Troubleshooting for very disappointing speeds fiber 1000

@jslow421 I shared my experiences with the same problem in this post: https://forums.att.com/t5/AT-T-Fiber-Equipment/Fiber-Speed-is-Way-Below-Advertised/m-p/5993153#M11300

The short is that for the most part I believe the speedtest on the gateway and that there are a lot of things on your devices and internal network that can be the issue.  Ideally, your device that you test with is one that you've been able to prove it can consistently get gigabit speeds (either with your prior carrier or at a different location) and it is directly plugged into the gateway with a the same cable that you were able to previously get gigabit speeds (i.e. you need a "standard" that you know works - not just in specs, but proven to work).  I realize some of these suggestion may not apply to you:

• On the gateway
  • disable IPv6
  • only attach the one device via wired connection (and disable wireless) during testing
• On the test machine
  • everything I said above about using a proven gigabit capable device (with the exact same configuration, cable, etc.)
  • use wired connection plugged directly into gateway - it is very unlikely for you to get gigabit speeds with wireless (unless you have the right router and right wireless NIC as well as the right environmental conditions)
  • turn off all power saving settings (for the CPU)  (i.e. you want to be in max performance mode)
  • if laptop, plug it in as it may use a different power setting when plugged in vs not plugged
  • if using USB ethernet adapter, you may have a limitation of the USB port or other issues
  • turn off all power saving settings on your NIC (make sure that you have any "energy saving ethernet" setting on your NIC disabled)
  • update the NIC drivers to latest (or google to see if the version that you are using could be the problem and switch to a previous version)
  • i personally think that laptop NICs are generally crap - do you have a computer with a high quality Intel full duplex gigabit NIC that you can test with?
• If you are using a router
  • google to verify maximum WAN-LAN throughput (there are sites like www.smallnetbuilder.com that test and publish this)
  • if using dd-wrt or other open source firmware, make sure it supports hardware NAT on your router (or you'll be limited to about 350 mbps)
  • google to investigate issues with your router firmware - try testing a newer or older router firmware
  • try testing using the speedtest client on your router if it exists

Appreciate the time, @n2itus . Most of this has been covered, but maybe I can be a little more clear.

As I mentioned this test was performed directly on the gateway with a wired connection which was validated to support gigabit speeds. One thing I didn't mention is I disable the wireless radios on the laptop during this process. The test was also performed on a completely UNMODIFIED gateway initially. By that I mean the installer left and I walked over with my laptop, plugged it in and ran the test from AT&T's test site. This is that same laptop and cable that I validated as supporting gigabit throughput. My assumption was that if the laptop and cable are capable of sustaining gigabit transfer speeds for a decently sized file I could safely assume a speed test would be no problem. Seems to me that throughput is throughput.

I mentioned briefly I'm not consistently getting 500mbps which would absolutely make me wonder about my hardware. Depending on the test site I can see 800mpbs+ on runs here and there. I typically alternate between fast.com and speedtest.net, but I've added the AT&T test as well since I had heard nobody from AT&T would care if the numbers were from anywhere else. Curiously the AT&T results are always the worst by far in the group.

I am currently using IP pass through with my Orbi system, so open source firmwares aren't involved (my old bricked Nighthawk is a testament to why I'm not allowed to play around with those anymore  ). The same speed test issues arise, but before anyone flies off the handle after reading that sentence, please remember that these issues were present before I had touched a single setting on the gateway.

I've seen the IPv6 thing on a couple of these threads. I've been using IPv6 for about a year on Spectrum now. If someone wants to explain to me how Spectrum is able to handle this and consistently deliver more speed than I paid for while AT&T sees speeds cut in half please let me know. I could buy a slightly longer latency, but once the DNS is resolved in what way does that impact speed? I also do not use AT&T DNS resolution on my router, but the laptop that was connected directly to the gateway would have so if resolution is the issue considered there that can be safely ruled out. Again, please remember that before using my usual DNS resolver I had run these tests on a stock gateway. Side note - I've had to say this before, but "disabling IPv6" isn't a long term solution to anything as this will in fact be a required technology in the not too distant future.

Hopefully that clears up questions on my methodology a bit?

Again thanks for taking the time.

Speed tests have so many performance variables. Everything between your client computers and the test servers matters, including the clients and servers themselves.

The internal speed test on the AT&T gateway is probably the most accurate for testing your AT&T service. I presume that the server(s) is well maintained by AT&T and is dedicated to gateway speed tests. This test includes your last mile connection, be it copper or fiber and a small portion of AT&T's internal network that is between you and their test server(s). If it says you are getting consistently close to 1 Gb/s, then your fiber service is fine.

To isolate performance factors at the far end, try testing with "speedtest.net" as you can choose from a variety of servers hosted by many organizations. Test servers are frequently very busy or the networks leading to them are congested. This will yield unusually low throughput results so it is a good idea to test against several servers and find ones that are consistently fast. Many times, the "who is this company?" servers are the best because fewer people use them. Additionally, whenever possible, use the speedtest.net app instead of testing through a browser for best results.

Make sure that your computers are fast enough. Sending very large files between them and timing the transfer is probably accurately enough. For an accurate test, install iperf on them and measure their performance against each other. If you only have one computer, it will obviously be tough for you to validate your side.

https://www.att.com/support/speedtest/
https://iperf.fr/

@johnc60Thanks for your reply and your time. It is very much appreciated.

I understand the difficulty of validating speeds with external services which is why I've used a few different items. It's not that I necessarily doubt the number the gateway provides, it's just also just a black box, so I don't have a lot of reason to trust it completely. I had hoped to achieve a similar speed via the AT&T speed test, but that wasn't the case.

I also completely understand run to run there will be variances due to various factors, but my cable connection on the same tests will vary maybe 10% and the fiber connection will vary 50% or more. Again my concern here isn't the raw numbers necessarily so much as the inconsistencies that I simply do not experience with my other provider running these same tests.

I have numerous devices at my disposal and all are more than capable of supporting sustained gigabit transfers. Any numbers I refer to have been validated on a minimum of two systems.

I guess my question really becomes this - I've established my hardware is capable of the speeds. Sometimes the network is capable of the speeds. Is a download speed of 300mpbs, and a variation of 50% or more something I should expect to see on this service from AT&T's own speed test app and their support site? By that I mean the support site says:

What specifically does this test run on? The "Rg speedtest" would, to me, indicate that this is not factoring in my personal hardware.

@jslow421 

Quick question.....is this an Internet-only connection or do you have Uverse TV and/or Phone service being delivered through the AT&T gateway (you don't mention which gateway you have)?

Ooops, I lied, one other question.....which Orbi system do you have?

I don't know where that "Rg speedtest request" page is.  Needless to say, I don't use any of the AT&T website tools.  🙂  Let me know the URL of that page and I'll take a look at it and find out what it actually does.

The name does imply that it will trigger a speedtest on your gateway itself, which is the test that I would trust the most.  Have you performed this test?  If so, what are the results?

My comments were just to make sure you have all your bases covered on your equipment and it looks like you do.  The 1 Gb/s rate is only consistently achievable when measured on your fiber link.  Even using AT&T's own test site (speedtest.att.com) does not yield consistent nor fast results.  For example, I have the Internet 1000 service.  I can consistently at any time achieve 920+ Mb/s up/down to a few specific "speedtest.net" sites and also on my gateway's internal test.  However, I have never been able to get anything close to that on AT&T's site and on other "speedtest.net" sites.  I have not looked into cause; I didn't care because other tests are fine so I know that my 1000 M service is up to specs.  I can did deeper into it for the general good, I suppose.

The first step is to verify that you are getting close to 1 Gb/s on your fiber connection.  The internal gateway test is a good measure.  Can you post that result?  You should perform many tests in a row, one after the other, because each test only sends two seconds of data.

Curiosity got the best of me so I took a peek at the AT&T diagnostic tools at "ufix.att.com".

The "Slow Browse" check actually reaches out to my Arris gateway at home and invokes a download-only speed test; I watched it in real time.  Here are the results (see screen shots).  I did four consecutive tests and each time it reported that "the internet speed you are receiving is less than the speed you are subscribed for."  Meanwhile, on my gateway, the results of the four tests are logged with each showing over 960 Mb/s, which is about as good as one is going to get.  AT&T is doing themselves a great disfavor by the inaccurate assessment on their web diagnostics tool which I am sure is causing unnecessary tech support calls.

I also compared the test results of "www.att.com/support/speedtest/" and "speedtest.net".  The AT&T test was always substantially lower and varied quite a bit from test to test.  The "speedtest.net" results were always higher and more consistent and across several of my favorite servers.  I did not do detailed analysis on AT&T's testing methodology but I suspect that their servers are a bit under powered or their network connection to the servers is getting congested by thousands of AT&T customers doing speed tests.  I wonder how much of the overall Internet traffic is just speed tests?  🙂  Even on "speedtest.net" some servers were slower than others, particularly those that are further away.

The take-away is that AT&T speed test results and assessments can't always be trusted unless one understands how the test is being performed and what seemingly uncontrollable factors are influencing the tests.

I am interested in what your gateway speed test results are.  If they are in the 900 Mb/s range then you are getting what you are promised, as far as your fiber connection goes.  Then, we move outward and start narrowing down where your slowness comes from.  It may be as simple as you being in a busy AT&T area and their network infra is congested.  I'm not certain how far their speed guarantee goes; I'm guessing just on the circuit to your premises.

@tonydiIt's a BGW-210 and just internet, no other services beyond internet. Orbi system is an RBK50, but tests are performed both on and off this system and with/without it even attached to the gateway so aside from offering up the thought that I'm relatively happy with it but do have a few gripes there's not a lot relevant to share.

@johnc60Yes that's the test I'm speaking of. I was able to verify by watching network traffic that it is indeed not browser based as you said. What bothers me most is that I've never seen a single test on a multitude of servers exceed 800mbps. This isn't necessarily a problem in itself if I wasn't also regularly getting 300mbps. I mention this because you said "I can consistently at any time achieve 920+ Mb/s up/down to a few specific "speedtest.net" sites" and I've never hit close to that on a single one (in fairness I've only tested the 10-15 closest, and my fastest tests have never been remotely consistent. Also your admittedly "slow" AT&T speed test blows any test I've ever run to that one out of the water.

I had replaced the gateway last night at was running some tests about 1am local time which I would have assumed would be ideal for lower traffic. It was not. Fiber 1000 has only been available in my neighborhood for maybe a month or two tops.

My current issue is that my anecdotal tests indicate an inconsistency/problem, AT&T's internal help test indicates a problem. The only thing in the world that indicates there isn't a problem right now is the gateway's own internal test. I certainly don't disagree that all testing is inherently iffy. If someone from AT&T could provide me with the IP the gateway's internal test goes to I could follow the paths and determine where there might be a slowdown.

My issue really is that my previous service was consistent. It was consistently faster than what I was paying for but I don't expect that. I would be happier even if it's only consistently in the 600-700mbps range, and I'm not getting that. I get 300mbps about as much as I get 600mbps.

If the answer is that this is what happens with the service that's fine. I'll just run my year out and go back to my previous provider. I guess I just expected more.

In any case thanks for everyone's time. It is much appreciated as always!