1

I have a Linksys WRT54G and I pay for a 12Mbps connection. I've been testing my connection using speedtest.net for many days and always get 8Mbps.

I called the support and they told me to bypass the router and test. I did it and got 16Mbps (much more than I pay for), so I thought "this guy just changed my speed so can he blame my router", and he blamed it. But to my surprise, everytime I bypass the router I get 16Mbps and when I use the router I get 8Mbps.

Is this guy trolling me somehow (configuring the VOIP-modem-stuff to different profiles depending o the MAC address connecting to it) or is my router a POS? How can I find out?

I don't know what's the thing the router connects to, it's a kind of VOIP adapter; the link is this one, but unfortunately I don't think you'll understand because it's in Portuguese. I know they can remotely connect to it, that's the origin of my conspiracy theory :)

I just tested wired to the router and got 10Mbps (and still 8Mbps on wifi and 16Mbps without router) O_o

I'm 5cm away from my router, so no obstacles to interfere, right?

------ UPDATE -------

It's a WRT54G V8, I'm using firmware v8.00.8.

Results:

IPerf LAN-LAN: 80Mbps

IPerf LAN-WLAN: 19Mbps (therefore we can ignore wireless issues/settings)

I wasn't able to make the (W)LAN-WAN NAT-enabled test with IPerf, I get a connection refused error. I'm not sure if did it right: ran in server mode, configured router to forward that port to my IP and tried to connect to my internet IP that got from this site.

I don't think there is a way to disable NAT using this firmware.

Question:

Let's suppose it's an underpowered hardware issue. Is it right to assume that custom firmwares could resolve the issue because they are possibly better implemented and would make better use of the router resources?

I couldn't find any references pointing to wired performance improvements with the use of custom firmware.

1
  • For the WAN to [W]LAN tests with NAT on, it sounds like you did it mostly right, except that it sounds like both your IPerf client and server on the LAN side of your AP. What I was going for would have been to plug your IPerf client machine to the WAN port of your AP (maybe give your AP a static IP address on the WAN port for this test, and put the IPerf client machine on a static IP on the same subnet). Then have it connect to the WAN IP address of your AP, which gets port forwarded to your IPerf server on the LAN side.
    – Spiff
    Mar 23, 2012 at 3:04

1 Answer 1

2

I think you could have any number of things going on here.

NAT adds latency, and home gateways are often underpowered

Note that pretty much all home gateways that do NAT are going to add some latency, and most home gear is extremely price-sensitive so they tend to use underpowered CPUs. I'd like to think that most home gear can do wire speed when just bridging (not doing NAT), so it might be interesting to do some LAN to LAN, LAN to WLAN, and (if the WRT54G supports it) bridge-only (non-NAT) WAN to LAN IPerf tests, plus a NAT mode WAN to LAN IPerf test, to really characterize the revision of the WRT54G you have, and the firmware image that's on it. Note that "wire[less] speed" over 802.11g is about 25mbps due to 802.11's extraordinary overhead (~50% of your signaling rate).

Your router is ancient

Note that the Linksys WRT54G was introduced in 2002, so it's a decade old, which, with Moore's law and all, makes it decrepitly ancient. Back in 2002, people's home broadband connections were probably just 1.5mbps down, 512kbps up DSL, so it didn't need to be all that fast. It's seen several hardware revisions since then, but I don't know if any of those revisions have had any effect on performance. More reason to run some clean tests to characterize the box you have.

Your client is too close to your AP

Putting your wireless client just 5cm from the AP is too close. You can actually overload a radio and temporarily desensitize it if the signal is too strong. Keep your Wi-Fi gear at least a meter away from each other when you can. Watch what your software reports for the signal strength. Use tool that reports true RSSI in negative dBm (deciBels relative to 1 milliWatt), not something that reports a meaningless percentage. If you ever see an RSSI about -20 dBm, your signal could be too hot for some cheaper radios. Good radios can handle RSSI's up to 0 dBm (== 1 full mW), but if you've never measured that aspect of your radio, figure -20 dBm to be the max signal strength it can handle. Also use software that lets you see which of 802.11g's signaling rates you're getting, and make sure to position your client where it is consistently getting the 54mbps signaling rate (it's okay if it occasionally dips down to the second-highest rate, which is 48mbps).

Your AP is probably on a polluted channel

Make sure you have your AP on the cleanest channel available to you. Don't just use a Wi-Fi-only tool like inSSIDer to find interference, because you might be dealing with non-Wi-Fi interference. If you don't have access to a real spectrum analyzer such as a Wi-Spy or better, then, try LAN to WLAN IPerf tests on channels 1, 6, and 11, and manually configure your AP for whichever channel you get the best performance on.

Your AP firmware probably isn't great

In case it's a firmware problem that can be fixed (as opposed to just underpowered hardware), if you don't have your WRT54G on the latest Linksys firmware revision for your hardware revision, try that. If after-market open-source third-party firmware distributions interest you, try the latest stable version of the distribution that appeals to you the most and runs on your hardware. Again, use a good performance tool like IPerf and only change one thing at a time so your results are as comparable as possible.

All or Several of the Above

Don't forget that the overall speed decrease you're seeing could be caused by a number of the above things in combination.

You're not being trolled

It's exceedingly unlikely that your ISP is trolling you. Remember Hanlon's razor.

1
  • Thank you very much for your answer. I updated my post with test results and more questions, can you please comment?
    – Roberto
    Mar 22, 2012 at 3:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .