I have a feeling it is related to your internet connection given these results. Is the rig hardwired to an LTE modem or is it on WiFi too? Have any way to try it on a good old wired or fiber internet line? I can’t rule out the OS or the drive it is installed on either though.
I did have one rig on WiFi for a short time and it behaved similarly with lots of rejected shares although not this many. I ran an ethernet cable to it and that problem went away.
basically mobile internet not make reject shares, as internet usage it very low i use before 4g for a month at it use just 10gb of data, it was rejected shares i have right now 1400mbs fibre optical still same missing shares on my new rig
right now i try figure out with one card make reject as on second ring was same problem ans everytime rejected 1 card as you see in my photo i have 585 and 10 reject so im watch almost 1h but still dont see rejecf sharesyou need to check ping on your lte possible ping is to high if you use mobile networking you shut use just for rig nothing else connected to network other wise increasing lowering unpload speed, as shares download fast but to slow return to pool and pool reject
with model motherboard you have?
I had this same problem last week with one card. I lowered the memory by 10 MHz and it went away.
The rig is cabled through a switch directly into LTE central router by ETH cable
Sadly I don’t have any “standard” connections to test it with, as my location doesn’t have any possibility to do so (digital divide area)
I don’t know tho, as soon as I set the vdd, them goes pretty much ok
Some incorrect, but atleast not 50k+ like I found before in a couple hours 
Taking out the vdd them start again to go crazy and incorrect all the time
change to other os to see or still same
or try run windows straight to phoenix to see or it different
I don’t think my rig can get any better than this
in 14h30m 676 incorrect shares is still better than finding them at 50k+ or even 1k+.
I would’ve like better to have an average of… zero, but I do think it’s related to LTE connection.
switch os and will see get missing shares or not any way you losing profit is you stop ring for 1h nothing bad as any way 1h your profit almost 0
try to ser vdd at 860. i do rigt now on my rig cos zi get same reject share as clock voltage is to low and rejected shares im try run stock 1000mv core and1000mv memory no kore rejected before i was set 856 now up to 861 no more reject so second rig workig properly
i do more search about reject shares and talk to other forum.reject shares make wrong underclock or undervoltage settings in my case to low core voltage stales shares from internet connection if use wifi or internet ping is hight so in your case shut be also same core voltage to low
I’m sorry to say but that is one ugly set of numbers. And Asiukas6 is right, those errors are a result of some glitch related to overclocking or undervolting, caused either through aggressive settings or possibly an underperforming PSU.
A basic RX5700XT runs at a stock core frequency of about 1600 mHz and a memory frequency of 1750mHz. Your picture shows you have a core of 1350 set and you’ve put the voltage to 850mV which should be fine (although you should be able to go lower on the core frequency and not impact hash rate - thereby lowering power usage).
What I can’t work out is why your mem column shows 930 because that is not your memory frequency - it must be the memory voltage. That said, 930mV is high enough to run the memory at pretty big overclocks and shouldn’t be what’s causing the problem - particularly as you don’t seem to have overclocked the memory (which is what will get you an increase in hash rate).
Sorry I can’t help you tweak things in Linux but I’m strictly Windows - if you decide to try operating on Windows shoutout here and I’ll walk you through a few tweaks.
I run 6 x RX5700 non-XT which are exactly the same cards as the XTs with some higher functions disabled. On my cards, I can get the core comfortably down to 1200mHz and voltage at 800mV with the memory wound up to 1825mHz and the voltage at the pre-set 850mV. These settings give me 52.5mH/s and power draw of only 75w with no errors. Your cards should be able to get at least the same.
Is there any software like hiveos to track history, infos etc on windows?
if there is i might consider give it a try, i can partition the SDD and install both windows and linux, so not an hassle
If you mean to track your NiceHash mining info then yes the website logs you in to a portal that has a range of info, although it doesn’t provide as nice an interface as the one you have.
Not sure if it helps or not but I’m running 1 xfx rx5700xt thick iii…I’m mining Ravencoin right now @24.8 Mh/s at 1360/760/1800 at 118w…that’s the best on kawpow I’ve come up with that’s stable
I think your settings display different numbers than HiveOS because I run my memory between 900 and 930 also and I run my core at 1300
So, bumping up the vdd it stayed stable around 80% accepted/incorrect ratio without reboot for more than 24 hours, so I really think it’s the vdd causing all those incorrect.
Thinking about it tho, it doesn’t make any sense to bump it up so they consume 180W each just for 49-50Mh/s, that sunks all the profit into electricity.
@Badfish as for the Windows solution you had in mind, are them controllable from an interface remotely or something like that? Cause soon I’ll be having more than 10 rigs, and having to get into 1 by 1 to reboot, tweak etc, would cost me hours of time which I don’t have 
Dude you should get your internet sorted before you build more rigs.
