How does cable modem power cycling improve the net speed?

Question

After a while (can be days, can be weeks), my network access speed slows down to ~2Mbps. Power cycling the modem (router and modem power off, then modem on, then router on) improves the performance to about 15Mbps. I'm curious what is the underlying reason for the slowdown...

Edit: Router and modem are both Linksys (CM100 and WRT320N), the OS is Windows 7. Provider is Comcast. I've had the same symptoms with the old hardware (Microsoft router, and rented RCA modem). If you search for "power cycling" you can find a lot of articles, speculating if it helps or not. In my case it clearly helps - but why?

Answer 1

Use SpeedGuide.net TCP/IP Analyzer.

This will analyze your connection's parameters and tell you what is the limiting speed possible for you. Compare the values you're getting from a time when the connection is fast and when it's slow. This might tell you where the difference is.

Also, your router probably has a status page. Compare also the values found on this page. If the router indicates that the connection-speed has dropped, then it's your ISP that's throttling you down. Sometimes the connection starts faster than what you paid for, then drops down to the correct speed.

If no difference is found, it will help to know what is your O/S, the make of your modem and its firmware version.

EDIT

See this article : Comcast Cable Modem-High Speed Internet?
I believe it discusses exactly your problem.

Answer 2

One possibility, depending on how a cable modem works, is that it re-negotiates its speed downwards during transient periods of high noise on the line, but is unwilling (or unable) to negotiate back up when it passes.

I've experienced something like this with DSL devices before - the web interface would be saying that it's connected at a much lower speed than normal, with poor line characteristics, but resetting the device would clear it up. On one occasion this was due to my neighbour having work done on his phone line.

But then, this is comparing apples to oranges; I've never seen a cable modem (NZ isn't big on cable TV), and have no idea how they work :)

Answer 3

I bought a Linksys WRT160N, and had to reboot it every three or four days; it would become unresponsive after that time. I flashed it to DD-WRT, and haven't rebooted it in two months.

Reliable software running over long periods of time is a difficult thing to do, not to mention creating reliable operating systems and networking software. It takes decades to get it right and reliable. The firmware in most consumer level routers do not begin to approach this level of testing and reliability.

Answer 4

I've heard that it may have to do with log file buffers on the devices. On a power cycle the logs are wiped and this freed space can then be used by the other functions on the device. I do not have any hard proof or evidence this is the case. Do any others have ideas regarding this?

Answer 5

Some ISPs, eg Comcast in the US have a Speedboost facility, whereby the first, I dunno 10MB of a download are boosted and then downloaded at a regular speed. Is it possible that this is happening?

Answer 6

From what you describe, there could be a problem with either the router or cable modem or both or even your OS.

• Do either get hot after a while? Do you leave them on all the time?
• How long does it take for performance to drop off after power cycling?
• Have you tried running with just the cable modem to rule out the router?
• Do you get the same degradation if you boot from a live CD?
• What type of things do you do with your net connection? Surfing/emails? Big/frequent downloads (and when)? Maybe you're affected by your ISP's traffic shaping policy?

Answer 7

I'm speculating here but from a hardware point of view, electronic devices do drift under increased temperatures. More so for analogue devices, which is what most communications gear is made of. However, the units should have been designed to work within accepted tolerances. Check to see if the temperatures are exceedingly high. If it is, something else might be wrong, causing the parts to overheat.