Gordon Laing

From processor clocks to digital camera resolutions, the technology industry constantly plays a numbers game, using ever-increasing figures to encourage sales. And why not? Bigger numbers normally mean a superior product, right? Well, not always.

As we've discovered over the years, plenty of marketed numbers don't always translate that well into reality.

Take the previously quoted size of monitors for instance, which in recent years had to be changed to genuine viewable areas. I strongly hope that a similar truth is soon applied to networking technologies, wireless in particular.

Comparing quoted and actual network performance is not dissimilar to the difference between your gross and net salary. You're told one number, but reluctantly accept that you'll receive something at least one-quarter less.

Surprised? I was, but I discovered that this was a common reality. Most experienced networking folk understand, and reluctantly accept, a loss due to overheads.

Once various layers and protocols are taken into account, not to mention switching or hub performance, it's not uncommon to find a 100Mbits/sec wired network delivering around 60Mbits/sec in practice.

Something similar happens with wireless networking, where the 11Mbits/sec of the common 802.11b standard only delivers around 6Mbits/sec in practice.

While testing six wireless 802.11g routers something in me snapped. I was prepared for a significant performance difference between marketing and reality, but what really got me was the amount involved.

As hard as I tried, I just couldn't coax more than around 15Mbits/sec out of the recent and highly promoted 802.11g wireless networking standard - popularly described as offering both 54Mbits/sec and five times the speed of 802.11b.

Most networking folk reluctantly accept a performance loss of around 40 per cent, resulting in 802.11g speeds of 32Mbits/sec, but this was double what I measured.

Even when comparing speeds, I was only getting around three times faster performance than 802.11b. What was going on?

I decided to contact some manufacturers and ask them whether they felt describing 802.11g performance as being five times faster than 802.11b was at all misleading.

D-Link reckoned it wasn't misleading as the 15Mbits/sec to 20Mbits/sec we achieved was on average five times faster than what you'd get with 802.11b.

SMC admitted that, while bit speed really is five times faster, data throughput is not. It reckoned its Nitro technology could see greater than 20Mbits/sec in real-life environments.

US Robotics believed it was a little misleading and claimed that this is why it had not marketed its products this way. It too measured throughput in the region of 15Mbits/sec with standard 802.11g products.

That said, USR believes that its Accelerator Technology delivers around double the performance of normal 802.11g devices and it uses a figure of 100Mbits/sec on its marketing.

In our tests, the USR products were some of the fastest, but only delivered double the performance at longer distances by maintaining their speed while others halved. So it may have performed better than the rest over longer distances, but still delivered no more than 16Mbits/sec.

The interesting aspect was that all three companies described 802.11g data rates in the range of 15Mbits/sec to 20Mbits/sec. Couldn't anyone get it to go any faster? One solution put forward was to activate a feature called frame-bursting on both the client and access point.

Frame-bursting can increase data throughput by transferring packets continuously and, while supported by most 802.11g products, isn't usually enabled as standard.

I activated the feature on both the wireless client and access point and repeated the tests, but measured barely any difference. I tried activating the 'turbo' mode offered by most 802.11g access points which supports 802.11g clients only, but again it made virtually no difference.

Everyone's environments are different and this can greatly affect wireless performance, but from discussions I've had, it appears that few are squeezing much more than 20Mbits/sec out of 802.11g under default settings and normal conditions, and most are ending up with something closer to 15Mbits/sec.

Even taking the best-case scenario for 802.11g and a poor one for 802.11b, the difference remains closer to three times than the often quoted five times.

And remember, this is three times something that in practice runs about 40 per cent slower than quoted, which subsequently makes 802.11g a bit of a disappointment.

Don't get me wrong. I'm not saying avoid 802.11g. On the contrary, the prices of 802.11g kit have already fallen to a point where it's a no-brainer to choose them over 802.11b.

So the fact you're getting something around three times faster for little or no extra cost is a result - just perhaps not the result you were hoping for.

You may be luckier in your surroundings, but in our tests, 802.11g delivers 40 per cent of its quoted 54Mbits/sec at best, but more like around 25 per cent in practice.

While this is quicker than its predecessor and well worth having as a consequence, I reckon it's a technology ripe for some honest advertising.

Like this story? Spread the news by clicking below:

 del.icio.us     Digg this     reddit!