From: Michael Quinn on
Have P4C800-Deluxe that refuses to run at 1GBit/sec- old bios, but when I
autonegotiate with my brand new Gigabit switch, it refuses any more than
100Mbit/sec. When I fix the speed to 1Gbit/sec, the interface fails
(Windows XP says cable unplugged). I thought this board supports 1GBit.
This is the first time I've tried to run the LAN interface at 1Gbit/sec -
Uses the 3Com Gigabit LOM (3C940)

All diags says its fine ??

thnks all


From: Paul on
Michael Quinn wrote:
> Have P4C800-Deluxe that refuses to run at 1GBit/sec- old bios, but when I
> autonegotiate with my brand new Gigabit switch, it refuses any more than
> 100Mbit/sec. When I fix the speed to 1Gbit/sec, the interface fails
> (Windows XP says cable unplugged). I thought this board supports 1GBit.
> This is the first time I've tried to run the LAN interface at 1Gbit/sec -
> Uses the 3Com Gigabit LOM (3C940)
>
> All diags says its fine ??
>
> thnks all
>

Verify the cable has eight wires, and that all eight contacts
are present. GbE requires all eight wires, while 10/100BT can
run with only four of eight present (1,2,3,6 ?).

One of my GbE machines, has a cable tester feature. The first
time I tried to use the LAN interface, I got an "open circuit"
indication for one pair, from the cable tester. I had to insert
the cable a half dozen times, until some manufacturing grime was
removed from the motherboard Ethernet connector.

So those are a couple possible reasons.

Paul

From: Michael Quinn on
Thanks for the info
1 - have 2 PCs on the network, ones a DL140 G3 server (HP) and its
running fine at 1Gb
2 - grabbed the cable from my DL140 and used that in the PC4800-Deluxe -
no luck, stuck at 100MBit

Used the VCT utility on both cables (the original and the one from the
DL140) and it reports
wires 3&4 are open ??

I fear that the onboard NIC is faulty ??


"Paul" <nospam(a)needed.com> wrote in message news:fjb7nf$ukm$1(a)aioe.org...
> Michael Quinn wrote:
>> Have P4C800-Deluxe that refuses to run at 1GBit/sec- old bios, but when I
>> autonegotiate with my brand new Gigabit switch, it refuses any more than
>> 100Mbit/sec. When I fix the speed to 1Gbit/sec, the interface fails
>> (Windows XP says cable unplugged). I thought this board supports 1GBit.
>> This is the first time I've tried to run the LAN interface at 1Gbit/sec -
>> Uses the 3Com Gigabit LOM (3C940)
>>
>> All diags says its fine ??
>>
>> thnks all
>>
>
> Verify the cable has eight wires, and that all eight contacts
> are present. GbE requires all eight wires, while 10/100BT can
> run with only four of eight present (1,2,3,6 ?).
>
> One of my GbE machines, has a cable tester feature. The first
> time I tried to use the LAN interface, I got an "open circuit"
> indication for one pair, from the cable tester. I had to insert
> the cable a half dozen times, until some manufacturing grime was
> removed from the motherboard Ethernet connector.
>
> So those are a couple possible reasons.
>
> Paul
>


From: Paul on
Michael Quinn wrote:
> Thanks for the info
> 1 - have 2 PCs on the network, ones a DL140 G3 server (HP) and its
> running fine at 1Gb
> 2 - grabbed the cable from my DL140 and used that in the PC4800-Deluxe -
> no luck, stuck at 100MBit
>
> Used the VCT utility on both cables (the original and the one from the
> DL140) and it reports
> wires 3&4 are open ??
>
> I fear that the onboard NIC is faulty ??
>

Then, have a good look at the RJ-45 connector on the motherboard.
Get a strong source of light (I use a 50W halogen lamp for
inspection work). Maybe some dirt or damage will be obvious when
you inspect it. (Remember that it took me at least five or six
tries of plugging and unplugging the cable, before my Asus board
would work and pass the VCT test.)

The pair orientation on RJ-45 is: 1-2, 3-6, 4-5, 7-8
There has to be continuity on those pairs, for a pair
to be usable. All four pairs have to work, to run at 1Gbit/sec.
Only 1-2 and 3-6 have to work for 10/100 operation.

How VCT works, is described here. Just a FYI, and not crucial to
solving the problem. Only if you're curious.

http://web.archive.org/web/20051101231749/http://marvell.com/products/transceivers/singleport/VCT_White_Paper.pdf

The RJ-45 Ethernet connector is transformer isolated, from the rest
of the motherboard. When probed with an ohmmeter, you should find
a dead short between 1-2. And between the 3-6 pair. Etc. It is because
the ohmmeter tests at DC, that it looks like a short. (If you test from
pin 1 to pin 3, that should be an open.)

The transformer behaves differently, when pulses are sent, such as
normal Ethernet packets, or VCT test pulses. Then the differential
pairs look like 100 ohm transmission lines or thereabouts, and no longer
look like a dead short. That 100 ohm value is what VCT is looking for,
when the cable is connected to another GbE device.

Probing the Ethernet connector on your motherboard, runs the risk
of damaging the contacts. (I've damaged RJ-45s before, by doing that,
so I don't recommend it.) To do it properly, I'd want a stereoscopic
microscope, and electronics labs use low power microscopes for fine
work like that. A set of very fine probes (sewing needle tips), might cause
a little less damage. If you purchase a $20 multimeter at the hardware
store, the probe tips are bigger than the RJ-45 spacing.

HTH,
Paul
From: Michael Quinn on
Paul
fantastic information, I'll persist down the path you suggested.
Sometimes you get gems in newgroups, thanks heaps and kudos to
you

I've had the motherboard for 3 years now, and cannot fault it, its been
running at 100Mbit for years with no prob. I'd like to retire it and
use it as a
NAS, (use promise RAID etc) but would like to get 1GBit working.

cheers
"Paul" <nospam(a)needed.com> wrote in message news:fjdqkq$ua9$1(a)aioe.org...
> Michael Quinn wrote:
>> Thanks for the info
>> 1 - have 2 PCs on the network, ones a DL140 G3 server (HP) and its
>> running fine at 1Gb
>> 2 - grabbed the cable from my DL140 and used that in the
>> PC4800-Deluxe - no luck, stuck at 100MBit
>>
>> Used the VCT utility on both cables (the original and the one from
>> the DL140) and it reports
>> wires 3&4 are open ??
>>
>> I fear that the onboard NIC is faulty ??
>>
>
> Then, have a good look at the RJ-45 connector on the motherboard.
> Get a strong source of light (I use a 50W halogen lamp for
> inspection work). Maybe some dirt or damage will be obvious when
> you inspect it. (Remember that it took me at least five or six
> tries of plugging and unplugging the cable, before my Asus board
> would work and pass the VCT test.)
>
> The pair orientation on RJ-45 is: 1-2, 3-6, 4-5, 7-8
> There has to be continuity on those pairs, for a pair
> to be usable. All four pairs have to work, to run at 1Gbit/sec.
> Only 1-2 and 3-6 have to work for 10/100 operation.
>
> How VCT works, is described here. Just a FYI, and not crucial to
> solving the problem. Only if you're curious.
>
> http://web.archive.org/web/20051101231749/http://marvell.com/products/transceivers/singleport/VCT_White_Paper.pdf
>
> The RJ-45 Ethernet connector is transformer isolated, from the rest
> of the motherboard. When probed with an ohmmeter, you should find
> a dead short between 1-2. And between the 3-6 pair. Etc. It is because
> the ohmmeter tests at DC, that it looks like a short. (If you test from
> pin 1 to pin 3, that should be an open.)
>
> The transformer behaves differently, when pulses are sent, such as
> normal Ethernet packets, or VCT test pulses. Then the differential
> pairs look like 100 ohm transmission lines or thereabouts, and no longer
> look like a dead short. That 100 ohm value is what VCT is looking for,
> when the cable is connected to another GbE device.
>
> Probing the Ethernet connector on your motherboard, runs the risk
> of damaging the contacts. (I've damaged RJ-45s before, by doing that,
> so I don't recommend it.) To do it properly, I'd want a stereoscopic
> microscope, and electronics labs use low power microscopes for fine
> work like that. A set of very fine probes (sewing needle tips), might
> cause
> a little less damage. If you purchase a $20 multimeter at the hardware
> store, the probe tips are bigger than the RJ-45 spacing.
>
> HTH,
> Paul