Unstable Internet Connection

[HeartBreakerTV]

Alright, so I moved into this house and have been trying to stream to Twitch (as I've been doing for years) with a 300down / 30up connection. It has been awful. We have the most unstable internet known to mankind. I used to have a more stable connection on 60/6. It doesn't matter what I set my bitrate to..I've had it coast at 6k for minutes and then drop to 38kbps...I've set to to 3k and it'll run and then drop to 1k. With a 30up, even if I'm not getting it all..that's a TON of head room for a 5k bitrate..25mb of wiggle room..yet somehow I lose basically all of it.

I've tried a laptop and a PC. I've tried ethernet and wireless. Ethernet cat-5 and cat-6.

We've had 3 COX technicians out (they've replaced everything in the house and all the way out to the street), been on the phone/chat/Twitter Dm's with support for a month. They finally said oh, we know what it is. We have a bad node somewhere else. Well, they fixed that today..and although the situation is different..it's not good.

I was getting 2% packet loss randomly prior to the node fix. Haven't tried to test that since.

This is just now running "cmd /k ping google.com /n 100": Packets: Sent = 100, Received = 99, Lost = 1 (1% loss),
Approximate round trip times in milli-seconds:
Minimum = 84ms, Maximum = 222ms, Average = 97ms

I went from just really big (and sometimes very consistent) drops..to just about nothing and would end up with like 30%+ frames dropped in OBS. Now it started off slow..and I played for about 20-30min with no frames dropped..then it went to .1%....and stayed..then 10min later .2%..and has climbed since, seemingly getting faster. It is not up to 1.8% dropped frames at 1hr and 50min of broadcasting.

Pre Node Fix: https://imgur.com/a/O9lyO

Post Node Fix: https://imgur.com/a/5q6cu

PC specs:

GPU: AMD 390x
CPU: i7 4790k (Not OC'd)
RAM: 16GB DDR3
HDD: 500gb
SSD: 120gb
Win: 10

 

[Gryz]

Keep using wired connections to do your tests. WiFi can have its own problems (where your ISP can do nothing, it's not their responsibility if your WiFi doesn't work properly). So don't get those 2 separate sets of problems mixed up.

If it's not a WiFi problem, then it's a problem in "the network". In that case only your ISP can fix it. All you can do it help your ISP locate the problem. Yes, you can help them.

Just pinging stuff is not gonna help you locate the problem. It helps seeing if there is a problem. But it won't help finding out what needs to be fixed.

Learn to use a tool called "traceroute". On windows its called tracert. There are graphical applications that do the same thing, but with nice pictures.

Pick a few places that you find important, and where you think there is a problem. Figure out their IP-addresses. Use those during testing. Some websites use varying IP-addresses, to make sure you get connected to one of their servers that is "closest by" for you. Google does that. Pinging to google.com is kinda meaning-less, because that can be anywhere. Use ip-addresses, so you know what you are testing.

With traceroute keep tracking a handful of ip-addresses. Maybe write a script to do this. (Maybe there are tools that will do it for you. I don't know). If there are packet-drops, the traceroute-output should show you where the drops are. Maybe it is a particular link in your ISP's network. Maybe it's a peering between your ISP and another ISP. Maybe it's on the 2nd hop on the way. That would mean it's the cable-network itself, that connects you to your ISP's backbone. It could be anywhere, If you don't measure it, you won't find out. And I'm pretty sure your ISP won't do such in-depth analysis for you.

If you find out exactly where packets are dropped, you can ask your ISP to fix it. They might, they might not. Good luck.

 

[HeartBreakerTV]

Keep using wired connections to do your tests. WiFi can have its own problems (where your ISP can do nothing, it's not their responsibility if your WiFi doesn't work properly). So don't get those 2 separate sets of problems mixed up.

If it's not a WiFi problem, then it's a problem in "the network". In that case only your ISP can fix it. All you can do it help your ISP locate the problem. Yes, you can help them.

Just pinging stuff is not gonna help you locate the problem. It helps seeing if there is a problem. But it won't help finding out what needs to be fixed.

Learn to use a tool called "traceroute". On windows its called tracert. There are graphical applications that do the same thing, but with nice pictures.

Pick a few places that you find important, and where you think there is a problem. Figure out their IP-addresses. Use those during testing. Some websites use varying IP-addresses, to make sure you get connected to one of their servers that is "closest by" for you. Google does that. Pinging to google.com is kinda meaning-less, because that can be anywhere. Use ip-addresses, so you know what you are testing.

With traceroute keep tracking a handful of ip-addresses. Maybe write a script to do this. (Maybe there are tools that will do it for you. I don't know). If there are packet-drops, the traceroute-output should show you where the drops are. Maybe it is a particular link in your ISP's network. Maybe it's a peering between your ISP and another ISP. Maybe it's on the 2nd hop on the way. That would mean it's the cable-network itself, that connects you to your ISP's backbone. It could be anywhere, If you don't measure it, you won't find out. And I'm pretty sure your ISP won't do such in-depth analysis for you.

If you find out exactly where packets are dropped, you can ask your ISP to fix it. They might, they might not. Good luck.

Thanks for the response. I have actually done some trace routes and found that there were issues. I sent the information over to COX and they said:

"After analyzing the data, it seems that the latency begins to experience issues starting at the 209.85.x.x ip block, which is owned by Google. This behavior seems to be mirrored on the IPv6 test you provided where the ping doubles starting at 2001:4860::c:4000:da1a (Google)."

"Thank you for the trace routes. With a traditional connection your internet traffic should not deviate off the Cox network when you are tracing to us. Are you aware of any VPNs or changes to the DNS settings?"

I, of course, haven't done anything with DNS or VPNs. We moved in..set up the new stuff and have had issues. Anything done was done by COX. I thought we were getting somewhere and then it just went back to typical "we'll send a tech out"

 

[HeartBreakerTV]

bump

 

[VirtualLarry]

OP, does your modem have a "Puma 6" (Intel) chipset? Those have known latency issues. Replace it with something with a Broadcom chipset, if you want to get rid of the problem.

 

[HeartBreakerTV]

OP, does your modem have a "Puma 6" (Intel) chipset? Those have known latency issues. Replace it with something with a Broadcom chipset, if you want to get rid of the problem.

Hey,

So I am currently using the COX Panoramic WiFi Modem/Router combo. I can't seem to find those specs anywhere.

 

[Despoiler]

The Cox Panoramic is just a rebranded Arris TG1682G. It does use the Puma chipset. Send it back to Cox and buy your own modem. I just picked up the Motorola MB8600. Great modem.

https://www.cox.com/content/dam/cox/residential/support/internet/print_media/TG1682G_ DataSheet.pdf

https://wikidevi.com/wiki/Arris_TG1682G

 

[pad23amg]

I, of course, haven't done anything with DNS or VPNs. We moved in..set up the new stuff and have had issues.
We've had 3 COX technicians out (they've replaced everything in the house and all the way out to the street), been on the phone/chat/Twitter Dm's with support for a month. They finally said oh, we know what it is. We have a bad node somewhere else. Well, they fixed that today..and although the situation is different..it's not good.

After analyzing the data, it seems that the latency begins to experience issues starting at the 209.85.x.x ip block, which is owned by Google.

Even AT&T tech has latency problems in my area. This behavior seems to be mirrored on the IPv6 test you provided where the ping doubles starting at 2001:4860::c:4000:da1a (Google).
Anything done was done by COX. Even reconfigured Cox modem. https://www.corenetworkz.com/2017/12/setup-cox-internet-troubleshoot-wifi.html I thought we were getting somewhere and then it just went back to typical "we'll send a tech out

OK before buying a new modem, could you check the DNS settings set on your current modem ? Trying open DNS may be good to troubleshoot.
Or try reset your modem and reconfigure it with the help of COX tech support from scratch. If still the issue persists, you may change the modem.

 

[Gryz]

The OP is complaining about packet-loss. What is a different DNS server gonna do about that ? Nothing.

 

[Ranulf]

Heh, when I saw this thread and read OP I immediately thought its probably a 32x modem by Intel. I saw in my local Costco the other day that they are now just selling the Arris 32x Surfboard model that has been giving people so much trouble for over a year now. I was tempted to put up a sign that said "DO NOT BUY, save money and get a 16x". Maybe they've fixed it, its been awhile since I've checked dslreports.com.

Oh my, up to page 256... some light reading for you OP. Hopefully this is what fixes your issues. As Larry/Despoiler said, just get a broadcom based modem and a real router/wifi setup. The lease fees kill ya.

http://www.dslreports.com/forum/r31122204-SB6190-Puma6-TCP-UDP-Network-Latency-Issue-Discussion

Edit 2: More reading about the Puma 6/7 chipsets from Intel:

http://www.badmodems.com/

 

[sonitravel09]

There are too many reasons WiFi can be unstable so without a doubt the fix is to get wired. If you can't run wires, use powerline networking.