Re: Tulip ethernet performance issues
On Sun, May 28, 2000 at 12:53:32PM +1000, Damon Muller wrote
> Hi gang,
>
> I know that this has come up on the list recently, but I haven't really
> seen anything that has helped me solve this little problem.
>
> I have a couple of tulip-based ethernet cards (I think they are made by
> Acton, and may certainly be re-badged), one in my Debian box and one in
> my win98 box. They are connected together by an 8-port 10/100 switch
> (specifically a LanTech MINI Switch 800 (8 port 10/100 Base-TX Switch)
> according to the front panel).
>
> The performance that I'm getting through this network is significantly
> less than I'd be expecting. In fact, it seems to be slower than the 10M
> hub that I had previously.
>
> I've just transfered a fairly large file from my Linux machine to my
> Win98 machine using the Win98 FTP client (the console based one), and
> here is what it said when it finished:
>
> 370238462 bytes recieved in 1872.52 secs 197.72 Kbytes/sec.
>
> With only 2 pcs, both with 10/100 tulip cards, over a 100M switch, I
> would have expected a much better transfer rate. (there is nothing else
> connected to the switch).
>
> Here are some diagnostics:
>
> Linux rei 2.2.15 #1 Fri May 5 18:30:12 EST 2000 i586 unknown
>
> tulip-diag.c:v1.19 10/2/99 Donald Becker (becker@cesdis.gsfc.nasa.gov)
> Index #1: Found a Digital DS21143 Tulip adapter at 0x6c00.
> Port selection is 100mbps-SYM/PCS 100baseTx scrambler, full-duplex.
> Transmit started, Receive started, full-duplex.
> The Rx process state is 'Waiting for packets'.
> The Tx process state is 'Idle'.
> The transmit unit is set to store-and-forward.
> The NWay status register is 41e1d2cd.
> Internal autonegotiation state is 'Negotiation complete'.
> Use '-a' or '-aa' to show device registers,
> '-e' to show EEPROM contents, -ee for parsed contents,
> or '-m' or '-mm' to show MII management registers.
>
>
> CPU0
> 0: 460597 XT-PIC timer
> 1: 10326 XT-PIC keyboard
> 2: 0 XT-PIC cascade
> 4: 5129 XT-PIC serial
> 5: 6112 XT-PIC soundblaster
> 8: 1 XT-PIC rtc
> 10: 286851 XT-PIC eth0
> 12: 8685 XT-PIC aic7xxx
> 13: 1 XT-PIC fpu
> 14: 151229 XT-PIC ide0
> 15: 738283 XT-PIC ide1
> NMI: 0
>
> eth0 Link encap:Ethernet HWaddr 00:40:C7:9A:01:5F
> inet addr:192.168.13.1 Bcast:192.168.13.255 Mask:255.255.255.0
> UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
> RX packets:181175 errors:80 dropped:0 overruns:0 frame:81
> TX packets:259454 errors:12590 dropped:0 overruns:4 carrier:12586
> collisions:820 txqueuelen:100
> Interrupt:10 Base address:0x6c00
>
>
> Particularly worrying, I guess, is the error, collisions, etc on the
> ethernet card (I had rebooted just before this large transfer). On my
> machine at work, with has a 3Com 509 which is also attached to a switch,
> with 3.5million packets transfered, there has not been a single error or
> collision (There should never be a collision with a switch, should
> there? I though that was the idea!).
>
> Can anyone suggest any possible solutions? Is it likely the card is
> dodgy (it worked fine, and fast with a 100M hub that I had, but I
> couldn't attach my laptop to that, as it only had a 10M pcmcia card), or
> might it be an interaction between the card and the hub? Is it work
> shelling out for a new card?
>
You should take a long, hard look at your cables. At 100M cables
that work fine at 10M can turn your data to mush. Verify that
they are "genuine" CAT5, and if you have the opportunity verify that
they give satisfactory performance at 100M with different gear; even
if the cable is CAT5, poor terminations or connectors can kill them.
John P.
--
huiac@camtech.net.au
john@huiac.apana.org.au
http://www.mdt.net.au/~john Debian Linux admin & support:technical services
Reply to: