Unable to take advantage of high speed internet

"Tom Barreca" hath wroth:

What happens when you plug directly into your router with an ethernet cable (i.e. no wireless)? That will determine if it's a router problem or a wireless problem.

That's nice. Is there some reason why you're not disclosing the model number? You did so well with the HP laptop and internet provider. Yet, it seems that the selection of wireless equipment is such an embarrassment, that many users will do anything to hide the details. I realize that it might be difficult, but could I trouble you to reveal the model number? I won't tell anyone you have this model.

I'll assume that you're checking the speed on the HP zv6270us laptop. Be advised that the speed often changes when you start moving traffic. In general, if there are errors, the speed will slow down. When traffic stops, it sometimes goes back to 54Mbits/sec. What the customer usually sees is only the starting and ending speeds. It makes it look like you have a wonderful 54Mbit/sec connection, when in reality, you only get that when it's NOT moving data.

These drivers?

Did you also install all the latest Windoze updates? (Just checking).

Sure. It would have been much easier if you had bothered to supply the DLink model number. So, you get generalities instead:

My guess(tm) is that your unspecified model DLink router might be gutless and can't handle the speed. See the table at:

and note that a large number of routers can't go fast enough. Is your router among these?

There's also a possibility it's a wireless problem. See table at:

At best, you'll get about 24Mbits/sec with a 54Mbit/sec connection. If you have 802.11b compatibility enabled, and have an older DLink router, you'll get at best 14 Mbits/sec thruput. If you have any form of interference or reflections, you'll get even less. However,

7Mbits/sec is positively dismal. The problem will be figuring out why it's so slow. First, try an ethernet cable to the router and see if the router is slowing you down. I consider this highly likely. If the DLink can deliver the same performance as a direct connection to your cable modem, then come back and I'll suggest some wireless setting changes and tests. One thing you can easily do is try a different channel (1, 6 or 11) in your DLink, in case it's some manner of interference that's slowing you down.

Jeff,

Thank you very much for your help.

When I connect directly to the cable modem with an Ethernet cable, I get 25 to 27 mbps. But when I connect via the wireless network, I only get 3 to 7 mbps.

The D-Link 802.11g/2.4 GHz wireless router is model # DI-524. It's not on that list that you suggested that I look at.

I had previously downloaded a new driver for the Broadcom NIC by finding it on the list in Device Manager, right clicking on it, selecting properties and selecting update driver. I went to the website you suggested and downloaded a new driver again. It didn't seem to increase the speed.

I assume that all windows updates have been previously downloaded. My computer is configured to automatically download updates, and periodically it gives me a message that it is doing just that.

The router was connected to channel 6. I tried channels 1 and 11 as you recommended, but that didn't make any significant difference.

When I connect the computer directly to the wireless router with an Ethernet cable, I get a significant increase in speed, 12.72 Mbps, but that's still less than half of what I get when I connect directly to the cable modem with an Ethernet cable.

Do you think that a new router would do the trick? If so, do you have any recommendations?

Thank you in advance. I really appreciate your help.

Tom

On Fri, 16 Feb 2007 13:51:10 -0500, "Tom Barreca" wrote in :

When the wireless distance is only (say) 5 feet or so? If so, then interference is the likely culprit. Try different channels (1, 6, 11). Or see Interference in the wikis below. Otherwise you may need a stronger signal, by means of a better antenna or cheap reflector.

First thing to do is set your wireless access point to "G" only and not both B and G modes. If your access point detects any B clients connected to it it will operate in the B mode for most communications.

Second even though the over the air data rate is 54 mbits per second you can never get anywhere near that. The 802.11 equipment is all half duplex and has other issues as well that limit the bandwidth such as beacons that are sent out as well as periods where both the client or access point has to listen for traffic or to switch from transmit to receive. 802.11 equipment acts as a wirelss hub and not a wireless switch. They are subject to constant collisions which limit throughput. The faster the rate that traffic is trying to be sent the higher the number of collisions.

Even good quality equipment such as Cisco 802.11 G acess points model

1240 series will have a very difficult time getting more than about 15 mbits per second even over very short distances. The best case I have seen is about a 20 mbits per second rate in one direction under optimal conditions.

Eternal praise and gratitude are always graciously accepted.

Ok. It's not the ISP or cable modem. That leaves the DI-524 router or wireless section.

Bummer. The DI-524 is not the worlds greatest router. However, it should not be slowing you down that far.

Oh well, it was worth a try.

Do you have a spare computer handy that has an ethernet connection? If so, I would like to try some benchmarks without going through the router section using IPerf:

To start, connect both computers to the DI-524 using CAT5 ethernet cable. Run the server on the spare computer as: iperf -s on one machine. Run: iperf -c ip_address_of_server You should get a report of the download speed via 100baseTX. It should be at least 80Mbits/sec. That will prove that the spare machine is not a limiting factor and that both machines have enough horsepower to run the test.

Next, disconnect the ethernet cable from the wireless laptop and connect via wireless. Run the test again. If you're really getting a

54Mbits/sec connection, you should get about 25Mbit/sec thruput. If there are 802.11b devices interfering with your DI-524, then perhaps 17Mbits/sec maximum.

We can get fancy with the options (i.e. bi-directional, UDP instead of TCP, longer duration, multiple streams, etc) later.

If you still get 3-7Mbits/sec, then you need to try some changes. John brought up the possibility of interefence. Try different channels (1, 6, 11). Try isolating the radios from a view of the city such as away from windows. Try it in the same room where you're guaranteed to have a good signal. Try turning off 802.11b compatibility "mixed mode" (unless you really need it).

That only gets the "critical" updates and not the optional updates. It doesn't get the optional or hardware updates. Apparently preformance issues are not considered critical. Try going to:

IE6 or IE7 -> Tools -> Windoze Updates

If it mumbles something about downloading the new and improved "Microsoft Update", do it. It's a bit of an ordeal but it's worth the effort slogging through the legal disclaimers and dumb questions.

Ok, then's it might not be interference. One gotcha that's becoming more apparent is interference from municipal wireless mesh networks. If you have one nearby, be advised that they tend to run high power (1 watt) but usually don't hog all the channels. Look out the window to check.

Other sources of interference:

See if any apply.

Oh-oh. It appears that the DI-524 is not capeable of going fast enough. That's why I mentioned that the DI-524 is not my favorite router. I couldn't find any reviews on SmallNetBuilder.com on the DI-524 that have benchmark tests.

If you feel ambitious, use the method I described with IPerf to measure the WAN to LAN throughput. Basically, setup your space PC on the WAN port, with a static IP address in both the PC and the WAN side of the router. The LAN side gets the other machine with a CAT5 ethernet cable. No internet and no wireless. I you get about

13Mbits/sec thruput with IPerf, the problem is that the router is gutless.

Yes. However, I'm not going to guarantee it. I have no customers or experience with 30Mbit/sec systems.

Obviously, one of the routers on this list that can do more then

30Mbits/sec WAN to LAN.

I really like the Buffalo WHR-HP-G54 (this week) but mostly because I can install DD-WRT alternative firmware. The problem is that I don't know if it can do 30Mbits/sec. I can try simulating it with Iperf this weekend, but only if you say you're interested in the results. I've got way too many projects pending.

Thank you very much for your reply.

How do I set my wireless access point to "G" only?

Tom

"Tom Barreca" hath wroth:

See:

It's the line with "G mode" and "Mixed" mode. Mixed mode means it will accept 802.11g and 802.11b connections. "G mode" means "G" only.

Thanks again for all the advice!

I changed it to "G" mode only, and I'm getting 7.3 Mbps, a slight increase, but still not as fast as I think I should be getting. When I have a little more time, I'll perform the diagnostic test that Jeff recommended in a previous post. If that doesn't work, I'll either have to get a new router, or just go back to the ordinary Optonline service. No reason to pay extra for the "Boost" service if I can't utilize it!

Thanks again,

Tom

Well, some is weird here. You get: 27 Mbits/sec connected to the cable modem. 17 Mbits/sec connected via ethernet to the DI-524 router 7 Mbits/sec connected via wireless. I would expect the directly speed connected to the router either via wireless or CAT5 to be exactly the same. At least that's what I see when I do internet download speed tests on various router.

Never mind, I see what's happening. When the WAN->LAN part of the router is hogging all the CPU cycles, there's nothing left for the wireless section. So, it slows down even more. I never see that on my tests because I never seem to run into any 30Mbit/sec internet connections or routers that are speed limited at that speed.

If I have time, I'll setup a similar simulation test at home and try a few routers I have floating around. It's a very simple test, where most of the effort is in the initial setup. It would be nice to have my own numbers and see what my routers can do.

Also, my guess(tm) is that you're heading for a new router purchase.

Jeff Liebermann hath wroth:

I hate days like this (when things don't work in accordance to my view of reality). This should have been easy.

The major effort was not the initial setup. It's turning into a troubleshooting session. I setup a Buffalo WHR-HP-G54 wireless router with DD-WRT v23 SP2 (09/15/06). I've used a bunch of these before, so this should be a no-brainer. Setup as below:

PIII/1000 W2K desktop 10.0.0.11 IP Address 255.255.255.0 Netmask none Gateway

Buffalo WHR-HP-G54 10.0.0.1 WAN IP (static assigned) 255.255.255.0 WAN Netmask 10.0.0.1 WAN Gateway 192.168.1.1 LAN IP address 255.255.255.0 LAN Netmask

HP ze2000 laptop with Broadcom wireless. 192.168.1.xxx IP address (DHCP assigned) 255.255.255.0 Netmask 192.168.1.1 Gateway

I run the IPerf server on the desktop PC with: iperf -s and run the IPerf client on the laptop with: iperf -c 10.0.0.11

With a wireless 54Mbits/sec connection, I get 23Mbits/sec which is normal. If I setup the DMZ to point to this IP address of the HP (it can change), I get 24Mbit/sec in both directions, which is normal.

Using just the ethernet switch section, I get about 90Mbit/sec thruput in both directions. That's normal for 100baseTX.

However, when I disable the wireless (using the on-off push button on the laptop) and connect to the Buffalo router with an ethernet cable, the best I can do is 7.8Mbits/sec. That's about 1/10th of what I would expect. UDP is even worse than TCP with 1.05Mbits/sec. Something is very wrong here.

The problem is that I can't seem to figure out what's broken. I've tried disabling SPI as suggested in:

played with various Iperf options, checked my cables, but I still get awful performance. I'll try some other software, methods, computers and routers (later), but I'm mystified. Anyone have some clues, hints or answers?

Jeff Liebermann hath wroth:

Never mind. Idiot error on my part. I had my laptop ethernet card set to 10baseT-HDX (as I usually do when I connect through marginal ethernet wiring system) and forgot to put it back to auto.

After switching to 100baseTX-FDX and juggling routers, I now get WAN->LAN: WHR-HP-G54 28.0 Mbits/sec WRT54G v1.1 16.0 Mbits/sec BEFW11S4v4 19.5 Mbits/sec (The first two have identical versions of DD-WRT v23 SP2 (09/15/06) installed.) Better, but not really good enough for 30Mbits/sec broadband.

I'll try to nail down the numbers a bit better, but it appears that even my favorite routers are not going to break the required

30Mbits/sec. There are also a few more routers in the pile.

Meanwhile, UDP is still giving me about 1Mbit/sec WAN->LAN which is all wrong. Also, repeated IPerf tests of the WHR-HP-G54 yield random results varying from 23 to 29 Mbits/sec. More troubleshooting. Grumble...

Jeff Liebermann hath wroth: (...)

Some more numbers and some surprises. Both routers have DD-WRT v23 SP2 (09/15/06) installed. The line with "mixed mode with b connection" is the results from an 802.11b connection from my XV6700 cell phone, but not moving any traffic. Routers were rebooted before and after this part of the test.

PIII/1GHz acting as server on the WAN port running: iperf -s HP Ze2000 laptop as client on a LAN port running: iperf -c 10.0.0.1

WRT54Gv1.1 WHR-HP-G54 Mbits/sec Mbits/sec

Ethernet WAN->LAN 18.5 28.5 Loopback 443 (on desktop)

Afterburner and Frame Burst Enabled Mixed Mode (g+b) 18.3 26.2 G only 16.8 25.5 Mixed with b connection 17.6 21.1 B only 5.1 5.5

Afterburner and Frame Burst Disabled Mixed Mode (g+b) 17.2 25.4 G only 18.0 26.0 Mixed with b connection 14.9 19.9 B only 5.8 5.8

Some other routers: BEFW11S4v4 Network Anywhere NR041 Ethernet WAN->LAN 19.5 10.9

802.11b at 11Mbits/sec 5.5 (no wireless)

Performance without the router section (wireless only acting as an access point (LAN->Wireless) and direct connected CAT5.

WRT54Gv1.1 WHR-HP-G54 Mbits/sec Mbits/sec LAN->Wireless 23.5 23.8 LAN->LAN(ethernet only) 90 92

Notes:

I was getting really erratic results with both wireless routers for quite a while. I was only 2 ft away from the router and suspected I might be too close. So, I moved back to about 7 ft away and found that the numbers are more stable.

Conclusions:

There doesn't seem to be a huge penalty for using Mixed mode (802.11b and 802.11g) even with an 802.11b client connected. So much for that theory.

Turning on/off Afterburner and Frame Burst (defaults are off) also has a very small effect on performance. So much for that theory.

Overall WAN->LAN performance is marginal and will be a problem with

30Mbit/sec "burstable" cable modem broadband. The router section appears to be the limiting factor as the wireless section is somewhat faster than the router section.

This is also a great reason to get a wifi access point that's separate from the router itself. Spread the workload.

Yep, that's a dirty little secret about the soho-class routers. Their performance across the lan/wan router is really, really bad.

Here's a comparison:

-Bill Kearney

"Bill Kearney" hath wroth:

Yep. It's far worse than I expected for current hardware. Naw, I'm being too nice. It positively sucks. I never noticed the problem because the local cable system is only 6Mbit/sec. Most of the routers listed would be just fine at that speed. However, for a 30Mbit/sec broadband service, only the recent MIMO and Turbo infested routers seem to have the necessary horsepower.

My IPerf test numbers for my WRT54G v1.1 (about 18.5mbits/sec) come sufficiently close to their WRT54G v1 IxChariot results (20mbits/sec). Good to have a sanity check.

What also bugged me was the inability to obtain fairly repeatable WAN->wireless test results. The default 10 second IPerf test resulted in about a 30% spread in values. Increasing this to 1 minutes was only slightly better. What was odd is that it would always start out with the fastest numbers, and then slow down with repeated tests. After about 10-15 minutes of repeated testing, it would magically start over again at the highest speed. The only way I could get repeatable numbers was to power cycle the router, run a trial IPerf test (10 seconds), and then run the real test. The connection speed indicated at the WZC client was 54Mbits/sec but that can be a lie. I tried it again with the speed locked in the Buffalo WHR-HP-G54 router at 54Mbits/sec and got the same results. It's not 802.11b compatibility (mixed mode) slowdown effects because I did the tests in

802.11g only mode. It might be something at my location (interference, reflections, RFI, etc) so I'll try again elsewhere. Weird.

It would also be interesting to test a Buffalo WHR-HP-G54 with the stock firmware.

Here's an update - I purchased a new router - a Linksys WRT54GS - Installation was very easy via the setup wizard. If I run an Ethernet wire directly to the router, I get 25mbps, which is excellent. When I go wireless however, I only get 9-11 mbps, a nice improvement over the old router, but not as good as I had hoped.

I do have one additional issue with this new router. The wireless network works fine, but my old backup computer, which is located adjacent to the router and runs Windows 98, will not work with a wired connection to the router. I cannot use it with a wireless connection because the computer is too old for that. I called customer support, and they told me that the router has issues with Windows 98. He tried to walk me through the process of connecting it to the router, but he just went through the same steps that the setup wizard did. When he couldn't get it to work, he said that it must be an equipment problem. But I had no problem with this computer when it was attached to the old router. This computer is located adjacent to the cable modem, so I could wire it directly to the cable modem, but unfortunately, the cable modem only has one output, which is being used for the router. My main computer is the wireless laptop, so I rarely use this old computer, but I really would like to be able to use the internet from this old computer.

Tom

What hardware version of the WRT54GS? The v5 and v6 mutations are marginal. See the version number on the serial number label or use the serial number to decode the hardware version:

Something is wrong. 25Mbits/sec is limited by the broadband connection. However, if you have a 54Mbits/sec wireless connection (check the client status), then you should get about 25Mbits/sec thruput. 9-11 isn't much better than the 7 you were getting before. Something might be wrong with the setup, the way you're testing, the location (interference), or with your client computer. Hard to tell from here.

Do the front panel lights turn on when you plug in the ethernet cable? I have several Windoze 98/ME boxes around and they all work fine with both wired and wireless connections. Check your TCP/IP settings (should be IP address and DNS assigned by ISP (router)).

So, get an old wireless card. I still use my Orinioco Silver PCMCIA cards on the older laptops. However, that's with Windoze 98SE (second edition). The first release was fairly awful.

Translation: They don't have a clue, don't have an answer, and want you off their phone line. However, that might be a hint to update the WRT54GS firmware to the latest version.

Gaaaak. I don't use the setup wizard.

Unplug the router from the cable modem. Plug in the Windoze 98 laptop. If the laptop is setup for DHCP and the cable service provider does DHCP, and there's MAC (cloning) authentication issues, then you should be able to connect with the Windoze 98 laptop.

It should work. Make, model, ethernet card maker, exact Win98 version, ipconfig output, etc would be nice. Also, look on the "log" output page in the router to see if there are any error messages when connecting with the router.

OK, I got the old computer connected to the network. It was really stupid. I was following the suggestions of the customer support guy, so I did everything that he suggested, no more, no less. After I got off the phone with him I relaxed a bit, had some dinner, and then went back to work on the problem. The first thing I did was the old control-alt-delete. Son of a gun if that didn't do the trick! What's the matter with that guy that he didn't suggest that? What's the matter with me that I didn't try that before I called him?

Anyway, I still have the problem of a slow wireless network, even with the new router. The new router is Version 6.

Tom

After installing my new router, I didn't really notice a real improvement in my download speed. So I got a new Network Interface Card (Linksys WPC54GS v.2) which is the companion NIC to the router that I got. It was an easy installation - the setup CD that comes with it is very easy to use. At first, it made no difference to the speed, but then I had the bright idea to go into the device manager and disable the broadcom NIC that came installed with the laptop, Presto, my download speed jumped up to 17-19 Mbps. Would it help if I disabled any of the other network adapters? In the device manager, there's a 1394 Net Adapter, the Broadcom 802.11 b/g that I already disabled, A Realtec RTL 8139/810x Family Fast Ethernet NIC (which I believe is the NIC for when I'm using a wired connection), and the new Linksys that I just installed. Any other suggestions for any tweaks that might improve my speed even more?

Thanks for your help.

Tom