(08-31-2016, 09:15 PM)MarkHaysHarris777 Wrote:(08-31-2016, 08:49 PM)jandvs Wrote: a friend of mine also has a pine64. said he never had a problem with Gigabt. I asked him how long the cable was, and he said about 6 inches. I don't have a cable that short on hand. if I get a chance I'll make one and see what that does. I also asked him to try a longer cable to see what t does to his.
We know the shorter the better.
That's wrong. Currently there exists exactly NO knowledge about the GbE problem since no debugging took place and devs were not able to reproduce the behaviour. The usual workaround known since months (switching back to Fast Ethernet) does work and if you're satisfied with 100 Mbits/sec slowing performance down you can use a cable of any length. There's also no need to use Cat6, Cat5e is sufficient and you can even use old boring Cat5 for Fast Ethernet and in most situations also with GbE.
I did all my GbE tests back in March and this week with the first Cat 5 cable I got in my hands (bought in 1996 since we upgraded a Sun server and a few PowerMacs from 10BaseT to 100 BaseT). Just to ensure 'worst cable possible'. No difference whether short Cat6 or long Cat5. It either works or not (and for me it always worked with both Pine64+ I tested with, next week getting another one to test)
GbE requires PHYs on both ends of the cable to negotiate clocks and duplex settings. It's obvious that there's something going wrong here (might be fixable in software but for that we need to be able to reproduce it first) and it's not the cable's fault unless the cable is really broken.
So choose the cable you want and if you run in a problem then add the appropriate ethtool call to /etc/rc.local to force Fast Ethernet. Maybe the problem is fixable in software (we had something similar with other Allwinner boards recently, there forcing the GbE PHY always into master mode resolved the problem), maybe not. No one knows currently. But it's known that cable length does NOT matter.