What's a Gigabit Good For?

Eight weeks into being a gigabit Internet service user, I can confidently say that… not a lot has changed. (See Geeking Out Over Gigabit.)

That's not to say I'm down on AT&T GigaPower as a service or that AT&T Inc. (NYSE: T) as a service provider isn't delivering what they promised. In fact, with the exception of one notable hiccup (fiber-to-the-home is easy, local number portability is apparently hard), the transition from Comcast Corp. (Nasdaq: CMCSA, CMCSK) Xfinity to AT&T GigaPower was quite smooth.

I got white glove treatment from my team of installers -- not terribly surprising since I was apparently customer number eight in Illinois. Plus, even when I got cranky about the voice line issues, AT&T's customer service reps were calmly professional -- even the one young man who tried to explain LNP to me and was promptly lectured as to what I was writing about when he was probably in grade school.

It's just that, it turns out my current needs don't really require gigabit speeds, or, at the very least, don't seem to take real advantage of that kind of bandwidth. Looking under the covers a bit, here's what I'm getting this morning:

AT&T Speed Test - Using WiFi

Of course, that's using WiFi, and as the only user in the home at the moment. By disabling WiFi and relying solely on an Ethernet connection, this is the bandwidth picture:

AT&T Speed Test - Using Ethernet

That's clearly not a full gigabit -- but 500-plus megabits ain't too shabby and it's within the terms of service, so I'm not complaining. I'm just not seeing any major change in how I am able to do my work or, for that matter, pursue various hobbies online. Like, today, tracking the NHL free agency scramble.

I accept the slightly slower speed that WiFi enables because, frankly, I can't tell the difference and the wireless connection is what enables my printer.

Read more about Gigabit Cities and the expansion of gig services in our Gigabit Cities section here on Light Reading.

The bottom line is that content downloads and even the performance of online videos on my laptop and even streamed Netflix content to a wide-screen HDTV hasn't noticeably changed since we upgraded to gigabit service. That's not just my opinion but that of other folks in the household.

That likely has nothing to do with our Internet service and all the world to do with our devices and their limitations.

So are we ready to switch back? Hardly. This service is rock-solid and costs (for now) the same as what I was paying Comcast for less-functional DVR and in-home WiFi and slower Internet speeds. But after some time passes, I might bite the bullet and downgrade our speed to save money. Of course, then I'd have to remove the "Guess what, I got a Gigabit" sign from my front window.

But if any of you have ideas as to how we could be making better use of FTTH and gigabit Internet, feel free to make suggestions… just keep 'em clean.

— Carol Wilson, Editor-at-Large, Light Reading

Page 1 / 2   >   >>
brooks7 7/10/2015 | 12:37:16 PM
Re: AT&T's Speed Test Carol,

When we were working on FiOS we did some work on systems comparing and testing assymetry and latency in combination with systems to get what TCP considers bandwidth.  Most test sites have moved to UDP since those days.  One thing is that the one site might be a TCP site still.  Delays in receiving TCP ACKs will be viewed as congestion and the protocol with throttle bandwidth in response (which is why we have things like WRED).  The whole thing is very complex when trying to test TCP bandwidth from end to end and be able to determine all the factors (like hey if I do G.fast does the Group Delay of the encoding make a difference?).


cnwedit 7/10/2015 | 9:26:03 AM
Re: AT&T's Speed Test Apparently this discussion has caught someone's eye at AT&T. I had a phone call yesterday from a technician, offering to help me figure out why I'm not getting the full gigabit service. He said management had contacted him about my "complaints." 

As I said to him, the primary point of this column was not to complain that I wasn't getting a gigabit but to note that I'm not really putting that much bandwidth to much good use. I think the bandwidth I am getting from AT&T is within their terms of service. 

So I now have someone I can call if I want to push the bandwidth envelope even farther, if I do find uses for all that bandwidth.
fiberman 7/9/2015 | 6:19:13 PM
Re: AT&T's Speed Test Very interesting. The speed you get is affected by a number of factors, not just the speed of your connection. It is affected by the numbre of users online at the time you do the test and what they are downloading. It is affected by the speed of the ISP's connection, not generally a big problem with large ISPs with adequate bandwidth but a big problem with small ISPs. AT&T probably runs their tests only to the local servers, showing only the speed of the local connection which is what they want to showcase, and they seem to have a "fast lane" that give you priority over other users for the test. IT also depends on the various Internet links and the load at the servers on which the test software is running. You do set the record for the biggest variation of speed - we usually only see 2-6X.

When I questioned a tech at AT&T on this issue and gave them my result which were their speedtest was 2-6X what I got with other services, they blocked me from using it anymore!
cnwedit 7/7/2015 | 11:11:49 AM
Re: AT&T's Speed Test I tried a variety of speed tests, including Comcast's, and most of them came up in the 600-700 Mbit/s range EXCEPT the one you mention - www.internetfrog.com - which came up at 21 Mbit/s download. That seems really strange. 

In addition to AT&T's and Comcast's, I did the Speakeasy test (Megapath), Ookla, Broadband Reports and ZDNet. The rates were all over the map, but clustered in the high range. The only one lower than 500 Mbit/s was ZDNet, and that was 121 Mbit/s, still well above internetfrog.com.


brooks7 7/7/2015 | 11:06:01 AM
Re: A mathematical explenation ... speed doesn't matter - but latency do! Well, your math makes tons of assumptions:

1 - That the time to fetch each piece of data is fixed.

2 - That there is no jitter in the network.

3 - That the network is lossless.

4 - That there is no congestion during part of the transmission.

5 - (and this is the big one) That Human Beings can notice the difference.

There were productivity studies done in the 60s by IBM that showed that network responses for a screen full of information needed to be under 250 msec.  If your website loads in 1 microsecond or 1 millisecond, will you notice the difference?  The only case that I can see that matters in everyday use is downloading of new or updated software.  


janneh 7/7/2015 | 10:43:51 AM
A mathematical explenation ... speed doesn't matter - but latency do! This is a very interesting post.


You had 50 Mbps before and got 1 000 Mbps. But the speed did not play out as the "game changer" of your digital life as I understand the blog.

Could it be hat speed does not matter at those levels...

Could it be that laws f math and latency plays out at full effect...?

 Everybody talks speed those days. One should have a gigg to be really ahead of the pack.


Speed is of course good - if it is delivering and is not affected by other more important circunstances.


So we need to think about what speed is:


1. Speed is DataVolume/DownloadTime. That i simple enugh ;-)

Le's go further and think about whats in hat ormula.

2. "DataVolume" is what it is a - chunk of data. But when we surf the web hat chunk is a composition of sub-chunks (objects of text, graphs, images, vido ec....  Lets say there are n-number of dataobjects in the service you want to view eg in the Internetbrowser.


3. "DownloadTime" is composed of two parts:

- DataTransmissionTime: The time it takes to download the datachunk - composed of n-objects. (Datavolume/Transmissispeed)

- LatencyTime:  that it takes for the entire system (server in he network, the link, the device to view the content/object). One can (to simplify) compse the ime it takes to get the information like this n*Latency 

- DownloadTime = DataTransmissionTime+ LatencyTime= (DataVolume/Transmissionspeed) + n*Latency).

Note: A consequence of increasing the Transmissionspeed and not decreasing latency of each object have "severe" mathematica consequences. If Transmission speed goes toward infinity that part of the equation goes toards zero. If the latency of n-datavolumes are not changed it will "take over" the ToalTime it take to transmit the informatin and turn it into a user-service (eg the beautiful web-page). It also get's worse if the designer of eg a webpage have a lot of small databjects on the page. The number of (n) objects rises and so does the LatencyTime...

(The Transmissionspeed is the linkspeed in he aticle discussed as of being 50 Mbps in he cable deployment) and 1 000 Mbps in the fiber deployment - if the data-chunk downloadee is small (eg a 100 kByte webpage with 30-50 objects) the latency will be the dominant part of the Totaltime.

4. Let bring it all together. The formular for the

User Experienced Speed (Mbps)= DataVolume/(DataVolume/Transmissionspeed) + n*Latency)


5. At high speeds - latency of objects rules the experience

And let Transmission speed go to a high number.... The simplified (often practical stimate) version is:

UserExperienedSpeed= DataVolume/n* Latency


Simply speking: At high speeds - it is only Latensy that rules the experienced-speed. What one should attack is consequently the chain of object-latency - then the experience will improve.

Does this makes sense in this case...?


Kind Regards

Jan Hederén







brooks7 7/4/2015 | 2:06:52 PM
Re: What's Gig got to do with it... Video occupies almost all the requirement for non-Technology focused users who feel like they need to Upload 3-D CAD designs all day long.

So if Cable opened up its billing model on its current spectrum (which given the number of OTT plays by their vendors they might be able to do over time), do you think the average user will need more video than he gets from cable?


smkinoshita 7/4/2015 | 11:20:40 AM
Re: What's Gig got to do with it... The average user might not need it now, but what about a couple of years from now?  I bet it'd be a different story.
danielcawrey 7/3/2015 | 9:30:21 PM
Re: What's Gig got to do with it... While gigabit does sound nice, I don't think the average user needs it. This may be why although we read a lot about how the US lags behind in broadband speeds services providers don't really do anything about it: It's because they don't really need to.

There's probably 5% of the internet using population that really needs gigabit as a consumer purpose. And for service providers, it's probably better to not spend the money on it. 
fiberman 7/2/2015 | 2:05:48 PM
AT&T's Speed Test Carol,

Have you tried other speed tests on your connection? AT&T's speed test usually shows speeds 2-6X as fast as non-AT&T speed tests like internetfrog.com. It seems to be that AT&T uses "fast lanes" for speed tests - when I called them about it, they blocked my DSL and I can no longer use their speed test!
Page 1 / 2   >   >>
Sign In