With 802.11n routers and notebooks (and some desktops) appearing everywhere, I wanted to know how much faster an 802.11n 2.4GHz network would be, when it is running alongside other 802.11g networks. Would I be able to set up an 802.11n network and have it serve just my media centre, for example, for video streaming, while leaving my 802.11g clients undisturbed?
So I did some (non-scientific) speed tests on a Linksys WRT160Nv3 (dd-wrt.v24-13309_NEWD-2_K2.6_mini_wrt160nv3), when it was running alongside a WRT54GS (tomato 1.25) router. The WRT160Nv3 was configured with 40MHz (upper) channels to maximize speed, with the WRT54GS was on channel 10 to avoid overlap as much as possible.
I tested 3 different cases:
Configuration 1: Intel 5300 copying files over WRT160Nv3 (40 MHz upper)
Configuration 2: Intel 3945ABG copying files over WRT54GS (ch10, mixed bg mode)
Configuration 3: Intel 5300 copying files over WRT160Nv3 (40 MHz upper) while Intel 3945ABG copies files over WRT54GS (ch10, mixed bg mode)
That is to say, I wanted to collect some data on how this 802.11n network might co-exist with an 802.11g network in the 2.4GHz spectrum, and what the impact would be on performance.
Configuration 1: Intel 5300 copying files over WRT160Nv3 (40 MHz upper)
Elapsed time 1 min 22.0 sec
Incoming Outgoing
------------------------------------ ---------------- ----------------
Total of data transferred 363.1 MB 7.4 MB
Maximum transfer rate 6.1 MB/s 127.7 KB/s
Average transfer rate 4.4 MB/s 92.0 KB/s
Configuration 2: Intel 3945ABG copying files over WRT54GS (ch10, mixed bg mode)
Elapsed time 1 min 22.2 sec
Incoming Outgoing
------------------------------------ ---------------- ----------------
Total of data transferred 191.0 MB 3.8 MB
Maximum transfer rate 2.6 MB/s 52.8 KB/s
Average transfer rate 2.3 MB/s 47.7 KB/s
Configuration 3: Intel 5300 copying files over WRT160Nv3 (40 MHz upper) while Intel 3945ABG copies files over WRT54GS (ch10, mixed bg mode)
Intel 5300 / WRT160Nv3 simultaneous (ch5-2.432 40MHz upper)
Elapsed time 1 min 58.1 sec
Incoming Outgoing
------------------------------------ ---------------- ----------------
Total of data transferred 532.8 MB 11.1 MB
Maximum transfer rate 6.7 MB/s 142.2 KB/s
Average transfer rate 4.5 MB/s 96.2 KB/s
Intel 3945ABG / WRT54GS CGN1 simultaneous (ch10-2.457 mixed bg)
Elapsed time 1 min 55.9 sec
Incoming Outgoing
------------------------------------ ---------------- ----------------
Total of data transferred 49.6 MB 1.2 MB
Maximum transfer rate 1.4 MB/s 33.7 KB/s
Average transfer rate 438.4 KB/s 10.3 KB/s
Conclusion:
Even on 2.4GHz, 802.11n with 40MHz channel width is roughly double the speed of an 802.11g network. But when an N network is put next to a G network in the 2.4GHz space, the G clients lose 75% or more of the performance on the average case, while N clients are virtually unaffected; your N clients will run at full speed, and G clients can barely surf the web.
Anecdotally: I'm seeing similar performance degradation when an 802.11g client is connected to an 802.11n 2.4GHz AP. The 802.11n clients run fast when it's a pure 802.11n network. But as soon as some 802.11g clients show up, compatibility mode kicks in, and the 802.11g clients only get around 500-700kb/s, which is still at least a 50% performance hit, and the 802.11n clients lose around 20% of the speed. (Yes, this paragraph is completely unsubstantiated. That's why it starts with "anecdotally"...)
Recommendation:
Based on my (extremely) unscientific tests, I'd say that if you really want a faster wireless connection, keep your 802.11bg clients on your G access point, and get a 5GHz 802.11n access point for your faster N clients. Mixing clients on an AP seems to ruin things for everyone involved. Even putting a separate 802.11n 2.4GHz AP next to your 802.11g network doesn't help the 802.11g clients.
Or, just replace all your clients with 802.11n cards, and run a 2.4GHz 802.11n network for a nice doubling of network throughput.