As discussed on Byteside Tech #1, we recently conducted a series of network data tests using the iPhone’s Speedtest.net app.
We had concurrent access to four iPhone 3GS handsets, one on each of the four Australian networks — Telstra, Optus, Vodafone, and 3. Travelling around the Sydney CBD and Sydney suburban areas, we ran close to 150 individual speed tests. Tests ranged from Manly to Homebush, Annandale to North Sydney, and plenty in between.
At each location tested, we ran four tests in rapid succession to try and get a reasonably accurate data set at each location while still pushing onward with the test as quickly as possible. Rapid successive tests would also, in theory, minimise excuses for high latency due to packet connection initiation times.
UPDATE: One thing we forgot to mention is that everywhere we tested we only ran the test when we could find full bars of coverage strength for all networks. So we were never testing good signal versus bad. Data strength may have had its own fluctuations in availability, but for most purposes we felt watching the standard signal meter was the best way to ensure we aimed for a level playing field at all locations.
Note there are almost 40 less data points in the Optus list than the other three. We had a data problem and lost most of our recordings from the first day of testing. But things did perform in a similar fashion on both days, so we’re confident this had very little impact on the overall statistics. You can see in the raw data the exact parallel tests if you want to purely compare test for test.
RESULTS
The summary? You really do get what you pay for. Telstra blows away the competition with speeds almost double the next best network, while the best value cheapest data network Three delivered the slowest results. Indeed Three’s best single result was slower than Telstra’s average.
All that said, even the slowest network delivered reasonable test results. In the context of landline access, wireless speeds were better than a home 512K connection.
Other notable concerns were the very high pings experienced across all networks. If you need low latency, wireless has a very long way to go.
We should note that of the 110 Optus tests conducted, no result could be obtained on nine occasions. And this was a similar experience on the first day’s missing Optus results. So while their final results look quite good, there was close to a 10% failure experience during the period of testing. We marked these results as ‘zero’ for down and up, and left the ping stat blank.
KEY STATISTICS
| Network | Stat | Down | Up | Ping |
|---|---|---|---|---|
| Telstra | Average | 2681 | 263 | 2992 |
| Median | 2416 | 280 | 2358 | |
| Best | 6151 | 335 | 212 | |
| Optus | Average | 1637 | 218 | 4493 |
| Median | 1903 | 279 | 2636 | |
| Best | 3654 | 359 | 214 | |
| Vodafone | Average | 1283 | 140 | 7489 |
| Median | 1125 | 107 | 5398 | |
| Best | 3349 | 357 | 209 | |
| Three | Average | 629 | 128 | 2211 |
| Median | 525.5 | 128 | 2247 | |
| Best | 2290 | 315 | 206 |
For all raw data, which includes date, time and GPS coordinates plus more statistical anaylsis, grab the XLS here.

Subscribe
Were you surprised by the result or did you think the Next G network would be the fastest network?
No question that we expected Next G to come out in front. But by such a significant amount? That was a surprise. Definitely makes paying the premium worthwhile for certain kinds of users.
I appreciate your effort and results, but I’m sorry – no interwebs speeds are worth handing your money over to Telstra for.
As a Telstra employee, we pride ourselves on the works best in more places slogan. If you are referring to the iphone in particular, given this phone is a high data unit, I wouldn’t use it outside the Metro area in your case. Like Telstra or dislike Telstra, be it as it may, the network is unsurpassed and as such, you get what you pay for. You wouldn’t pay $50’000 for a mini moke would you? Why would anyone expect Telstra to charge a pittance for the utilisation of the best network in the country. All the best and happy browsing.
seems strange that you would be surprised by the NextG to Optus margin, the optus network is not capable of even half the speed of the NextG network.
If the chip in the iPhone supported the full speed of the NextG network we would be seeing bursts just under four times higher,
yet these tests represent the best the other networks can provide.
What unit are you using for ping? It can’t be seconds, can it?
One big problem with this whole survey is the test web site, http://www.speedtest.net. Sure it is an easy site to use and seems to give statistical information but it really isn’t suitable for measuring the speed of HSPA networks as it reports the peak speed and not an average over a period of time.
The next real problem as stated above is the platform used for this survey, the iPhone. The iPhone only supports 7.2Mbps down and 384Kbps up. This in particular will not show the upload advantage provided by Telstra and Optus, who can provide up to 2Mbps in many locations. Telstra may even have some site support 5.7Mbps. Of course Telstra also has some sites that may provide up to 12Mbps or so downloads.
One other limitation with this survey is the time of day. The greatest congestion on the networks occurs between 8pm and 10pm on Monday to Thursday. Many of the networks are unuseable at these times including Telstra in some locations.
It is like with everything, you pay a premium for the best, Telstra made a huge investment upgrading their network (around 2.5 billion) when the second largest Ausie telco just made an investment of 350 mil, it is reflected in the performance of the network. Maybe a teenager using his/her iphone to play wouldn’t mind, but for a business availability is critical… There is no brainier here.
Good work guys, but I think you missed a real opportunity here. What you should have done (or do in future) is have thrown it open to all users to submit speedtest.app results to you for collation – with this rich network of data, you could have built a ‘speed topography’ of Sydney (and wider) for each supplier (and the average) – how amazing would that be. Furthermore, to require five bars is unrealistic – coverage is just as significant an issue as speed.
Brilliant, if your in sydney. Is there a possibility for a test to see who (Telstra, Optus, Vodafone and Three) has coverage in most areas and not just the area’s acrording to the providers website, i bet that will make an impact on peoples views.
Were the accounts the tests were done on under contract or pre-paid?
Some providers use different APNs for pre-paid, with different network quality.
The pings are pretty slow on all the networks. I'm on Voda in country Victoria, and it isn't that bad.
Not sure if it would make a big difference but pretty sure the speedtest.net app's server is hosted by Optus, giving them a potential advantage over Vodafone.
Were there different views of the apps store when you were testing between the carrier locked devices?