[David Lee]

Let me pose a technical question for someone with a lot more knowledge than I have about ethernet networks. I have an iBook G4 (933 MHz) and an old Quicksilver updated with a Giga Designs 800MHz CPU running at 867MHz. The problem I have is that the ethernet speeds on the iBook are very slow compared with the Quicksilver, while both are hooked up by ethernet cables to a 100 base router, which is in turn hooked up to a 12MB ASDL modem. The difference is especially noticeable when using CaminoKnight to download the latest nightly version of the Camino application, with a difference of at least 4 -6 times slower on the iBook. The difference is also obvious downloading big files from the net, such as an Apple updater file. Transfer speeds from either machine to the other are roughly the same. Both machines are set to automatic in the network/ethernet preference pane. Anyone have any ideas as to why this happens?

[ghporter]

Settings on the iBook are different, obviously. Use one of the many tools available online to check out your configuration, starting with the "Tweak Tester II" at BroadbandReports.com. Test both computers and see where the iBook's settings are different from the Quicksilver's. There are also a couple of things to look at like are you using the Quicksilver to authenticate your ADSL connection?

[David Lee]

Thanks for the observations. Tried the page you suggested: results for the iBook were; "upload packet less (1488) than your MTU" (set to 1500 by OS X by default but I can set it to less in manual mode). and "Good data stream" (no/few rexmits). "RWIN now set to 65535 Choose RWIN between 113880 and 300760" (never seen a setting for this in the networking pref pane, must do this with terminal?) All of the tweakable settings were the same for the Quicksilver and the iBook. The results were completely different.
Quicksilver transfer rate: 139352 bytes/sec 1114 kbits/sec
iBook transfer rate: 22524 bytes/sec 198 kbits/sec
So that proves what I have suspected, now how do I go about resolving this? Any suggestions?

[ghporter]

Your MTU (maximum transmission unit) can be very important to your speed. I'm not sure but it looks like your MTU is too high, and thus you're fragmenting packets-and slowing yourself down. Try setting your MTU to 1488 and see what happens.

There is a GUI app (rather old but still effective) that will let you alter a lot of parameters. It's called RMAC, and though the documentation is in Japanese, it's pretty easy to use without the docs-I'd hazard to guess that you won't have any problems reading the docs, considering your location . Find it here about a third of the way down the page on BroadbandReports.com's OS X Tweaking FAQ.

[David Lee]

Thanks again. I tried RMAC, but it errors out saying there is a PERL error. It was designed to run on 10.2 so maybe 10.4.11 is too much to handle. The Tweak page suggests a MTU of 1450, some OSX error sets the recommended MTU to 1488 but that is not right evidently. So I tried that (MTU @ 1450), and Tweak then recommends resetting it to 1500! but the iBook transfer rate has increased by 3x to: 73870 bytes/sec 590 kbits/sec I will try it again at 1488 and see what happens. OK tried 1488 and the speed has jumped to 104566bytes/sec 836 kbits/sec, but it still recommends 1500 (suppose I could try that again). MTU at 1500 slows the speed way down again so I am going back to 1488!
Thanks for all the time you spent on this. Best regards

[ghporter]

In the States, just about all DSL needs a MTU of 1492 or lower because the PPPoE protocol uses up 8 bytes of the packet. The default of 1500 works fine on US cable systems. I have no clue about how broadband works in Japan, but if you got a better speed at 1488, then STAY THERE! Your system likes it and that's what matters. As a goof, I'd see what MTUs of 1490 and 1492 do to your speed.

I don't know why RMAC won't work for you-it works fine for me on 10.4.11... But the RWIN suggestion is probably related to what the incorrect MTU is doing to your connection.

[David Lee]

Now that the speed has increased, still not as fast as the Quicksilver but tolerable, I will stay with the slightly reduced MYU values, after further testing at 1490 and 1492. I think this case is a result of the ISP characteristics, my ISP, DION or KDDI, is not PPPoE, and as I am very close to a sub-station I get almost the full ASDL speed rating according to the providers test rates. OF course that was on the Quicksilver. One other option is to try Apple - Support - Downloads - Broadband Tuner 1.0
which I will do during the holiday season. Thanks again for all the help, much obliged.
Cheers and a happy new year

[ghporter]

Good luck and happy holidays!

[David Lee]

Finally got around to trying MTU settings at 1492 and 1490, the 1492 setting slows the machine down as did the 1500 automatic setting, and the 1490 (just like the 1488 setting did) jumps the speed up from 73,870 bytes/sec 590 kbits/sec to 104,566bytes/sec 836 kbits/sec. This speed difference is very obvious in actual perception, it actually feels more like a 2-3 fold increase than the approx. 30% increase the numbers suggest.

[ghporter]

I think you found your sweet spot then. Leave it there and enjoy!

[David Lee]

To ghporter: The neverending Saga continues; I have found that the MTU settings of 1488 is the optimum for my networking environment, but the actual speed of the iBook G4 is horrible compared to a G4 tower on the same router, so I did some more troubleshooting. Trying RWIN again (clicking the "optimized" button and the save settings option) I got an error message saying "can't write file" so I checked the permissions of the Startup Items folder in the root library, it was system and wheel. I changed that temporarily to admin (leaving wheel) and RWIN installed the setting file. After a restart I got a message saying permissions "needed a fix" and I let it do that. Another reboot and the machine booted normally - with the Startup Items folder in the root library and the RWIN folder inside set to system and wheel. The result was a considerable increase in network speed, still not quite as fast as the tower, but close. It seems these iBook G4s have some oddities. I often shook my head in confusion when two friends with identical machines said their network was much faster using wireless (old "b" type, later changed to a "g" airport). That just didn't sound right but now I wonder if they weren't right. Without the tweeking you recommended the speed was very bad. Anyway, thanks a lot.