Was curious on how to overclock AMD cards properly
I know nvidia cards in the pst I’ve overclock Ed as followed: bring cards to stock>set PL to 100%>bring up memory in increments of 50MHz till crash (I then drop 10%, then minus 25 MHz)>then start bringing core clock up in 50MHz increments till crash (then I drop 10%). Once I have my core and mem i then throttle my PL till I get an efficiency and hash rate I like.
What’s the procedure on AMD cards. I tried to do it the same way but had not much change from stock settings.
Can someone point me in the right direction to a post for a guide or word it out for me like I did for the nvidia above?
Same idea. Min/max core and mem clocks, then min voltages. The goal being the highest mem that’s stable, lowest core that maintains full hashrate and the lowest voltage for each parameter that is stable.
6700xt can typically do the max mem they’re capable of which is 1075.
So to start OC put to stock & start bringing memory up till crash & reduce. Then core clock. Then what voltage are you referring to? There’s a lot of other settings. AMD seems so much more involved then nvidia.
@keaton_hiveon
my approach on amd cards start inspectioning the bios limit with more power tool, like this (6800 xt)
By this I get the upper memory limit and the bottom core limit, so i start from the upper limit of memory and wait for crashes decreasing memory clock by 10.
When the gpu is stable i start lowering core clock until e get crashes or invalid
Last but not the least the downvolt. Starting from memory and mem controller (because it’s faster than core voltage), then the core voltage, and last the SoC.
This is my general approach, but with the amd 6000 series i move the memory clock by 1mhz-step because this gen is SUPER SENSIBLE
so first memory clock. ramp slowly till i get a crash, reduce by 10%
next memory clock till io get crash, reduce by 10%
then memory voltage. where do i start at with this? do i start at a max value then ramp down or start from stock then ramp up?
next memory controler voltage. again, do i start at a stock setting and work up or start at a max value and work down?
then core voltage ( again, start down and work up or up and work down?)
then SoC frequency or VDDmax?
AMD cards are making me feel super uneasy about overclocking because there’s so much more parameters to cover.
if you can please spell out like this is for someone who’s in preschool lol
thank you in advance to you and any other input from any other replys
I haven’t touched my 5 AMD cards yet because I’m just overwhelmed. I’m looking for someone to break it down with detail, which order to adjust parameters and how to do each one. Voltage, core, memory, etc.
okay, so just getting back to my rig. been out of the country but no issues while i was gone
so, to start OC AMD cards start with…
Mem
Core…
Then im lost from there! Do i mess with core voltage, mem controller voltage, and memory voltage? and what about the SoC frequency & Soc VDDmax?
i assume power limit/wattage is the last thing i dial in(lowest wattage for good hashrate and good efficiency.
The goal is to find the highest memory that gives you highest hashrate, then lowest core clock that maintains that hashrate. If you go too low you’ll lose hashrate, not low enough and you’re wasting power. Then lowest voltages on each value that is stable, if you go too low it won’t be stable. Not low enough and you’re wasting power. It’s trial and error. Every card is unique.
You can look up the default value for your card or in hive in the info bubble for each value it has default values for 5700 which is typically pretty similar to navi 2 defaults. You can go in as big or little steps as you want to. Doesn’t matter how you get there, but the more fine tuned you have it the better on the card, the less heat and less power costs. And yeah the goal is 100% stability at the highest hashrate with the lowest heat/power.
