Strange issue with my RX 570s. They've been happily mining for some years now on Claymore. With OC I get a bit over 30MH/s so 18x in total. This has been super stable.
But Claymore is broken, so I switched to Phoenix 5.5c. It's working ok, but after a little while, one of the RX 570s will adjust the cclock from 1150 to around 900 MHz. This card has the lowest error rate, and doesn't get warmer than any other cards.
I've tried setting the cclock to 1150 for all cards, but that doesn't help.
The cards are overclocked through OverdriveNTool. If I restart Phoenix, all cards mines at the corrects speeds again for a while. It's always the same card that is the problem.
Is there any way to turn off this (watchdog?) autotune GPU Core feature?
In my experiences now with the >4GB dag what worked before may not work now. Mining software like PM has to do things to keep hash rate up and wattage down that can make previous settings/card unstable.
You don't mention if this is Windows or Linux or what AMD drivers you are using - that might help. In Windows I have no issues using the 20.8.1 driver. What voltage are you using at 1150 and what is the memory OC? The voltage and memory OC are kind of more important than cclock on card stability. I could run HWiNFO64 when mining to monitor the gpus for reported errors. This helped me find cards I needed to lower the memory OC.
Just sharing how I've setup my 570's with OverdriveNTool at 1065/800 and 2000-2100/800 depending on the card and they run very stable at about 29.2-29.5mhs drawing ~112watts at the wall. I tested higher, like 1130 cclock the voltage need to be 850mv and I could get to 30.5mhs but that found the +1mh could make the cards more unstable, generate more heat, and draw as much as 10watts more - that wasn't worth it so I'm staying with the lower hashrate.