TomMN

Accuracy..

4 posts in this topic

Hi, I am a subscriber to Comcast cable's resedential internet service in Minnesota and I believe my tests here to be innacurate. I'm suspecting routing/distance issues, although I have tried testing with both the Dallas central server and the CDN push alternative route during 'optimal' load.

 

Both speedtest.net and Comcast's speed tests rate my connection at a steady 62 megabits download and 11.8 up. Shaperprobe detects no shaping on my uplink and 6000-7000 kilobit shaping on my downlink, rating the average speeds to be 11125 kbps up and 55000 down. I have calculated my speeds during FTP transfers of multiple large files on both my downlink and uplink after 2-3 mins, reflecting these same results. Pingtest.net shows no packet loss and minimal ping/latency (45ms/2ms) to an Ookla-hosted server located in Dallas.

 

However, when testing here I am averaging 38Mbps down and 7.6 up. Why is this?

 

 

Share this post


Link to post
Share on other sites

Hi, I am a subscriber to Comcast cable's resedential internet service in Minnesota and I believe my tests here to be innacurate. I'm suspecting routing/distance issues, although I have tried testing with both the Dallas central server and the CDN push alternative route during 'optimal' load.

 

Both speedtest.net and Comcast's speed tests rate my connection at a steady 62 megabits download and 11.8 up. Shaperprobe detects no shaping on my uplink and 6000-7000 kilobit shaping on my downlink, rating the average speeds to be 11125 kbps up and 55000 down. I have calculated my speeds during FTP transfers of multiple large files on both my downlink and uplink after 2-3 mins, reflecting these same results. Pingtest.net shows no packet loss and minimal ping/latency (45ms/2ms) to an Ookla-hosted server located in Dallas.

 

However, when testing here I am averaging 38Mbps down and 7.6 up. Why is this?

 

Hi TomMN, sorry for the long reply... maybe you'll stick with it and read it, maybe not.  I just want you and others to understand why the results here can differ.  Sometimes I start writing and don't realize how much I've written... 

 

First, where you see optimal load... it's going to take about 5 times my current traffic to get that to say anything else. It's programmed to say other stuff but the home server is so over powered right now it's going to be a while before it will have enough load on it to make that say anything else.  Even then it will still have more than enough resources.  Just a heads up... it's always optimal.  If it ever gets close I always upgrade my hardware and networking. I put that there in case the server spirals out of control but this new server is the most stable I've ever seen.  Pushed a few million visitors through it so far and I haven't rebooted it yet... no signs of needing to. Can't beat RHEL and a really good hardware combination.

TestMy.net is it's own benchmark. Look at the results here differently than you do other speed tests. TestMy.net by default tests a single thread, multithread testing options will be available in the future. I'm going to offer the option to multithread but the default will remain the same because I believe it's a better benchmark as it uncovers connection issues that a multithreaded speed test results mask.

 

Multithreading, in case you don't know, is where you open multiple connections, and combine the results.  Your connection should NOT need to do this to pull big numbers.  Look at my signature... many of my tests are taken from THOUSANDS of miles from the servers. If you want to REALLY benchmark your connection, you need to test it under real world conditions not best case senario.  TestMy.net is by design NOT a best case senario... the Internet doesn't operate under best case senario... 

 

Your provider has too much at stake, especially someone as big as Comcast. If they do the test in a certain way and display the results in a certain way they aren't really lying.... look at my TiP data with each download test.  That shows you clearly that it's ALL in how the numbers are interpreted.  I could show you any of those numbers and call it my result.... what I do with TestMy.net is show you EVERYTHING taken into account in your final logged number.  

 

There are lots of things that can cause your result to drop like that.  We have a 2006 Macbook that could only get around 30 Mbps on TestMy.net, on a 50Mbps cable connection. The connection was perfectly fine... I swapped out the HDD for a SSD, rebooted and was pulling over 50Mbps.  TestMy.net is weird sometimes... I designed it and it baffles me sometimes how good it is at it's job.  Trust me... you're getting that score for a reason.  You might be pulling your speed under the right circumstances but you're not truly running at 100%. If you fix the issues causing that it will make your browsing experience snappier.

 

Having said that, that's not a bad score.  It could be hard drive performance slowing you down... TMN is the only speed test that I know of that is effected like that.  As you use your browser things are being written to cache... if your HDD is under performing then that process is slowed.  If it can be changed and it effects the results... it's a variable that needs to be accounted for.  TMN does a good job of cluing you into stuff like that where others fail.    ... maybe they do that on purpose to cut down on service calls, I don't know.  I just know that TMN time and again shows a problem where others don't.  

 

Just makes me wonder about their motives sometimes.  Or maybe their test just sucks.  They have it built by someone else so they can always deny any purposeful wrong doing later.  Some of the things that they don't catch seem like simple fixes, if the call center techs had the right tool at their disposal they'd be able to better help the customer.  But many of these ISPs are forcing their call center reps to use their test and shoo away all others.  THEY ARE WRONG.   ... providers that sent tens of thousands of their customers here now tell their customers not to use it. Research deep in the history of this forum, you'll see their customers. "A ____ support rep sent me here." "_____ tech support said this was THE place to test."  ... stuff like that. The core principals of TMN have NOT changed.  What changed is that they got their own speed test... that often makes things look better than they really are.

 

Real techs don't use flash speed tests.  They can't be trusted.  TMN is 100% HTML5, PHP and Javascript... 100% native browser support across the board using only your Internet browser.

 

Something else to consider :: The majority of speed tests out there, including Comcast's test, run Ookla software.  Ookla was founded by Mike Apgar, founder and former CEO of Speakeasy... an Internet provider. I'll leave it up to you to decide who that test was designed to benefit.  If it was the consumer... it would be catching the things that TestMy.net catches.   :wink:

 If you're on Windows, try TCP optimizer from Speedguide.net -- we get lots of positive feedback on that, for MANY years. Most operating systems don't come optimized out of the box for connections like yours... if you want to really tune your performance you need your MTU set at 1500 on cable.  If you used to have DSL it may be set at 1492 which will degrade your performance on cable (and vice versa).  Something else that TMN detects... where others seems to fail to notice.    ... it can REALLY degrade your performance too.  Symptoms of MTU or RWIN settings set wrong will be a speed lock. I often see it displayed as between 8-10 Mbps on connections that should get 2X higher speeds.  If this is the case you will see the same approximate speed or lower to all servers.  With you running 38Mbps... it might be something else.

 

... by the way, Comcast is one of many ISPs that my host has direct peering with. So you should have a pretty direct route to all my servers. Try the server in Washington D.C. and Seattle WA... see what you get there

 

TestMy.net should be a tool in your arsenal, use it as a clue.  If you have an apparent 60 Mbps pipe into your house but can only pull 38Mbps, there is a reason.  Routing, client end settings, router issues, modem issues... we see it all.  One thing is always the same, TestMy.net is correct in its assessment.  I guarantee you that if everything is perfectly in order you'll pull your speed.  Trust me Tom... this isn't my first rodeo.  10 years before Comcast started sending their customers to their own speed test... their customer service reps were sending people here.  The truth is, I don't make them look as good. But I don't care, I didn't build this for the ISPs. I built it for you.

 

 

Related ::

TestMy.net Bandwidth Test Legitimacy

Why Do My Results Differ From Speedtest.net / Ookla Speed Tests?

rekyle and TriRan like this

Share this post


Link to post
Share on other sites

Thank you for such a comprehensive response.

 

I am testing via Win7/Firefox 20 on a SSD with 8gb of ram and an i7 cpu, so I doubt my computer is the bottleneck.

 

After running TCP Optimizer and switching to the Washington server, my latest download result is 27.2 Mbps; about a month ago a comcast technician deemed my signal quality to be 'excellent', so as you can imagine this issue is quite perplexing to me.

Share this post


Link to post
Share on other sites

Thank you for such a comprehensive response.

 

I am testing via Win7/Firefox 20 on a SSD with 8gb of ram and an i7 cpu, so I doubt my computer is the bottleneck.

 

After running TCP Optimizer and switching to the Washington server, my latest download result is 27.2 Mbps; about a month ago a comcast technician deemed my signal quality to be 'excellent', so as you can imagine this issue is quite perplexing to me.

 

Yeah, with SSD and i7... you're right, your computer probably isn't the bottleneck. Although, I've seen setups just like your where hardware was newer... sometimes ends up being the network card or the ethernet cable.  If you have access to a secondary control on board, have an extra card laying around or if you have a USB ethernet to try I suggest giving a different NIC controller a try.

 

Also, do you have a DOCSIS 3.0 modem?  I assume with the package your on that they'd make you get a 3.0 modem, right?  We've also seen where the cable techs have deemed a line to be okay... scores suffer here, after talking with us the customers exchanged their modems and BAM, issue resolved.  Again, no other speed tests seemed to detect that.  I have no idea why other speed tests don't see some of this stuff.  I have a feeling it has to do with multithreading or the fact that those tests ignore the worst of your results.  ...how the hell is it a test if a large portion of the results are omitted.  :uglystupid2:   ---  I don't know how the other guys do it... I just know that a large percentage of the overall posts to this forum are people asking the same questions as you are.  People always blame me, they think that my system is broken or something. After explaination the people who choose to dig deeper often write us back to let us know that in fact there was a problem that was undetected by other speed test.

 

Research old threads here, talk to veteran users of TestMy.net... they'll all back up what I'm saying.

 

Good luck.  Like I said, if it's not bothering you I wouldn't trip out on it.  That's not a big red flag score.  Don't let your TMN score frustrate you unless your Internet is frustrating you.  Now if you were reading 10Mbps, I'd be more concerned about it effecting your browsing.  Having said that, if it was my connection.... I'd crack out on it until I pulled perfect speeds, at least to the closest server.  But I'm OCD so it would bug the crap out of me until it was right or as close as I could get it.

 

Check this out, my own connection is supposed to be 150 Mbps down... I often can only pull 50 Mbps.  Sometimes as high as 95 Mbps on TestMy.net.  I test other places and always pull 90+ Mbps. I download files and sometimes see 90+ but more often around 60.  I have NEVER seen close to 150 Mbps, anywhere. If I multithread I pull around 100Mbps. Sometimes you get what you pay for, sometimes you don't.  But TestMy.net as a tool helped me to pull the most out of what they gave me. In the past, with the same provider, I often scored well over my quoted package.  This is the first time with

Cox Communications that I wasn't given more than I pay for.  

 

I'm not making a big deal about it, I'm way happy with my speeds... I just know that there is more potential, but fixing the issue isn't always within your control.   :tickedoff:  :shrug: 

 

If you figure it out and get improved results please let me know what it was.  ;)

 

- Cheers

- D

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now