Sign in to follow this  
Followers 0
CA3LE

Google fiber only getting 96 Mbps down 4 Mbps up

1 post in this topic

Hello, I have consistently used your site, but am now having an issue and wanted your opinion. I within the last week got the 1gb up and down of googlefiber, when I run the speedtest from your site, it only comes up with 92mbps down and 4.3 up, when I use speedtest.net, your competitor, it shows 926mbps down and 946mbps. Can you help me understand this? Thank You

 

Hi Troy, this email is going to be a little long but I want to explain a few things.  Hopefully you read it.

 
This is one of the most common questions I get.  There are many reasons why speedtest.net reads so high.  First read Why do my results differ from speedtest.net / ookla.  They admit in their own wiki that they alter the results.  They ignore the worst 30% and fastest 10% because flash protocol is an inaccurate testing platform that they must make adjustments for.  How can that be accurate?  You can't alter the results!  That's not a test in my opinion.
 
They also by default test you off the closest server.  Often within 50 miles.  This usually eliminates many factors that you need to test to get a full picture.  How can you test your Internet if you're not really reaching out to the Internet.  On TMN I could easily incorporate testing off the closest server... but I'm not here to make you feel good, I'm here to try to help you get the most out of your connection.  My competition is out to stroke egos... people feel good and share their results, "LOOK HOW FAST I AM!!" - bringing them more traffic and perpetuating the misconceptions.  They make it seem like you need to test close to you to be accurate.  The only reason I offer mirrors on TMN is for route comparison.  I believe that the best test is actually at distance.  Longer distance puts your hosts intermediary connections into play... testing how well they peer.  Let's be real, the sites you visit are likely to be pretty far away.  My servers are hosted in very popular Internet crossroads that any decent provider should be able to provide clean connection to... areas where millions of sites are hosted.  
 
Another reason may be multithreading.  You most likely can only get closer to gigabit speeds by opening more than one thread.  If you download lots of torrents or off usenet then that's not a problem because that's how those files download (split up and multithreaded and recombined later).  When you normally download a file, it's a single thread.  In my opinion you should always try to benchmark using my normal, default single thread download speed test... but compare the results to the multithread speed test.
 
Let me use my own home network as an example.  I just moved to Colorado Springs CO and got Comcast 105 Mbps cable Internet.  When I first setup I was only getting a max of about 30 Mbps, really spotty performance.  It was up, then down... then back up.  Something was wrong.  I had just installed a brand new Linksys AC 1750 router... TestMy.net showed me that router was just not delivering.  If I hooked directly to the ethernet I was getting nearly full speed.  I went through the wiring and replaced all the coax connections and added a splitter to drop the signal level down a bit (the readings on the modem were a little high).  Tested again and off ethernet I was now getting over the 105 Mbps Comcast promised.  But my wireless performance left much to be desired, even right next to the router I was getting degraded speed.  Changed channels and it helped quite a bit but I wasn't able to get the same speeds as direct ethernet or off the modem directly.   ... I ordered a Netgear Nighthawk AC 1900 off newegg and promptly returned the POS Linksys (I flip flop ver the years, sometimes I love linksys... but this time I was unimpressed).  As soon as I setup the Nighthawk I fell in love.  My speed now maxes out no matter where I am in my house.  I was especially impressed by the signal quality I get in my backyard, every other router I've tried has an issue with my patio.  The new energy efficient dual pane metallic coated glass that most sliding glass doors and windows now use is the problem... totally screws with the signal.  Been annoying the crap out of me for a long time.  I got the Nighthawk and it must just be powerful enough to not be an issue because I haven't had any issues like that since.  Best router I've ever seen.  First with a dual-core... but that's another topic all together.   :razz:
 
... okay, so my point.  During that testing, when I was getting degraded speed and was troubleshooting, I ran multithread tests and for fun I tested at my competition.  The results would have led me to believe that there wasn't an issue, there obviously was an issue... this is why I don't have the multithread option as default.  Just like in school, you get more help and learn more from a more difficult test.  My classic download test is the hardest Internet speed test to ace.  My competition multithreads by default, leading to a false sense of pride.
 
If you perform much better on multithread then work towards bringing up the single thread results.  TCP and MTU tweaking can be a huge help.  Try TCP Optimizer from speedguide.net if you use windows. It's free and has no install, it just helps tweak the settings.  OSX and Linux usually don't need much help but searching for Linux TCP gigabit tuning or OSX TCP gigabit tuning may help.  If you find settings for your OS that help you please share them with me ormy forum.  Future visitors with consumer gigabit will thank you.
 
Keep in mind that I've dedicated myself to this subject.  Although TMN started in 2001 I've been testing speed and helping people get more speed since 1996.  Much longer than my competition.  My servers have more than enough power to serve you, connected by a minimum of 1000 Mbps.  My main server is especially powerful, connected with 2x1000 Mbps uplinks, round robin load balanced, solid state raid 5 array and 24 cores... trust me, I have juice on tap at all times and upgrade my servers way before traffic is an issue.  You can't test bandwidth accurately unless you have excess bandwidth on the server end... I make sure it's always there and constantly monitor my servers.
 
I hope this helps with your question.  I wish you the best of luck pulling more out of that... your speed is awesome, don't get hung up just because you aren't able to pull gigabit out of it.  Google fiber could be shaping the bandwidth and only allowing you so much bandwidth per thread, I don't know for a fact but I've seen other ISPs do that.  You still have a gigabit uplink but they might limit the way you can utilize it.  It could also be that along the route somewhere you're crossing a 100 Mbps bottleneck, testing to my speed test mirrors can help you discover routing issues like that.  I would look into that horrible upload speed though.  Try testing with the largest size and see what you get.  I don't have a multithread upload test yet but I am working on it for the future. I'd expect much higher upload results than 4 Mbps.  Keep in mind that right now I have a limited 33MB upload test size, at gigabit speeds that goes too fast so the highest scores you can expect would be maybe 200-300 Mbps.  Upload size will be increased eventually.  Connections like yours are still rare so it's not at the top of my priorities.  If you get over 100 Mbps on the upload test... it usually means that you're pretty dialed in. :grin:
 
Good luck getting the most out of that insane connection you have!
 
Thank you for your continued patronage, happy testing!

- Damon - TestMy.net

 

 

Share this post


Link to post
Share on other sites
Sign in to follow this  
Followers 0