[MinnowBoard] [MinnowBoard Max] Ethernet controller

[Darren Hart]

On 6/23/14, 15:03, "selsinork at gmail.com" <selsinork at gmail.com> wrote:

>On 23/06/14 18:50, Darren Hart wrote:
>
>> For what it's worth, in our testing so far, the chip has performed well
>> and I haven't observed any problems. Is there a specific use case you
>> would like to see tested?
>
>copy some large files (10+ GB) both to and from the board using
>nfs/ftp/scp 
>while both source/destination are connected to each other with a gigabit
>network. Closer you can get to wirespeed the better, reading at 8MB/s off
>an SD card isn't going to cut it.
>
>I previously gave up with realtek gigabit adapters due to that sort of
>traffic causing either the transmit or recieve queue in the chip to stall
>and block all traffic. It took a large transfer to show the problem and
>would recover if you halted the transfer.
>The chip would still pass traffic in the opposite direction while one
>direction was blocked, but that's not particularly useful.
>Didn't cause an issue at 100Mb, so probably not seen by many people at
>the time as realtek chips were typically used on consumer hardware.

Noted, and thanks.

>
>I switched to intel e1000e cards which 'just work', so I've not tried
>realtek in quite some time, maybe things have changed. Either way, I'd
>be interested in your results.
>
>I'll admit that seeing your confirmation that it was a realtek chip had
>almost convinced me to cross the Max off my list..
>
>I take it there's no mac in the chipset unlike a lot of other intel
>stuff then?

Correct, the E38XX SoCs do not include a MAC.

> 
>Even if the chipset mac rarely seems to get used, I usually think it'll
>be better than a realtek :)

Those had their issues to, but mostly around sourcing the corresponding
PHY :-) We'll be paying very close attention to feedback on this MAC/PHY.
Thank you for raising the concern.

-- 
Darren Hart					Open Source Technology Center
darren.hart at intel.com				            Intel Corporation