Promised broadband speed vs reality

[dj_jyno]

Again Challenging the average of 3.29Mbps quoted by Rudolph Muller in this [-]areticle[/] article:

I travel a lot and let me give you some of my results between April and December this year:

MTN:
Isando – 3.2Mbps
Croydon – 1.2Mbps
Greenstone Mall Edenvale – 1.2Mbps
Uvongo KZN – 700Kbps
Margate – 700Kbps
Port Shepstone - 300 Kbps
Durban CBD – 1.2 Mbps
Cape Town – Sea Point 1.2Kbps
Cape Town N1 City - 700 Kbps
Cape Town International Airport - 1.7 Mbps
Port Elizabeth airport - 1.5MKbps
Nelspruit – Airport – 700 Kbps
Vermont / Onrus River - 700 Kbps
Hermanus CBD - 700 Kbps
Somerset West – Somerset Mall - 700 Kbps
Phalaborwa airport – 450 Kbps
Phalaborwa CBD – 700 Kbps

Vodacom:
Croydon – 2.1 Mbps
Greenstone Mall Edenvale – 2.1MKbps
Uvongo KZN – 1.2Mbps (pocket loss too great – rendered useless in the end)
Margate - 700 Kbps
Port Shepstone - 300 Kbps
Durban CBD – 1.4 Mbps
Botha’s Hill - 300 Kbps
Cape Town – Sea Point - 1.7 Mbps
Cape Town N1 City – 1.2 Mbps
Cape Town International Airport - 3.1 Mbps
Port Elizabeth airport - 1.2 Mbps
Nelspruit – Airport - 700 Kbps
Vermont / Onrus River – 1.2 Mbps
Hermanus CBD – 1.2 Mbps
Somerset West – Somerset Mall – 1.2 Mbps
Phalaborwa airport – 950 Kbps
Phalaborwa CBD –1.2

As Jannie has mentioned, don't take this up with Rudolph, take it up with Ookla who supplied the statistics.

How busy were the two cells that I bolded in the above quote? The more subscribers in a specific 3G cell, the lower the average speed will be. You don't mention when these speeds were obtained - it could have been on a Saturday afternoon, when everyone was connected to 3G in the malls. At home on MTN I get a consistent 2+ Mbps connection.

 

[Electron1]

Perhaps Telkom's lack of capacity to upgrade Back haul links,
Exchanges and DSLAM's and their "legal" 20:1 contention ratio
has something to do with with ZA ranking so low?
We should pay 20:1 since we only get 20:1 speed from ADSL.

I see that 3G/cell has nearly the same speed as 10Mbps ADSL on speedtest.mybroadband.co.za 4.57Mpbs vs 4.76Mpbs.

Yup if only Telkom would take more notice of their backhaul links and upgrade the oversubscribed exchanges.
At least the Cell networks are busy upgrading their backhaul links......

 

[jannievanzyl]

That's not Jannie's point.

BTW - where I stay 8ta outperforms CrapC. (see how easy that is? )

We've got over 10 million data users, MTN probably close to that. So, in total there are close to 20 million people using data services in SA and (I'd guess) a few mil of them are 'proper' broadband users, i.e. the guys who'd run an Ookla speedtest.

Yet, people still seem to think their own, personal experience is indicative of the whole country. Don't we teach maths at school anymore?

Speedtest results are only relevant if you understand the COMPLETE test setup, otherwise the numbers are pretty meaningless. Unless you have a very large sample, like Ookla.

 

[jannievanzyl]

Hi Jannie; looking at the link... how's this possible?
(All ISPs)
2. MWEB Connect - 27.31 Mbps
3. Mobile Telephone Network - 26.00 Mbps

Only 700-odd test in that MWEB sample. Some naughty boy's been running core tests.

 

[MickeyD]

Only 700-odd test in that MWEB sample. Some naughty boy's been running core tests.

Yep... so easy to manipulate these tests

 

[jannievanzyl]

As Jannie has mentioned, don't take this up with Rudolph, take it up with Ookla who supplied the statistics.

How busy were the two cells that I bolded in the above quote? The more subscribers in a specific 3G cell, the lower the average speed will be. You don't mention when these speeds were obtained - it could have been on a Saturday afternoon, when everyone was connected to 3G in the malls. At home on MTN I get a consistent 2+ Mbps connection.

Yup, without the full test methodology published the absolute values are pretty meaningless. But the relative values might be relevant if the tests were done in very quick succession (within a minute or two) on the same test rig with just a SIM swap.

 

[jannievanzyl]

Yep... so easy to manipulate these tests

Especially if the sample size is small. With a few hundred thousand tests, it becomes much more difficult. In this set, Cell C's got 123k and Vodacom nearly 250k tests.

 

[MickeyD]

The other issue I have with ookla's tests is that in many cases folks (me included) only run a speed test when they suspect that something is wrong with their line.

 

[FleaDoctor]

If testing through fixed lines especially 4meg adsl lines,you also need to take into consideration line attentuation,noise etc.Often on a bad quality line one would only achieve between 2-3.5mbps anyway on a 4meg line.Unfortunately as technology advances so do our expectations,i would be more concerned about the download thresholds and misleading adverts etc by certain ISP's advertising uncapped internet with the tiny "* t's & c's apply" .*Telkom supply you with a fixed line capable of reaching 4meg (depending on line quality as stated above),the rest is down to your ISP.

 

[jannievanzyl]

The other issue I have with ookla's tests is that in many cases folks (me included) only run a speed test when they suspect that something is wrong with their line.

That and the fact that the modem-type (speed) is not taken into account. For example, on the Vodacom network, around half the modems are 3.6Mb/s or less. Even if these modems are used in a 14.4, 21 or 42M-capable area, they can never report speeds higher than 1.8 or 3.6.

Compare that to a network that launched in the last year or so with a minimum modem speed of 7.2. Such a network should actually show a much higher average than the network with a large number of sub-7.2 modems driving the average down.

As always, the results of these tests are very useful. As long as you understand the full environment.

 

[MickeyD]

That and the fact that the modem-type (speed) is not taken into account. For example, on the Vodacom network, around half the modems are 3.6Mb/s or less. Even if these modems are used in a 14.4, 21 or 42M-capable area, they can never report speeds higher than 1.8 or 3.6.

Compare that to a network that launched in the last year or so with a minimum modem speed of 7.2. Such a network should actually show a much higher average than the network with a large number of sub-7.2 modems driving the average down.

As always, the results of these tests are very useful. As long as you understand the full environment.

They only have themselves to blame for not pushing higher capacity modem sales. Many folks (that I know) used older modems as they were not willing to bind themselves to a 24 month contract on a start-up network. Plenty of E220's being used

 

[SaiyanZ]

The other issue I have with ookla's tests is that in many cases folks (me included) only run a speed test when they suspect that something is wrong with their line.

This would apply to the other countries as well so it is a moot point.

 

[jannievanzyl]

They only have themselves to blame for not pushing higher capacity modem sales. Many folks (that I know) used older modems as they were not willing to bind themselves to a 24 month contract on a start-up network. Plenty of E220's being used

Not really a question of pushing higher-speed modem sales. At least in our case. When those E220's were sold they were the best modems you could get.

Rather what we have here is the non-emotional relationship between a broadband user and his kit and thus a lack of desire for the newest. With phones, people tend to upgrade as quickly as possible to get the latest and greatest but with modems we don't see it. An E220 works well as you said, so why replace it?

Years ago we swapped every single 384K 3G modem for a 1.8M modem. If we did not do this, I'm sure many of them would still be active.

 

[MickeyD]

Not really a question of pushing higher-speed modem sales. At least in our case. When those E220's were sold they were the best modems you could get.

Rather what we have here is the non-emotional relationship between a broadband user and his kit and thus a lack of desire for the newest. With phones, people tend to upgrade as quickly as possible to get the latest and greatest but with modems we don't see it. An E220 works well as you said, so why replace it?

Years ago we swapped every single 384K 3G modem for a 1.8M modem. If we did not do this, I'm sure many of them would still be active.

True, their biggest drive was to attract customers to their network. As for the E220s, a certain company had to get rid of plenty of "old" stock (branding issues), so I obtained quite a few of them. Handed a few of them out here (in TK competitions !!!) and gave the kids one each and keep one in my briefcase, just in case!

My point was that if the new operator pushed modems capable of higher speeds, their tests would have looked better. Instead many of their subscribers are used 3.6Mbps modems.

 

[jannievanzyl]

True, their biggest drive was to attract customers to their network. As for the E220s, a certain company had to get rid of plenty of "old" stock (branding issues), so I obtained quite a few of them. Handed a few of them out here (in TK competitions !!!) and gave the kids one each and keep one in my briefcase, just in case!

My point was that if the new operator pushed modems capable of higher speeds, their tests would have looked better. Instead many of their subscribers are used 3.6Mbps modems.

Ah, got it.

Basically what we're saying is the black modems should be averaging much higher than the red, yellow or pink ones.

 

[TimTDP]

If I purchase a 21Mbps line, and am told this speed is available in a specific area, I expect it to perform at 21Mbps

I am not interested in excuses like too many people being connected at the same time, back haul limitations, distance from exchange, copper quality etc

It must just work. That is what I am paying for! Simple as that.

 

[MickeyD]

If I purchase a 21Mbps line, and am told this speed is available in a specific area, I expect it to perform at 21Mbps

I am not interested in excuses like too many people being connected at the same time, back haul limitations, distance from exchange, copper quality etc

It must just work. That is what I am paying for! Simple as that.

That's why nobody will offer you a 21Mbps service. It will be stated as "up to" 21Mbps.

Unless you are willing to pay for it... dedicated point-to-point service without guaranteed thoughput. It is available.

 

[brentmin]

Take it up with Ookla, it's their stats quoted here.

But I'd like to see your rationale where and how their testing methodology is flawed.

They say the proof of the pudding really lies in the eating - well I am eating. The question really is, "What is the end user experiencing when he takes his HSDPA dongle and connects up to the internet. Stats mean nothing. The problem with these stats is that we don't really know the test crityeria. It's easy for MUller to quotye stats as he does, but these need to be given along with the test parameters and criteria or they are pretty worthless. For all we know when the tests were done the tester was standing close to thew tower he was using and everyone else had been kicked off the tower so he can do his test. I wonder hopw many more people are reading this and saying what I am saying - that "this is not my expewrience.?"

 

[jannievanzyl]

They say the proof of the pudding really lies in the eating - well I am eating. The question really is, "What is the end user experiencing when he takes his HSDPA dongle and connects up to the internet. Stats mean nothing. The problem with these stats is that we don't really know the test crityeria. It's easy for MUller to quotye stats as he does, but these need to be given along with the test parameters and criteria or they are pretty worthless. For all we know when the tests were done the tester was standing close to thew tower he was using and everyone else had been kicked off the tower so he can do his test. I wonder hopw many more people are reading this and saying what I am saying - that "this is not my expewrience.?"

A few relevant points:

1. Basically you're saying your own results are "pretty worthless" as you did not give us your "test parameters and criteria". For all we know, when you did your tests you were miles away from the tower or used an old PC or a slow modem. From your results it does seem you are using a test setup with substantial limits. Did you do the tests outdoors or were you deep inside a shopping centre? Ask yourself why are so many of your tests across different networks reporting EXACTLY the same number. Points to something common being the limiting factor. And the only common parameter is your test rig.

2. These are not "Muller's stats". You would be well advised to acquaint yourself with the Ookla test setup as you really just show your ignorance about the subject, where the results came from and what they mean. Ookla's speedtest.net test servers are internationally recognised as the leaders in this field. Unlike your own test methodology, theirs is well published, right down to protocol level.

3. The 'tester' you refer to is in fact hundreds of thousands of people running millions of tests.

Results were obtained by analyzing test data between Nov 30, 2011 and Dec 29, 2011. Tests from 1,178,151 unique IPs have been taken in this country and of 4,469,380 total tests, 222,050 are being used for the current Index.

But you believe all these millions of people are all wrong and only your own results (which in your own words are "pretty worthless") are indicative of the complete broadband landscape in South Africa?