Nothing related with the topic but I would share my experience with AMD support RMA since I fucked up my card see post on top
They are sending me a new cards in the next 2 days.
On my side I need to send the old one in NL (I’m french) they don’t paid for the shipping tho. ~18€ for me
Really good support from AMD. I hope the next card will be better 
I would like to share my experience in case that will help someone.
I got 2 reference RX 6800 that have both the SoC problem, on windows adjusting with MorePowerTool the SoC to 33A they get 62 MH/s (without fast timings) at around 110w which is pretty good.
On HiveOS without trying to fix the SoC problem they get max 59 MH/s which is not terrible but they can get higher hashrate.
Now both my cards are running since 3 days at almost 62 MH/s with 105w and that’s accurate because i checked with powermeter
In this screenshot you can see almost every settings except the SoC frequency so i am gonna post another screenshot just to be clear
So here are the steps i do to get this Hashrate with this power consumption
With those steps i “fixed” my cards multiple times without having any problem, don’t know the behavior but it seems that if you shut down the rig for a while when you turn it on the “Trick” is not applied anymore and you have to re do it again, but if you do a quick reboot it seems to work without redoing that
if you keep doing this you will have more and more stability problem at the end like people have reported here and maybe like me you will kill your card and unlock the RMA bios flash succes xD
Not sure you want that
It is for this reason that I think that the new method of increasing only the frequency of “F state” is a better option.
The problem I’m facing right now is that I can’t apply the settings at startup before the OC settings. I tried through crontab, init.rc, systemd but without any success. Just for the moment after rebooting the system I have to apply the settings.
For now, after 3 days, everything works normally, also a lot of restarts (when trying to apply the commands at startup).
i have done same stupide mistake what have you done did you fixed any one have any solution?
what is solution to fixed
Flash back original bios from your card.
i have try but cant in hive os i will see if the card work on windows
Hi.
I’m stuck at 57 with my Phantom RX6800.
Please could you point me to the specific BIOS that you used? Also did you flash it in windows or with HiveOS?
So far this is my experience too. I just shut down the rigs because of a lightning storm. when it came back up soc is low again. I am worried about bricking the card like the above user?
What is your soc running at? I’m running at 1066 and sometimes drops to 960. Does it need to be fixed at 1200? My 6600xt runs at 1200 soc
For what it is worth I have (1) 6800 reference which holds 1066 socMHz and gets just over 60MHs with 1275core/1060mem.
Note I also have 1200 socMHz cards in the 59’s with the same other Core/Mem specs. SocMHz is just one variable we are dealing with, see the F State/Clock above, along with voltage.
My target for the 6800 references is 61.01 or they get stuck back in Windows rigs for 63+.
If they are high maintenance to run 61, I’ll take the hit in a higher maintenance OS and move on to other problems.
It works for me, I got a stable 62,5 with this method (nbminer)
Yes you are right Nbminer also gives me 62.5 MH/s
First create script (I create my in /opt)
Go to /opt
cd /opt
Make a new folder with name like you want
mkdir script_upp
Make a new script
nano script.upp.sh
Put commands in this script (separate command for each card maybe can do it better with “for cycle” and index of GPU this will be the next job)
#!/usr/bin/env bash
upp -p /sys/class/drm/card1/device/pp_table set smc_pptable/FreqTableFclk/0=1550 --write
upp -p /sys/class/drm/card2/device/pp_table set smc_pptable/FreqTableFclk/0=1550 --write
Make the script executable
chmod +x script.upp.sh
Test it
bash script.upp.sh
Now if everything is ok you can do next job:
Now we put script before OC settings open hive script
nano /hive/bin/hive
Find a line
#Apply OC before miner start
Then put our script after this line: if [[ $GPU_COUNT_AMD -gt 0 ]]; then
echo2 "> Applying F state clk"
bash /opt/script_upp/script.upp.sh
It will look like picture
Then press Crtl +x , press Y, press Enter
If everything is ok after reboot in system journal log you can see our commands
Open journal
journalctl -u hive
Scroll down and you will see if it’s run normally
Thanks it works perfect!
Now i only have to change the memory voltage to let the F state become active
Can anyone please explain to me what is going on that I have on two gpu so much lower Hashrate? I tried the settings which were posted by gsico and they increased the hr a bit and reduced the consumption. I didn’t flash gpu just overclocked. With the settings as posted on the mining chamber, it holds me evenly at 58mh on each card with a power consumption between 113-124w. I use gminer.
I have the same problem, and the boards are from the same brand.
Can anyone say the reason for this?
It seems to me that a new version of Riser 010x has come out that solves any AMD problem, right???
Sorry for my ignorance, but what is the function of this script?
Running stock bios on XFX, I have not found great value or differences beyond 1300 Core or 1068 Mem. Yes, four of the six cards in this rig will get over 62 when I set at 1075, but they’ll give me trouble at some point 72 hours later. At these settings they’ll run until I upgrade the HiveOS version, weeks between care and feeding.
This rig does have (2) that tend to get unsettled so they are set at 1275 and 1060 respectively:
Risers are version 09s with the fancy blue and red LEDs fwiw.
Which miner are you using?
