General questions  MaxSpeed measurement  SurfSpeed measurement  Problems  Improvements  Throughput tips

 

General questions

What do the numbers mean? And the graphs?

See the Explanation-link when you have just done a measurement.

The download speed is what most people focus on. If you are overwhelmed by all the information then ignore the rest, just look at your download speed. The higher the number, the faster you can use the internet.

• Internet speed is expressed in bits per second or in bytes per second. A byte is 8 bits. It doesn't matter what bits and bytes are exactly, they're units just like weight, distance, and time.
• Bits and bytes are often confused. The convention is to use lowercase for bits (for example 'kb') and uppercase for bytes (for example 'KB'), but lot's of people don't know this. It's usually safe to assume 'bits' unless it is clear from the context that it should be 'bytes'.
• Thousand is abbreviated with 'k' (or 'K', or "kilo"). Million is abbreviated with 'm' (or 'M', or "mega").
• The "per second" part is often left out.

Abbreviation	Usually means
'bps', 'b/s'	bits per second
'Bps', 'B/s'	bytes per second
'k', 'kb', 'kbs', 'kbps', 'kb/s'	kilobits per second
'K', 'KB', 'KBs', 'KBps', 'KB/s'	kilobytes per second
'm', 'mb', 'mbs', 'mbps', 'mb/s'	megabits per second
'M', 'MB', 'MBs', 'MBps', 'MB/s'	megabytes per second

For example: "56kb" is the shorthand for "56000 bits per second". See the Unit Converter for calculations beteen bits, bytes, kilo, mega, and such.

Am I fast or slow?

It's technically impossible for this server to detect what kind of connection you are using (analog, ISDN, Cable, ADSL, network, wireless, etc.). This means I have no data, so I simply do not know what speeds are typical for a particular (your) connection.

Is my provider at fault?

People are quick to blame their provider for a low throughput. But a low throughput may be caused by things such as noise on the line, a bottleneck in your PC, or simply because the internet is busy. Try the measurement again at a quiet moment, for example deep at night. If you get a significantly higher throughput then this proves that your PC and local connection are allright.

Cable

Cable bandwidth is shared by all the subscribers that happen to be on the same segment of cable. A low throughput is probably caused by too many other users.

ADSL

The theoretical maximum speed is 8 megabit/second downstream and 1 megabit/second upstream at distances up to 3 kilometers from the telephone exchange. At 5 kilometers the speed has degraded to about nothing.

ADSL2+

The theoretical maximum speed is 24 megabit/second downstream and 1 megabit/second upstream at distances up to 1.5 kilometers from the telephone exchange. At 7 kilometers the speed has degraded to about nothing.

VDSL

The theoretical maximum speed is 52 megabit/second downstream and 12 megabit/second upstream at distances up to to 1.2 kilometers from the telephone exchange.

Who's the best provider?

I don't know. This website does not keep statistics about providers. The database would be huge, and only the largest providers generate enough measurements to be statistically significant. Furthermore, speed is not the only criterium on which you should base your choice. There is also the cost to consider, available services, helpdesk support, for consumer or business, local availability, etcetera, and I am not prepared to gather all that data to be able to give a balanced advice.

Other measurements say I'm faster?

The MaxSpeed measurement is much less sensitive to distance than other measurements, but measurements to servers nearer to you are still likely to show a higher speed. Also see the  Speed reduction by distance calculator.
The SurfSpeed measurement strongly depends on distance, which is why you have to select your country and/or language.

What is "latency"?

Latency is the time between sending a request for data and receiving (the first bit of) the reply. The lower the latency, the faster you can do things like chatting or playing games, and other low-data activities.

• Latency depends strongly on the distance between you and the server you are visiting. Electric signals travel at 299,792,458 meters per second, the speed of light through a vacuum. The circumference of the earth is 40,076 Km. If you visit a website at the other side of the world then your request travels from you to the website and the answer back to you for a total distance of (at least) 40,076,000 meters. Latency is therefore (at least) 0.134 seconds, or 134 milliseconds. If the signal travels through glass-fiber cables it's even slower, because light travels slower through glass than through a vacuum (about 70%).
• Another example: geo-stationary communication satellites are 35,768 Km far from the surface of the earth. Your request goes from you via the satellite to the server and the answer back via the same route for a total distance of 4 times 35,768 Km is 143,072 Km. Latency is therefore (at least) 477 milliseconds, almost half a second.

How about Ping measurements?

Many people (even specialists) use ping to measure speed. But ping measures network latency, which is a completely different property that should not be confused with surfing speed, bandwidth, or throughput. Please read my article  The Ping Fallacy for more information.

Is my distance to the Numion server relevant?

The MaxSpeed measurement is influenced by distance, although much less so than comparable measurements. The Surfspeed measurement does not depend on your distance to the Numion server at all.

Is the measurement influenced by caches and proxies?

No. The MaxSpeed measurement uses random data that will not be cached. The Surfspeed measurement uses a simple trick to fool caches, including remote caches such as proxy servers. A random string is added to the URL of every image. The cache will think it's a different image and will fetch it again. There's still the normal cache-overhead for storing the images and deleting old files, your harddisk is used exactly as if you were surfing by hand. Cleaning up the cache will temporarily increase your surfspeed because the cache-overhead is reduced (no old files to be deleted).

I'm not from that country?

See the  Country Explanation page.

How do I use the auto-repeat function?

Select an auto-repeat period in the "Options" screen, perform a measurement, and then keep the browser running on one of the results pages "Your Measurement", "Explanation", "More Details", or "Your History". The auto-repeat is cancelled if you surf away or stop the browser. You can minimize the browser if you want. To do other work start a new copy of the browser (as if it is not running yet), do not use "File -> New -> Window".

Note: there are no special facilities for dial-up connections, i.e. the measurement cannot hangup the phone when it has finished.

A kilobyte is 1024 bytes, right?

In data communications only the Metric definition of a kilobyte (1000 bytes per kilobyte) is correct. The binary definition of a kilobyte (1024 bytes per kilobyte) is used in areas such as data storage (harddisk, memory), but not for expressing bandwidth and throughput.

Why are the graphs logarithmic?

If you go from 64k to 128k then your speed is doubled. Similarly, going from 128k to 256k is a doubling as well. But if these speeds were plotted on a linear scale it would seem like the second doubling is much more of a gain than the first. Logarithmic graphs are perhaps a bit harder to read, but much better for depicting speed.

What is the timezone for the graphs?

All graphs and times are in your local timezone - all legends are automatically adjusted to the local time on your computer. The only exceptions are the Top 100 and Newest 100 pages, which use the timezone of the Numion server, which is GMT+0100 (Amsterdam).

 

MaxSpeed measurement

What do the numbers mean? And the graphs?

See the Explanation-link when you have just done a measurement.

How can it be accurate? It's too fast.

People are used to the simple and very common start-stop measurements that are available all over the internet, which are slow and need huge amounts of data to achieve even a low degree of accuracy. MaxSpeed uses a different and far more advanced measurement method. It analyses traffic as the download progresses, looking for a nice and regular bit of traffic. This bit of traffic is (assumed to be) representative for your maximum speed. The measurement stops and the speed is calculated using closest-fit linear regression. The measurement can therefore be very accurate in very little time with very little data.

How does it work?

The measurement is a small program (Java applet) that runs on your computer. For the download measurement it asks the Numion server to generate random and uncompressible data. While receiving the data it measures the speed and when it has enough information it stops, minimizing the time and the amount of data used. The upload measurement does the same, but this time the Java applet on your computer generates the random data and the Numion server does the measuring.

Example

At time=zero the measurement on your computer sends a request for data to the Numion server. The request travels through the internet and the server replies by sending a stream of data back at maximum speed. The graph is a timeline from left to right in seconds. The yellow line shows the total number of bytes received and goes up whenever a block of data is received.

There are two reasons why the yellow line forms a staircase. First, networks use blocking to optimize performance. Data is not received bit by bit, but in large blocks, hence the staircase. Second, the clock in your computer will "tick" every now and then, for example every 10 milliseconds, depending on operating system and Java implementation. Networks nowadays are so fast that lot's of data can be received during a single clocktick, again showing a staircase.

The white line in the graph shows your speed. The steeper the line, the higher your speed. The mathematical formula for a straight line is "Y = A + B * X", where A and B are constants. The formula can be rewritten as "Y = B * (X + A/B)", or in our case:

BitsReceived = Speed * (Time - Latency)

The program calculates your speed from the tops of the staircase with closest-fit linear regression for all possible intervals with a minimum intervalsize of 7 consecutive steps, and determines the interval with the highest R-square. The measurement will last for at least 10 steps, stops after 30 steps, or after 10 seconds, or when 10 megabyte has been sent, or when 7 consecutive steps are found with a very regular pattern (R-square better than 99.5%).

What does the "regularity" mean?

The measurement looks at how smoothly your connection transports the data. A high-quality connection has a high regularity. If there are pauses, glitches, hiccups, and other noise, then the regularity is less. If the regularity is less than "GOOD" you should disregard the results and measure again. The regularity is determined by calculating the closest-fit determination coefficient (also known as "R-squared"), a number that shows how much the calculated line deviates from the individual points.

If R² is	Then the regularity is
better than 99.8%	EXCELLENT
better than 99.5%	VERY GOOD
better than 99%	GOOD
better than 96%	FAIR
better than 90%	POOR
less than 90%	VERY POOR

Is it accurate at high speeds?

Yes, because the measurement calculates the speed by linear regression, and because it automatically adjusts the amount of data and the time interval. It can achieve a very high accuracy in a few hundred milliseconds, where other measurements need many seconds.

What can I use MaxSpeed for?

If you have a good connection then the numbers from the MaxSpeed measurement will be virtually the same between measurements. The measurement can therefore be used to:

• Check the claims made by your provider.
• Comparing your speed with friends (boasting).
• Quick checkups.
• Showing trends over time.

Other measurements can be useful for:

• Complaining to your provider (if the particular measurement is not influenced by things outside the control of your provider).
• Tuning your computer.
• Pin-pointing a problem.
• Measuring speed towards a specific server.
• Measuring other internet protocols (email, ftp, irc, etc.).

 

SurfSpeed measurement

What do the numbers mean? And the graphs?

See the Explanation-link when you have just done a measurement.

How does it work?

The measurement is a small JavaScript program running on your computer that fetches images (logo's) from 40 large and popular websites, measures the time, the size of the images is known, and from these two numbers your surfspeed is calculated in bits per second. Your computer is used exactly as when you are using it by hand, the measurement simulates normal surfing on the internet. The average of all micro-measurements is a very good (quantitative) representation of the (subjective) internetspeed as you experience it. The higher the number, the faster you can surf.

What is "surfing speed"?

Surfing speed is the speed at which you can surf the internet (visit websites). It is expressed in bits per second (or bytes per second) just like other internet speeds, but it is NOT similar to, say, download speed. At high speeds your surfing speed is determined (limited) by latency, not by throughput. A 1 Mbps connection will generally have a lower download speed than an 8 Mbps connection, but can have a higher surfing speed.

Tons of banners, bah!

The measurement uses deeplinked images (logo's) directly from their websites, see the  Imagelist for details. They're not banners, nobody is paying me anything. The system is universal and could measure any old images anywhere on the internet. I've chosen logo's mainly because I figured that the owners would then be less likely to complain about all the useless traffic. After all, it's their bandwidth and they're paying for it.

Why is my surfspeed so much lower than my download speed?

• Your surfspeed cannot be higher than the maximum speed of your connection, only lower.
• The measurement calculates your average speed to a mix of slow and fast websites. An average (speed) is always less than a maximum (speed).
• The measurement is an accurate simulation of surfing by hand and, unlike the MaxSpeed measurement, is therefore limited and influenced by things like your CPU, video, harddisk, browser, HTTP overhead, DNS, etc.
• The measurement uses small files, the most common files on the internet, which cannot be downloaded as efficiently as large files. See  Big files or small files?
• At low speeds your surfspeed can reach your download speed, but at high speeds it will be capped by latency (distance).

What makes YourSpeed so special?

• Measures your real surfing speed. It was specially developed to put a number on the "speed-feeling" you get when surfing the internet.
• Accurate surfing simulation. The measurement uses your computer exactly like you would and is influenced by exactly the same things affecting your surfing speed, such as programs running in the background, hardware (video, harddisk), and the efficiency of your browser.
• Measures many websites in a single measurement. Speed towards a single website doesn't say anything about your speed towards other websites due to the complex topology of the internet.
• Noise cancellation. The speed of the internet is far from stable, there's lot's of noise. By averaging lot's of micro-measurements YourSpeed manages to cancel out the noise.
• Multitasking. Other measurements are generally single-tasking.
• Uses small files for the measurement, not big files. Most files on the internet are small files, so that's what should be measured. See  Big files or small files?
• Users may compare their results. Comparing results from single-website measurements between users is meaningless because of differences in network-distance and network-routing.
• Also measures outgoiing data, not just incoming data.
• Also measures DNS, very important when surfing, and ignored by most other measurements.
• Also measures HTTP overhead, generally ignored by other measurements.
• Geographic location independent. It does not matter how far or near you are to the Numion server.
• Compatible with many operating systems and browsers.
• Connection independent, works on all types of connections, for example analog modem, cable, ADSL, satellite, etc.
• Very easy to use, no plugins or downloads to install.
• With personal, country, and worldwide statistics.

What can I use YourSpeed for?

The top speed of your car is what you boast with, but how long does it actually take to drive to aunt Emma today? Compared with other measurements YourSpeed can be used for:

• Measuring how fast you can really surf the internet.
• Quick checkups.
• Tuning your computer.
• Comparing your speed with friends.
• Showing trends over time.

Other measurements can be useful for:

• Complaining to your provider (if the particular measurement is not influenced by things outside the control of your provider).
• Pin-pointing a problem.
• Measuring big-file download speed.
• Measuring upload speed.
• Measuring speed towards a specific server.
• Measuring other internet protocols (email, ftp, irc, etc.).

 

Problems

• Check the  Newest 100 and  Top 100 lists. The problem is probably local to your computer and not a problem in the Numion server.
• I cannot solve your problems. You have to do it yourself.

The most common problem: Java and JavaScript.

This website makes extensive use of Java and Javascript. Both are standard programming languages included with many browsers and used by many websites. They are not the same, but 2 totally different programming languages.

To see Java error messages:

Java Console on Internet Explorer:	Sun Java: Tools -> Sun Java Console Microsoft Java: 1. Tools -> Internet Options -> Advanced -> Microsoft VM -> Java Console Enabled 2. View -> Java Console
Java Console on Firefox:	Right-click the Java icon in the Windows taskbar, select "Open Console"
Java Console on Opera:	Tools -> Advanced -> Sun Java Console
Java Console on Windows:	Control panel, Java -> Advanced -> Java Console -> Show Console
Java Console on Linux:	"/usr/lib/java/ControlPanel" or "/usr/lib/java/jre/bin/ControlPanel" and then check "show console".

To see Javascript error messages:

Javascript errors on Internet Explorer

Internet Options -> Advanced -> Browsing -> enable "Display a notification about every script error"

Javascript Console on Firefox

Tools -> Error Console

Javascript Console on Opera:

Tools -> Advanced -> JavaScript Console

Tips to get Java working on your computer:

• Do you see a "Java Required" in big red letters on the  YourSpeed Homepage? Then Java is not installed on your computer, or it is disabled. You can check if Java is installed on the  Test your Java page on the www.java.com website. Just below "Test your JVM" you should see "Congratulations. The latest version of the JRE is installed and running correctly on your system". If you don't have Java or want to upgrade then click the button to the right:
• Windows XP/2003 Internet Explorer Enhanced Security Configuration: goto the "Control Panel", "Add or Remove Programs", "Add/Remove Windows Components", and de-install the "Internet Explorer Enhanced Security Configuration".
• Internet Explorer security: goto "Tools -> Internet Options", the "Security" tab. Click "Trusted sites" and the "Sites..." button, uncheck the "requires server verification" checkbox and add "http://www.numion.com".
• Java is installed but disabled. In Internet Explorer open the internet options, "Advanced" tab, look for a heading called "Microsoft VM", and check the "JIT compiler" item. Also see the "Security" tab, custom-level button, a heading called "Scripting", check the "Scripting of Java applets" item.
• Some firewalls will block Java applets. If you have a firewall then check it's settings, or ask your system administrator.
• The internet settings may be damaged. In Internet Explorer open the internet options, "Programs" tab, click the "Reset Web Settings" button.
• The browser may be damaged. For Internet Explorer open the Control Panel, "Add/Remove programs", select "Microsoft Internet Explorer", "Change/remove" button, "Repair Internet Explorer" option.

The 2nd most common problem: software interaction.

An old joke is that a computer with no problems is a computer turned off. The more software you install, the more problems you will get from programs unexpectedly interacting. If you have a problem then try to turn off, or even de-install, as many programs as possible. Especially look for hidden software, there is lot's of junk around that can install itself on your computer without you knowing it. Other examples to look at are: security programs, pop-up blockers, accelerators, toolbars, download helpers, content filters, messengers, RSS feeds, internet phones, internet radios, peer-to-peer networks, firewalls, virus checkers, and automatic updaters.

The measurement jumps without pausing.

Select "Auto-repeat = Do not repeat" in the Options page. The measurement will then pause and wait for a mouseclick between the MaxSpeed and SurfSpeed measurements.

The measurement does not continue.

This website contains advertising banners to pay for the hosting. The measurement will wait for the banner to load before it starts, or the measurement results would be influenced by the loading of the banner. Banner-protection software on your computer may cause the measurement to not continue.

Download and upload speeds are not measured.

• Make sure the MaxSpeed measurement is enabled (checked) in the Options page.
• See above: "The most common problem: Java."

The graphs on the results-page do not show.

See above: "The most common problem: Java."

Impossible surfspeed.

This is usually reported by users of Safari and Konqueror browsers. I'm sorry, but I have no access to machines with those browsers and cannot solve the problem. Try another browser, such as Mozilla, Internet Explorer, or Opera.

Impossible download speed.

This is usually reported by users that are behind a cache, such as a cache in a firewall or a proxy server. Caches should pass incoming traffic as they receive it, but some caches will collect all the data before passing it on. The measurement then receives all the data all together at local network speed, and thinks you have a very fast connection. There's nothing I can do about this, it's a bug in your cache.

"The speedmeasurement only works if your browser supports JavaScript."

If you see this message when starting the measurement then the most likely cause is an ad-blocker on your computer. Turn the software off or add a rule to "trust" the Numion website. Some firewalls (for example the Norton Personal Firewall 2004) have a built-in ad-blocker.

Wrong date/time on the graphs

Set the clock on your computer to the correct date/time. There are many free programs available on the internet to automatically synchronise the clock in your computer with atomic clocks on the internet, see  Tucows for a collection.

"The page cannot be displayed"

This error may occur if you are trying to run the measurement while too many other programs are running. Your computer is full. Close as many programs as possible and try again.

 

Improvements

But I like my idea better. -- Calvin

Suggestions for improvements are always appreciated. Only with your help can I further improve this website. Please enter your comments, remarks, tips, and ideas on the  Support form.

Saving the measurements in a logfile

Due to security limitations on the internet it is impossible for me to write anything to your harddisk. And a good thing, too, or the internet would be even less secure than it is today. Sorry! You will have to save the measurements by transcribing them, or something. I could save the measurements on the server and make some kind of download button, but according to my estimates a single week's worth of measurements would eat up more than 60 megabyte!

A better graph

I've tried lot's of different graphs, all kinds of different ways of presenting the results. If you have visited this website in the past you will have seen them. In my view the current graphs are the best possible compromise between the technical possible, space on the screen, details visible, easily understood, and general usefulness. What can I say, I've done my best.

 

Throughput Tips

Most people are about as happy as they make up their minds to be -- Abe Lincoln.

• Check your expectations. Providers and modem manufacturers advertise the gross maximum speed of their system under ideal circumstances. Do not expect that to be your netto average speed (SurfSpeed measurement) or netto maximum speed (MaxSpeed measurement).
• Check your computer for viruses, spyware, adware, scumware, and such. Many viruses spread by sending email in the background. This takes bandwidth. Other kind of programs can similarly eat up bandwidth. If your speed suddenly drops then the first thing to do is check for infestations.
• Try the measurement again at a quiet moment, for example deep at night. If you get a significantly higher throughput then this proves two things: your PC and local connection are allright, and your provider has insufficient capacity for prime time.
• Download and install the newest drivers and firmware for your modem, router, ethernetcard, and other equipment to do with your connection.
• Run as few background programs as possible. This will free memory and CPU, making your PC faster. The cleaner your PC, the faster it'll be.
• Close instead of minimize. If the icon of a program is visible in the Windows taskbar then it's minimized, not stopped, and the program is still eating up memory and CPU.
• Buy more memory for your PC. It's cheap and on undersized computers can dramatically increase your surfing speed. On Windows 2000 and XP you can see the memory useage in the Task Manager.
• Try another browser. For example: Explorer is significantly faster than Netscape.
• If your provider has a proxy server then compare your speed with and without it. A proxy server is no guarantee for higher surfing speeds, it may be overloaded.
• Try another dialin-provider, such as a "free" provider. You won't get all the extra's, but I've seen free providers that were a lot faster than subscription providers.
• Tweaking your computer (changing system settings in the registry) is a bad idea. You can seriously damage your computer, and for most people the speed increase will be negligeable. The only way to significantly increase your internet speed is to get a faster (more expensive) subscription.
• Do not expect miracles from speed enhancement software. They simply cannot increase the speed of your connection, it's impossible. There are some subscription programs available that let you surf faster by compressing everything, but they have limitations and are not cheap.
• Advanced users may find the information from a  reverse traceroute (from the Numion server to you) useful.