Question Weeks later, still having odd Internet issues

Oxicoi

Honorable
Feb 7, 2017
426
13
10,815
8
Hi,

So as title says, my Internet is being somewhat stupid. Almost all non-portable devices are connected via Ethernet while phones and other devices are wirelessly connected. We have 1000mbps fiber internet, so having multiple devices shouldn't be a problem.

When it comes to games, TV's, computers, phones, etc., the speeds are perfectly fine. However, there is random packet loss, a SIGNIFICANT delay in a lot of day-to-day activities. Most of this is on gaming, but our brand-new Samsung smart TV can't even connect wired and bugs out with a blank screen.

ATT is our provider. I know many don't like this provider, but it's the only one affordable in our area. The other is Cox, but they always increase price after your contract, so we left that long ago. Arris BGW210-700 is the router they gave us and is on IP Passthrough to a Netgear Nighthawk XR500, which again, should be plenty. We also have a plug'n'play ethernet switch as we have a lot of devices.

Here is a packet loss test and to me, it's weird.




Is this normal? Shouldn't it be consistent, not up and down?


Testing a ping test to google and noticed this:


Using Ping Plotter, it shows a really bad latency issue (if i'm correct):
 
Last edited:
Your test results show almost a perfect connection so you must be imagining things, ticket closed. :)

How are you seeing the packet loss. Do you know if it is only to say games or other applications. I would do the standard leave constant ping run to various IP addresses in the background and then quickly check when you think there is a issue. I would start with ping to the router IP and a common ip like 8.8.8.8

I have not seen that router but the newer one has a massive number of information. If it has the ability I would a speedtest on the router itself. This one should show very close to the full gigabit up and down. ATT somehow rigs it so it tests to a really fast server. Also check the signal levels on the fiber. It should actually tell you if
the numbers are within the recommended range.

What you kinda need to hope is that you are getting packet loss in your house, very unlikely, or on the fiber itself. If you for example only get loss to riot games but not steam it will be very hard to get fixed.
 

Oxicoi

Honorable
Feb 7, 2017
426
13
10,815
8
Your test results show almost a perfect connection so you must be imagining things, ticket closed. :)

How are you seeing the packet loss. Do you know if it is only to say games or other applications. I would do the standard leave constant ping run to various IP addresses in the background and then quickly check when you think there is a issue. I would start with ping to the router IP and a common ip like 8.8.8.8

I have not seen that router but the newer one has a massive number of information. If it has the ability I would a speedtest on the router itself. This one should show very close to the full gigabit up and down. ATT somehow rigs it so it tests to a really fast server. Also check the signal levels on the fiber. It should actually tell you if
the numbers are within the recommended range.

What you kinda need to hope is that you are getting packet loss in your house, very unlikely, or on the fiber itself. If you for example only get loss to riot games but not steam it will be very hard to get fixed.
It would be odd to say if every game has packet loss. I don't want to say that and be like "well it's the games then" and call it a day. Although that is exactly the case, it just doesn't make sense to take all of these tests and then nothing be wrong. I added another test with Ping Plotter and it shows a really bad latency figure. I can't say if what I'm seeing is true, but 100ms+ of a latency just to google seems a bit absurd.

Maybe the nighthawk router just isn't worth the hassle? Sometimes I wonder if putting ATT's router on IP Passthrough makes it worse and slower than what it should be, and doing so on purpose. However, the nighthawk has UPnP, ATT does not. We have multiple devices that game, so UPnP is kind of necessary because you can't port forward 1 singular port to multiple devices.

It's just such a huge mess. Back in the day, I could have little to no problems since it was easy to edit these types of things. Now, ISP's just control so much with little freedom to give to the customer. It's super annoying.
 
You have to be very careful about using tools like this without fully understanding how they work.

Issues in intermediate nodes that do not actually affect the end traffic are not really valid. Think of it this way. Say instead of routers these were street lights on the way to work and you measured the delay to each. If one was broken then it would also make you later to work. It can't delay you and allow you still get there with no delay.
This could be very random delays as the data is being tested but much more often it is that the router delayed responding for other reasons. It is very common for routers to delays responding to ping/trace so they can use their cpu to pass actual traffic. Also many times routers restrict the rate they will respond to avoid denial of service attacks.

In the end you pingplotter also show you have no issues. A bad one would show say a delay of 100ms in hop x and at least 100ms if not more in every hop past there including the last hop.

The ping you posted shows normal variations in the test. You can not measure variations that low with ping and trust it is a real value. Actual problem will show 100ms or even 200ms spikes. Most times you can not even detect stuff less than 100ms in a game.

You are going to have to continue to test to try to find the problem. If you still can't detect it with a simple ping command you need to start thinking the game is telling lies. Many games are really stupid, they will get stuck say processing a complex video frame and when they finally get around to looking for the response to their ping they will blame the delay on the network when the data was sitting there all the time just waiting to be read out of the buffer.

Att is indeed strange when it comes to how they implement their so called "bridge" mode. From other tests I have seen on the newest att connections that are claiming 5gbit you see people passing well over 2gbit even in passthrough mode. So far I have not seen someone test that has equipment that has 5g ports and att has just started to role out this faster offering. I guess you could turn it off and connect directly to the att router and see if this packet loss persists. You could I guess also turn off just the passthrough part and let it run router behind router to test. Many games do not need/use UPnP so you could test on those games.
 

Oxicoi

Honorable
Feb 7, 2017
426
13
10,815
8
You have to be very careful about using tools like this without fully understanding how they work.

Issues in intermediate nodes that do not actually affect the end traffic are not really valid. Think of it this way. Say instead of routers these were street lights on the way to work and you measured the delay to each. If one was broken then it would also make you later to work. It can't delay you and allow you still get there with no delay.
This could be very random delays as the data is being tested but much more often it is that the router delayed responding for other reasons. It is very common for routers to delays responding to ping/trace so they can use their cpu to pass actual traffic. Also many times routers restrict the rate they will respond to avoid denial of service attacks.

In the end you pingplotter also show you have no issues. A bad one would show say a delay of 100ms in hop x and at least 100ms if not more in every hop past there including the last hop.

The ping you posted shows normal variations in the test. You can not measure variations that low with ping and trust it is a real value. Actual problem will show 100ms or even 200ms spikes. Most times you can not even detect stuff less than 100ms in a game.

You are going to have to continue to test to try to find the problem. If you still can't detect it with a simple ping command you need to start thinking the game is telling lies. Many games are really stupid, they will get stuck say processing a complex video frame and when they finally get around to looking for the response to their ping they will blame the delay on the network when the data was sitting there all the time just waiting to be read out of the buffer.

Att is indeed strange when it comes to how they implement their so called "bridge" mode. From other tests I have seen on the newest att connections that are claiming 5gbit you see people passing well over 2gbit even in passthrough mode. So far I have not seen someone test that has equipment that has 5g ports and att has just started to role out this faster offering. I guess you could turn it off and connect directly to the att router and see if this packet loss persists. You could I guess also turn off just the passthrough part and let it run router behind router to test. Many games do not need/use UPnP so you could test on those games.
I suppose. I'm not really new to networking in general, but the programs and tests... yes. I know games don't need UPnP, but it always helps to keep everyone on an open NAT type instead of just 1 out of X amount to find every lobby that exists for a fuller and better experience.

Indeed frustrating.
 
It is not so much a long term solution it is more to do testing and see if it is the portforwarding/passthrough. I doubt it but it is one less thing.

I would still leave the ping run to 8.8.8.8. or google.com if you want. What you need to check is when you see issues in the game if you quickly tab over to this window does ping see the same issue. This is so you don't spend huge amount of time trying to fix the wrong thing. If you see issues in the ping....note it will be a large and very obvious spike..then you can look for network issues. If it does not show issues then it would be better to look for software on your machine causing it.
 

ASK THE COMMUNITY