router on the other you got acceptable speeds. You wanted to "clean things up" and terminate that cable into wall outlets and use a patch cable from one wall outlet to your switch/router and another patch cable at the other outlet to your PC. Unfortunately doing this resulted in your speed suffering.

Is this correct?

If so, questions

1. Is the in wall cable solid core or stranded? I see Samir suggests solid but I'd like for you to confirm that. It's NOT recommended to use keystone jacks with stranded wiring. Also, there are different RJ45 plugs for solid and stranded. YES, it can and usually does make a difference.
2. Can you see enough of the wire in the wall to see if it is CAT5, CAT5E, CAT6, or other? You've added 4 more connections in your path so if the connection was marginal to begin with you may have pushed some limits through these connection losses.
3. Are you using keystone terminations or screw down?
   
4. Are you using "home made" or factory patch cables? If home made, see comment on different plugs for different wire types. Also, check the patch cable direct between PC and switch/router for your speed targets.
5. There are different keystone jacks for the different wire terminations (5 vs 5E vs 6). Make sure you've got the right ones ("better" is OK). If you're using screw down, switch to keystone (unless stranded and then expect to take a hit).
6. As suggested, if all else fails go back to one change at a time. Re-terminate the in wall cable with RJ45 plugs, test for your original speed and begin at the PC end replacing the connections one at a time testing after each change.

That was the plan for tomorrow.

Tried that, changed the cable between the router and wall jack as well, twice...

I have no idea what's inside the walls where the cable runs, but the cable itself hasn't changed in any way, except the RJ45 connector on one end, so I don't think that AC power lines is the issue...

 

[ColinTaylor]

@Smokindog I think we can eliminate all of those points as the cause because of the information in post #3. He can confirm that when he does his end to end iperf test. No need to start changing the hardware unnecessarily.

 

[coxhaus]

If the wire was at fault you would see errors in the stats. If you really want to know then rent a fluke tester to certify the cable.

 

[Smokindog]

I'm sorry but I don't see post 3 answering any of my questions and I only suggested going back to square one if nothing improved.

I also see no statement indicating what the link speed is for the connection.

I'll stand by by comments/questions for now unless someone can explain exactly what the problem actually is!!!!

@Smokindog I think we can eliminate all of those points as the cause because of the information in post #3. He can confirm that when he does his end to end iperf test. No need to start changing the hardware unnecessarily.

 

[ColinTaylor]

I'll stand by by comments/questions for now unless someone can explain exactly what the problem actually is!!!!

"I've lost 30-40Mbps in download speed and 10Mbps in upload speed."

I also see no statement indicating what the link speed is for the connection.

"I should point out that the router sees the connection as 1000Mbps, but for some reason I don't get the full speed out on the internet."

 

[dosborne]

My money would be on an issue with the wiring.

I would definitely get a proper hardware level cable analyzer to check things out. The original cabling may not have used the same pairs the way you think it did ( ie may not have been per the specs you used). Crossing a couple ground wires could be enough to cause crosstalk although that *should* show up as resent packets.

But, this (connectors) was essentially the only change, according to your statement. You have to start with the basics,which is the physical layer in this case.

 

[Samir]

It because if issues like this I usually leave wiring alone if it's working.

 

[TheLostSwede]

Probably me but I'm still not clear exactly what you have. I THINK you had a cable run that had RJ45 connectors making it a patch cable from one floor to the other. IF you connected it to your PC on one floor and your switch/router on the other you got acceptable speeds. You wanted to "clean things up" and terminate that cable into wall outlets and use a patch cable from one wall outlet to your switch/router and another patch cable at the other outlet to your PC. Unfortunately doing this resulted in your speed suffering.

Is this correct?

If so, questions

1. Is the in wall cable solid core or stranded? I see Samir suggests solid but I'd like for you to confirm that. It's NOT recommended to use keystone jacks with stranded wiring. Also, there are different RJ45 plugs for solid and stranded. YES, it can and usually does make a difference.
2. Can you see enough of the wire in the wall to see if it is CAT5, CAT5E, CAT6, or other? You've added 4 more connections in your path so if the connection was marginal to begin with you may have pushed some limits through these connection losses.
3. Are you using keystone terminations or screw down?
   
4. Are you using "home made" or factory patch cables? If home made, see comment on different plugs for different wire types. Also, check the patch cable direct between PC and switch/router for your speed targets.
5. There are different keystone jacks for the different wire terminations (5 vs 5E vs 6). Make sure you've got the right ones ("better" is OK). If you're using screw down, switch to keystone (unless stranded and then expect to take a hit).
6. As suggested, if all else fails go back to one change at a time. Re-terminate the in wall cable with RJ45 plugs, test for your original speed and begin at the PC end replacing the connections one at a time testing after each change.

1. Solid core
2. No markings on the cable, oddly enough. As I mentioned, it was installed by the cable TV company a few years ago, but it's not that old and has worked at Gigabit speeds and still syncs at Gigabit speed.
3. Factory, tried three different ones.
4. 5e keystone
5. That would mean I have to go out and buy some tools, as I don't have an RJ45 crimping tool, nor RJ45 connectors.

 

[TheLostSwede]

It because if issues like this I usually leave wiring alone if it's working.

Normally I would've, but the installation was really ugly with three cables sticking out of the wall in the living room, something I didn't want to have after having made the rest of the place look really nice and tidy.

 

[Smokindog]

Just for grins, can you test each of those 3 patch cables you have directly between your PC and your switch/router and check the speed? I know you said the entire setup still syncs Gb but I'm pig headed to test each piece separate, especially when something awkward is afoot

Also, can you see the wire pattern in the factory cables? Can you tell if they're "A" or "B".

Normally I don't worry about it but this case doesn't seem normal. I ran into a similar issue over at my neighbors awhile back. Real head scratcher in that the connection worked fine but periodically just froze until the NIC was rest at the computer. After watching the cable tester for awhile I saw what I thought was a periodic "two LED flash" within a pair, the second LED being VERY faint. We then back tracked recent events and determined that this all began when they had some work done in one of the rooms. Found a nail barely nicked the wire in the wall and periodically would cause issues. Your issue is hopefully contained to the two outlet boxes!

Anyway, it could be a bad punch/crimp at one of the keystone jacks or something like that. How did you install the wire into the connector? Many of the keystone jacks give you a plastic punch down tool and they can be problematic. Recheck those connections carefully. Sometimes a small flat blade screwdriver can be used to better seat the wires in the channels. Just be careful of the keystone and your fingers Look for possible points of cross connect like excess wire extending from the keystone punch points. They can be finicky.

I really encourage you to get at least a cable tester and crimper tools similar to what I've linked but I understand it's not in everyone's budget. I'm not endorsing these products, just using for example and they look similar to what I've had/used in the past.

https://www.monoprice.com/product?p_id=38029

https://www.monoprice.com/product?p_id=3350

https://www.monoprice.com/product?p_id=7041

1. Solid core
2. No markings on the cable, oddly enough. As I mentioned, it was installed by the cable TV company a few years ago, but it's not that old and has worked at Gigabit speeds and still syncs at Gigabit speed.
3. Factory, tried three different ones.
4. 5e keystone
5. That would mean I have to go out and buy some tools, as I don't have an RJ45 crimping tool, nor RJ45 connectors.

 

[TheLostSwede]

No need, I can see issues running iPerf3, I peak out at sub 900Mbps client to server (server upstairs) and doing a reverse test I end up with peak speeds of 300Mbps, so something is going on here.
Going to have to get some new sockets and some tools... Worst case I'm going to have to try and run a new cable

I used a cheap punch down tool, but with a metal part for the punching.