How to measure your true Internet speed

Topic

New Reply

	TOP STORY How to measure your true Internet speed
	By Woody Leonhard Curious about your Internet speed? Most Internet service providers (ISPs) offer some sort of throughput test tool on their sites — just click, and you’ll get a couple of often-impressive numbers. It’s in an ISP’s interest to provide the best speed numbers possible; your actual throughput is probably something quite different.

---

The full text of this column is posted at WindowsSecrets.com/top-story/how-to-measure-your-true-internet-speed/ (opens in a new window/tab).

Columnists typically cannot reply to comments here, but do incorporate the best tips into future columns.[/td]

[/tr][/tbl]

Reply | Quote

Replies

I think Woody made some calculation errors here. Speed of light is 300.000km/s (roughly). Thats 300km/ms. With an outer span of the earth of around 40.000km, you need 133ms to go completely around. Only problem is, 300.000km/s is only reachable in vacuum. Down here on earth, the speed of light will reach about 200.000km/s in fiber. That’s 200km/ms and takes 200ms for a round trip.
I haven’t seen a stretch of 40.000km fiber yet, and any switching equipment in the mix, however fast, will introduce extra latency. In addition, fibers aren’t taking the birds path, but “slither” along the ground. 250-300ms is an extremely good result for a roundtrip.

So a perfect internet connection, takes 200ms for a roundtrip. Real connections, at least 300ms. And thats when everyone on the planet is asleep, or doing things we used to do back 30 years ago.
Cheers!

Reply | Quote

GACK! You’re absolutely right. I didn’t want to get into a discussion of the speed of light in copper or fiber cable, much less in the atmosphere, but in a vacuum it’d take 134 ms for a circumnavigation. Er, my slide rule slipped…

Reply | Quote

I remember a few years ago when dial-up was how we got on the internet I had a problem with “internet speed” and called Bell South to send a tech out to check on the problem. He connected his instruments and checked my internet speed and then he checked on a couple of online speed tests and my speed was quite within the advertised range. (Was 600 to 800KB or Kb per second good internet speed then? I forget that number.) BUT! My “data transfer speed” was like 12 to 25 KB/s (or was it Kb/s). The end of that story was that my ISP claimed that their service was normal and you cannot expect an internet download speed to transfer to your computer harddrive at that rate. The point of all of this history is to indicate that your data transfer rate is the only thing that tells you how good your internet connection/service is. BellSouth was delivering to me at the same speed that Verizon was delivering to my computer through my tethered cell phone. As soon as I switched to cable internet with a 1 MB/s (Mb/s) download speed my transfer rate jumped to 300 to 500 KB/s. In the end, internet speed means nothing to how many seconds or minutes it takes to download a given file. Transfer rate is the only measure that has concrete meaning. Why don’t authors who report on internet speed include transfer rates with internet speed when they are comparing numbers? If I have a 500 KB/s transfer rate, I can know that a 10 MB file will take 20 seconds plus a few seconds connection time to download. Can you comment on this aspect of “internet speed”?

Reply | Quote

Sorry for the pedantry but the A in ADSL is for asymmetric rather than asynchronous. Asymmetric = differing upload/download speeds.

David

Reply | Quote

I always thought looking at your router status was the easiest way to find your speeds.

32540-RR

Reply | Quote

I’ve been using testmy.net. (It is often recommended on DSLReports.com.)
They make a good case for why their tests are different here: http://testmy.net/legit-speed-test.php

Some fo it may be marketing, but I do know this: I get consistently lower results from this test than from anywhere else. (Also, I get consistent results).

– Chris

Reply | Quote

Years ago about a dozen cities got together and formed a ‘fiber to the home’ entity. Their philosophy is the city provides pipes for water and sewer, why not a pipe for all communications? After many battles with the local ‘copper to the home’ telco they just succeeded in building out our city with fiber to your home if you want to sign up. The city owns the fiber and any company can sign up to provide services. The city is not the ISP. UTOPIA is the entity and the city of Centerville is where I call home in Utah.

I have a single fiber strand that terminates inside my home. It provides full symmetrical 100 meg ethernet! I also cut my phone over to the fiber and have free LD and about every phone feature you can imagine and then some! I also have a block of 16 static IP addresses. The cost of the 100 meg ethernet, phone, and IP addresses is around $88 a month!

I don’t use any of the copper or my local phone company anymore and will never look back. For $300 a month I can get a full 1 gig ethernet both up and down – full speed in both directions.

With a 100 meg pipe to my house for uploads and downloads, speed is essentially governed by the remote server I’m connecting to. I downloaded over 5 gig of data faster then it took to copy the data from a thumbdrive to a PC! I have empathy for the rest of you with your DSL, cable, and other service providers. As for me – I’ll take glass any day!!!

Reply | Quote

Woody Leonhard’s column in internet speed is interesting and valid. However, I suspect that he’s measuring too many events in one measurement.

For me, there are only two important speed measurements.

[*]What is my speed (upload and download separately) to my ISP, which is controlled by my hardware, the ISP’s modem, and the ISP’s lines to my broader internet connection.
[*]The second is the speed with which my ISP contacts the backbone in my area. All else is out of my ISP’s control. Yes the other speeds are interesting, and affect me, but I have found great differences in, for example, servers, particularly during busy times.

To summarize, I’m paying my ISP to connect me to the backbone. Anything from there is my issue. I can choose to have different DNS servers, etc. to optimize my information processing times, but, basically, I can control only these two issues (by changing my hardware, and/or changing my ISP). The rest is only open to praise or complaint.

Reply | Quote

Woody, you left out a good, non-biased, test set that has existed for some years: Argonne (Illinois) National Laboratory Network Diagnostic Tool (NDT)

http://ndt.anl.gov:7123/

Which also has links to other NDTs in the US & Switzerland.

Yes, it is Java & based on one file up & one file down, but at least it’s not run by an ISP that may make the results look more favorable for itself & less so for its competitors. Always found it to be pretty consistent.

Bill

Reply | Quote

Lots of good points here. I’ll add mine.

Years ago, when cable internet was still fairly new and unstable, I would ftp files from my ISP using the command line. I turned on hash marks to see the progress. I could then “show” the cable tech how there were “pauses” during the transfer. So small files transferred quite quickly but large files would incur speed bumps thus bringing down the overall speed. Since this was direct to the ISP it really was what they were delivering to the house. No browser interference either.

Everyone else is right about the external affects. It doesn’t matter what your ISP gives you. If you’re going to busy servers somewhere down the line you mileage will vary.

Ric

Reply | Quote

Two years ago I came across somewhere that mentioned “Sam Knows” an FCC sponsored program that included free speed monitoring and a free Netgear router. I have had it installed for 2 yrs and every month I get an e mail with graphs of the last months performance.
I don’t have worry about third party ratings as it is continously monitored.

Reply | Quote

Very worthy subject and great follow-on conversations. Thank you!
One of the interesting (newer) twists to this topic that should not be overlooked is the “JITTER” measurement and specification.
Recently, when I was attempting to rid myself of POTS Telco line that has gotten unbearably expensive [omitting the company name intentionally], I came across this requirement for the Ooma VoIP hardware that just was not cooperating. Everyone I called, using Ooma, kept telling me that they were hearing themselves echo but my side was surprisingly clear!
After multiple tests and trials (and desperate calls to Ooma) finally revealed that their VoIP box is recommended to be used on Telco lines that have 5mS (milliSecond) or less of Jitter.
I live in a beach community with lackluster (antiquated) copper POTS lines and the salt air probably does not help matters.
I finally called the phone company and their technicians tried everything inside and outside the house and even climbed the pole in the alley
They verified that their lines are within performance envelope and there was nothing that could be (or needed to be) done!
Using http://www.pingtest.net, my phone line consistently showed 15mS of jitter (3x higher than Ooma requires).
I was forced to return the Ooma box. Ironically, Vonage does not have this issue although voice clarity (IMHO) is not as good as Ooma’s.

Reply | Quote

Since 1998, I have been using a PAYware program called DUmeterthat boots up with my WinOS system automatically and always shows me network activity in a 4x2cm widget on the lower right hand corner of my screen. There are other FREEware versions of this program but the features of this utility have worked for me all these years.
YMMV :mellow:

Reply | Quote

Since 1998, I have been using a PAYware program called DUmeterthat boots up with my WinOS system automatically and always shows me network activity in a 4x2cm widget on the lower right hand corner of my screen. There are other FREEware versions of this program but the features of this utility have worked for me all these years.
YMMV :mellow:

Network Activity is not what this thread or the article is about. Network Activity does not measure directly your actual connection speed.

What Network Activity monitoring programs and applets are designed to do, is to monitor your throughput and usage per period of time. This is useful if you have restrictive bandwidth usage caps, but not really very useful even there.

When you measure throughput at the computer, you get what the traffic is into and out of the computer, not what the throughput is at the router. Measuring at the router, you are assuming that every packet which goes through (incoming or outgoing) is full of data. This is rarely the case in practice. Most ISPs send out and receive data packets which are up to one-third empty of data. This means that your ISP is measuring up to one-third more bandwidth usage than your in-house monitoring utility is reporting. So you may go over the cap, while your own measurements say you are under the cap. DSL Reports is full of threads which reveal just how prevalent this sort of overbilling is.

I use Softperfect Research’s Networx to monitor my usage and alleged speed. Unless I am downloading a file, I rarely see in the utility’s display anything near my true upstream or downstream speeds, and it is hard to say that any given Web server is delivering downloads at the maximum possible download rates for my level of DSL service.

Also complicating speed testing is the feature offered by most ISPs these days, which accelerates the first few MB of any download to tremendous initial speed bursts, only to level off to much slower steady-state download speeds for the rest of the download. So, any speed test which does not take this speed burst technology into consideration is essentially lying about the steady-state Internet Connection Speed. The article did not cover this effect.

So this isn’t an exact science we are doing here. It is at best a rough estimate, no matter what methods we use to determine or true connection speeds.

The effects of raw speed pale in comparison with the throughput advantages of accelerators and the counteracting limitations of download speeds imposed by the servers from which content is being downloaded. If I am downloading several large files from different servers at the same time in my Chrome Browser, the actual rates of data received do not add up to my overall bandwidth usage nor my overall connection speed. (The totals are usually much higher than the overall reported rate.) And this is totally normal, given the behaviors of Web servers and browser accelerators.

The same arguments apply to video streams, torrents and uploads to Cloud-based services like Skydrive. In fact, depending on how busy a Web server is, speed of transfer can vary tremendously, or the connection can be lost altogether.

The amount of data actually transferred per unit of time, all factors considered, is the only “speed” we end-users will ever really notice. And the lion’s share of this “speed” is entirely out of our hands, and out of the hands of our ISPs.

-- rc primak

Reply | Quote

Mr. Bob Primak,
Here is DUmeter at work while it is showing the activity of the network connection along with the actual data being d/l + u/l.
32546-201212_DUmeterExample
Once familiar with its inner workings, it can do a lot more than just show activity and also provide for a fairly accurate bps data (in my case; it is close to speedtest.net results)
Cheers!

Reply | Quote

Superbly useful article which sheds light on a complex subject. Maybe the best answer to the speed question is “it depends…”
Woody’s article explains why that is, in easily understandable terms.
By the way, I’m a bit surprised at the tone in some of the correction notes. Yes, Woody may have made a “slide rule” error but he did a lot of work to put this all together — with no $$$ reward for it. So, lets be grateful for smart folks who are willing to share their good work.

Reply | Quote

… I have empathy for the rest of you with your DSL, cable, and other service providers. As for me – I’ll take glass any day!!!

My computer’s network card’s top speed is 100 mbps (megabits or million bits per second).
I have a coax cable (copper to home) connection with my ISP Telenet.be. It offers me telephone (POST telephone set), analogue TV channels, interactive digital TV channels, on-demand TV programs, internet access (Telenet promises 120 mbps down and 5 mbps up) with uncapped down/uploads.

Just today I downloaded a 508 MB file from a GigaNews server in 46 seconds, or 88 mbps, which is typical. I usually get GigaNews download speeds between 85 and 90 mbps. :bananas:

For this service subscription Telenet.be charges me € 73,- per month (+/- $ 95).

Not all that bad, eh?

Reply | Quote

Especially since this is a technical topic let us all strive for accuracy. “M” is for “Mega” or 10^6. “m” is for “milli” or 10^-3. Not a lot of difference, just a factor of a billion.

Reply | Quote

Measuring Internet speed depends on many variables, such as traffic at source and traffic ad destination, so it is difficult to get an accurate figure. With that said, a possible alternative not discussed in Woody’s article is to try downloading a fairly large file from a web site that has fast servers. I daily download multimedia files from a commercial company that has its own software including a download manager. My computer has a 15MB/s connection. I normally get anywhere from 1200 KB/s to 1700 KB/s per download. The speed is measured over 10-20 minutes. I have two desktop gadgets which I use as a second source to compare download speed. However, you don’t need a commercial source. All you need is a download manager that shows download speed and any site that provides downloads via good servers. I use Speedtest and the results there closely correspond to my download and gadget figures. You can also manually measure download speed, by noting the file size, time used to download the file and dividing the size by the time.

Reply | Quote

I concur with the observations that different tests provide different answers. The Ookla test at speedtest.net has been the most consistent for me. It will also provide different results if you choose different servers, so check around a little bit to see which is the fastest. That will be, after, the best test of YOUR downstream bandwidth. Anything less is not measuring YOUR bandwidth, but rather choke points somewhere upstream of your ISP. The Comcast server in Chicago is fast if Chicago has a fast backbone to your city. It easily provides >40Mbps downloads to St. Louis. A cable user in St. Louis reports that the Charter server in Olivette delivered >90Mbps.

“Accelerators” in browsers appear to be merely caches. They should not affect your measured download speed because the speed tests ignore them. If the caching takes place at your ISP, the bandwidth of the loop from your ISP to you is still the limiting factor. Ditto “multiple streams” — they still have to share that local loop.

Some antivirus implementations will affect the apparent bandwidth because they may not release the file back to Windows until it has been scanned. Try testing with A-V turned off to see if that is true in your case. In one case a defective speed test was reporting the speed of the transfer from the A-V program to Windows! Wow, 50Mbps on a 6Mbps DSL line!

Once you have determined your FASTEST download speed (mine is 5.2Mbps on a 6Mbps-rated DSL line), you may be assured that anything lower in the same time frame is the result of being tested by a slower server, a slower backbone path to your ISP, or a defective test.

This whole topic points out how deficient the US is in providing broadband to the public. My 5.2Mbps used to be considered fast. Two years ago in a cheap hotel room in Copenhagen, Denmark, I tested the FREE Ethernet and received >7Mbps. At speedtest.net review the speeds reported by users in other countries. You may be ready to emigrate!

Reply | Quote

I’ve used a variety of the speed test sites, especially when I was on DSL and suffering terrible performance. But as a day-in-day-out measure I find it much more effective to simply (running Windows 7) open Windows Task Manager to the Networking tab*. Right now my local network connection is 100 Mbps, and my internet service is advertised as 12Mbps. That means when the graph approaches 12% I am getting all the performance available. When I stream a basketball game in HD I can see that it runs in short bursts to the limit alternating with intervals of no activity. Downloads by Windows Update pretty much stay right up there near the 12% level. I start Task Manager manually after logging in. I find the Network tab doesn’t seem to start tracking until I have opened it the first time. Of course this only shows the activity of my computer, so if the TV is streaming Netflix my performance would take a hit.

*I don’t know if it makes any difference, but I always check the “Show processes from all users” option on the Processes tab.

RH in CT

Reply | Quote

I always thought looking at your router status was the easiest way to find your speeds.

RR, the image you posted was of your DSL Modem’s connection status as part of your DSL combination modem/router. I believe that the downstreamconnection figure supplied is what your ISP provisioned you for as a maximum speed, not what you are actually getting.

Jerry

Reply | Quote

RR, the image you posted was of your DSL Modem’s connection status as part of your DSL combination modem/router. I believe that the downstreamconnection figure supplied is what your ISP provisioned you for as a maximum speed, not what you are actually getting.

Jerry

Very true Jerry, I use it to compare downloading speed usually using Utorrent.

Reply | Quote

You missed one of the best, if not the best, (and free) speed measurement tools- Shaperprobe, from Georgia Tech. It measures you maximum upload and download speeds, as well as the maximums ultimately permitted by your ISPs traffic shaping rules. See http://www.cc.gatech.edu/~partha/diffprobe/shaperprobe.html

Reply | Quote

pseudoid posted (#18):

Ironically, Vonage does not have this issue although voice clarity (IMHO) is not as good as Ooma’s.

No, this is not ironic. It’s simple math. Faster bitrates (Ooma) give better clarity, while poorer quality (Vonage) results from slower bitrates. The slower bitrates tolerate more “jitter” than the faster bitrates. Hence the differences you noted. It’s even more noticeable with my slow DSL service and video streams. I have to choose low bitrate Standard Definition because High Definition hesitates and drops out too much.

-- rc primak

Reply | Quote

Again, these are raw throughput results, not actual real-world data transfer rates. My Networx graphs show me exactly the same data. But these data (in my tests) do not correspond at all times with the online test results or my ISP’s reports (included in monthly billing) of bandwidth usage. It’s a good First Approximation, but not the official reality.

-- rc primak

Reply | Quote

I found the Verizon Internet speed test gives about the same results as several of the independent speed tests. So it appears reliable. However, I also found that my speed is only about 1/3 of what they promise for the FIOS Quantum service I upgraded to.

Reply | Quote

To follow up, I went to Verizon FIOS Technical Support. They were able to fix things at their end so that with a wired connection, I was getting better than they promise for download speed (56-58 Mbps compared to guarantee of 50) and upload speed (31 compared to 25). They only guarantee a wired connection. I found the fix did not affect wireless speeds, which are not guaranteed. These were about the same as wired speed, but now are only a fraction of wired speed, having not improved whatsoever. Does anyone know a reason why wireless might not be able to support a speed above 18 Mbps?

Reply | Quote

…Does anyone know a reason why wireless might not be able to support a speed above 18 Mbps?

Is this 802.11g? Despite the 54 Mbps rating, net throughput will be less:
http://en.wikipedia.org/wiki/IEEE_802.11g-2003

It could be an older router with WEP/WPA turned on. Or interference from other wireless networks.

Reply | Quote

I use speedtest and speakeasy and between the 2 get results that make sense for me, including when up against some of the tests you suggested. I have Optonline and upload speed is as Cablevision specifies, and download is noticeably higher. As opposed to claims another provider has made about CV.

Reply | Quote

I tested my Frontier DSL line, which is way out in the boonies, and rated/contracted for 3.0 Mbps down using all the speed tests in the article plus two additional sites (Frontier and Speakeasy). All came back with a download speed of between 3.14 and 3.3 Mbps and an upload speed between .34 and .38 Mbps. Only Bandwidthplace gave somewhat inconsistent results with 2.88 Mbps down and .1 Mbps up.

Regardless, this is fine for our family and we only pay $25 a month for our DSL line since we can’t get a higher speed.

I forgot to add that for a real time speed monitor I have used Netpersec which has been around since 2001. It is accurate, monitors both up and down speeds simultaneously and the real time graphs and numbers reporting (actual and average) are quite informative. Works in XP, Vista and Win7. Get it free here: http://download.cnet.com/NetPerSec/3000-18512_4-15690.html

Reply | Quote

Great article. I use DUmeter from Hagel Technologies (http://www.hageltech.com). to monitor my download / upload speed.

Reply | Quote

No Woody, “For example, the “A” in an “ADSL” connection means “asynchronous” the A does not mean asynchronous. I means “Asymetric”[/SIZE][/SIZE][/SIZE][/SIZE][/SIZE][/SIZE][/SIZE][/SIZE][/SIZE][/SIZE][/SIZE] That simply means that download and upload speeds are not the same. Asynchronous means something entirely different.

Reply | Quote

Thanks for the notes, everybody. I did, indeed, mis-calculate the speed of light, and I used “Asymetric” instead of “Asynchronous.” Ooops. I particularly appreciate all of the independent tests on the speed test sites. I’m still at a loss to choose one site over another!

Reply | Quote