Why does wireless performance suck so bad...

[Lazybones]

So I setup my new router over the weekend, and it does perform better than the old one, but that is not saying much..

Is anyone else upset about how ridiculous the performance claims on wifi are compared to reality?

How is it they can get away with advertising wireless N (comes in many flavors) as being 300 Mbit in theory when real world averages between 20 and 70? Maybe peaking at 100Mbit?

If you want a sanity check on real world speed SmallNetbuilder has a benchmark database of routers they have tested
http://www.smallnetbuilder.com/lanwan/router-charts/bar/58-2_4-ghz-dn

I was able to push a peak of about 64Mbit with a 40-48 average out of my Asus RT-N16 witch is better than the netbuilder chart, but I am not using the standard firmware.

Also for those that do not know, N has several modes or flavors depending on your hardware
2.4Ghz single channel 20Mhz (150Mbit connect)
2.4Ghz dual channel 40 Mhz (300Mbit connect)

5Ghz single channel 20Mhz (150Mbit connect)
5Ghz dual channel 40 Mhz (300Mbit connect)

Note that the net builder charts still indicate that even 5Ghz dual channel mode on higher end units still is disappointing.

[Mr. Analog]

I often wonder how isolated these tests are, I don't know about most of you but there are a lot of conflicting sources within the same band at my house. At any given moment there are more than a dozen access points not to mention other sources of EM interference that could be causing latency issues.

The most frustrating part is that when moving in I could have had network cable run to all my rooms but time and motivation was low, looking back I should have had this done (I may still), I mean sure some devices like my DSi and my Xoom are wireless only but there are so many other devices now that could have worked over wireless but because the connection is so unreliable I simply ran cable to my laptop/netbook/DVD player.

At any rate, I still need to upgrade my router, the D-Link I have has some routing problems and there hasn't been a firmware upgrade since 2007...

[Tom]

N is best off on 5Ghz only, when you shut off the 2.4Ghz part all together. No G, no 2.4Ghz N. Nada. IIRC its called "Green Field Mode".

Also, to get 300Mbps, you have to have a MMIO router, with multiple radios and antennas. Many routers are not MMIO and can only get a max of 150Mbps theoretical. But like with G, the max in real life isn't likely to be near the theoretical max. I get about 20Mbps off my wifi, and thats about the max anyone can hope for (out of 54Mbps).


append: I would recommend not using the 2.4Ghz part of ANY N router. Keep a second router for G if you need G still.

[Lazybones]

I often wonder how isolated these tests are, I don't know about most of you but there are a lot of conflicting sources within the same band at my house.

The real world has interfering access points, that is the point... A saw an article that stated Intel only expected up to 100Mbit in ideal real world situations for its own cards. That is 3x less than the advertised 300mbit that IS NOT ACHIEVABLE short of maybe owning an interference free bunker with no other 2.4Ghz devices (bluetooth, cell, baby monitor).

So far my Asus RT-N16 is performing well with the TomatoUSB firmware, to early to recommend yet.

[Lazybones]

N is best off on 5Ghz only, when you shut off the 2.4Ghz part all together. No G, no 2.4Ghz N. Nada. IIRC its called "Green Field Mode".

Also, to get 300Mbps, you have to have a MMIO router, with multiple radios and antennas. Many routers are not MMIO and can only get a max of 150Mbps theoretical. But like with G, the max in real life isn't likely to be near the theoretical max. I get about 20Mbps off my wifi, and thats about the max anyone can hope for (out of 54Mbps).

You can't achieve 300 let alone 100 even at 5 Ghz and no other access points around. I have the necessary 3 antenna card 2.4 / 5Ghz card in my laptop and we have tested with dedicated enterprise access points that cost over $1500 each. We still achieved sub 100Mbit speeds.

[Tom]

N is best off on 5Ghz only, when you shut off the 2.4Ghz part all together. No G, no 2.4Ghz N. Nada. IIRC its called "Green Field Mode".

Also, to get 300Mbps, you have to have a MMIO router, with multiple radios and antennas. Many routers are not MMIO and can only get a max of 150Mbps theoretical. But like with G, the max in real life isn't likely to be near the theoretical max. I get about 20Mbps off my wifi, and thats about the max anyone can hope for (out of 54Mbps).

You can't achieve 300 let alone 100 even at 5 Ghz and no other access points around. I have the necessary 3 antenna card 2.4 / 5Ghz card in my laptop and we have tested with dedicated enterprise access points that cost over $1500 each. We still achieved sub 100Mbit speeds.

And you had the 2.4Ghz part completely disabled on the router/AP?

[Mr. Analog]

Lazy would have to build a faraday cage around his house, and unplug any other devices like wireless phones etc to even get close to the lab conditions used to test the max.

Perhaps people living outside of urban centres see speeds close to the maximums advertised, however I doubt even that would be perfect conditions :shrug:

[Tom]

If you don't have a router with dual radios, one for 2.4 and one for 5, you pretty much NEED to turn off 2.4 to get anything approaching theoretical. (as the one radio has to switch back and forth between 2.4 and 5 to do both bands).

[Mr. Analog]

I think there are more factors involved but I don't disagree with you.

Radio interference is a big problem...

[Tom]

I think there are more factors involved but I don't disagree with you.

Radio interference is a big problem...

The 5Ghz band isn't nearly as overloaded as the 2.4Ghz band. So interference isn't nearly as large of a problem (yet).

[Mr. Analog]

I think there are more factors involved but I don't disagree with you.

Radio interference is a big problem...

The 5Ghz band isn't nearly as overloaded as the 2.4Ghz band. So interference isn't nearly as large of a problem (yet).

It's true that there are less devices, but it's still a problem in high density areas.

Bottom line though is that the max speed you read in the spec is not ever going to be the max speed unless fairly specific conditions are met.

[Lazybones]

If you don't have a router with dual radios, one for 2.4 and one for 5, you pretty much NEED to turn off 2.4 to get anything approaching theoretical. (as the one radio has to switch back and forth between 2.4 and 5 to do both bands).

The 5Ghz setup was at work, using two different vendors access points priced between $500 and $1500. We ran the test at 5Ghz Dual band 40Hz with a likewise high end notebook with 3 antennas running Intels high end N chipset at the time.. The results where similar to the smallnetbuilder numbers, we never really even came close to 5Ghz, and we did several scans of the area and there where no nearby 5Ghz APs.

So spending loads on 5Ghz gear will only help to still get under 100Mbit speeds but that might be still better than nothing if you are saturated by near by 2.4Ghz access points. Some access is better than no access.

So again, the SPEEDS on WiFi hardware are total BS as you can't even break 100Mbit with anything in the real world.

[Lazybones]

Bottom line though is that the max speed you read in the spec is not ever going to be the max speed unless fairly specific conditions are met.

In the Case of the WiFi b / g /n specs, they arn't even in the same ball park...
Yes N devices are faster than G and those are faster than B, however the spec speeds keep getting further and further from reality.

[Mr. Analog]

Bottom line though is that the max speed you read in the spec is not ever going to be the max speed unless fairly specific conditions are met.

In the Case of the WiFi b / g /n specs, they arn't even in the same ball park...
Yes N devices are faster than G and those are faster than B, however the spec speeds keep getting further and further from reality.

The next step of course is Blast Processing with tons of migs and megs of memories

[Thorin]

The next step of course is Blast Processing with tons of migs and megs of memories

Ah yes, wireless via Russian fighter planes...