Experimental Physics and
Industrial Control System

Michael Davidsaver via Tech-talk <[email protected]> · Sun, 5 Jan 2020 21:13:34 -0800

On 1/5/20 6:33 PM, Mark Rivers wrote:
>> Are any of the 10G links fiber?
> 
> 
> I have tested with 2 different Linux machines.  On one of them the link from the Linux machine to the first switch is fiber, on the other it is copper.  Both fail the same way.  I have also tested 2 different devices that are connected to different 1 Gbit switches with (obviously) different cables.  So I don't think it is a defective cable or switch.  I would also point out that the problem is not at all random, it is completely deterministic.  Each time I send a specific packet I get a frame error on the NIC, which does not sound like collisions to me?
> 
>> I suspect that if you were able to eliminate the switches, and connect a 10G NIC directly to the device, things would work.
> 
>> Or fail explicitly with no link if auto-negotiation isn't working.
> 
> 
> My understanding what you propose is impossible.  I have not seen any 10 Gbit copper NICs that support 10 Mbit.  They all say they only support 10000/1000/100 Mbit.  The same is true for the 10 Gbit switches I have seen.  In order to physically connect the device to a 10 Gbit NIC I believe I need at a minimum a 1 Gbit switch.
> 
> 
> Have you seen 10 Gbit NICs that can directly connect to a 10 Mbit device?

Forgive me Mark, the last time I had a 10M ethernet NIC in my hand I was dealing
with Windows 3.1.1 and IPX protocol.  And I was removing it after lighting damage
(it would transmit, but not receive).

I suspect that the host side difference you describe is a symptom of a problem
between the device and the switch it is directly connected to.  I shouldn't be
theorizing further on so little information, so I'll leave it at that.

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System