Here’s the problem I see. We don’t have a specific post where people can share their STABLE OC settings for different AMD cards. I would like to see a few options and have something as a starting point but sometimes it takes days before you would be able to find STABLE config for your specific GPU.
Here’s the solution I see. Share your GPU model and your stable overclock settings here. I will attach each response to the very top so people can quickly find it instead of going through a ton of posts.
I have
Asus, XFX, MSI RX 6600 Hynix memory. A good starting point to get 29Mh/s at 40w (software):
Core 901
650/650/1100
Mem 940
MSI RX 6600 XT Samsung memory. A good starting point to get 32Mh/s at 47w (software):
Core 1100
650/680/1300
Mem 1120 or 1130
XFX 5700 XT Samsung memory. A good starting point to get 57Mh/s at 100w (on moded vBIOS):
Core 1350
750/750/1350
Mem 900 - 950. My card did not go above 930
Radeon RX 6700 XT 12272 MB · Sapphire
Samsung GDDR6
Core 1150
675/675/1250
Mem 1074
46-47 Mhz (eth) @ ~100W (86 on hive, but not the actual value)
Let’s share our OC and I will try to keep it organized.
100% true but we need a starting point. Would you agree that most of the time for ETH you just have to play around core and mem a bit and it will run great?
That’s why I’m suggesting to help and build one post where people can find something like a “starting point” and go from there.
Core and Memory frequency settings for many cards require power adjustments…and then you have differences in local environments which impact thermal variables.
Ok. So you’re saying no way we can share “our best” and “our worst” settings so people can play around and find something in a matter of hours not days?
You can, but it’s likely faster to find the highest memory, lowest core and lowest voltages yourself from scratch instead of trying others ocs. Start high and go down on core and voltages, Start low and go up on memory. Open the miner in shell while tuning so you can see if your adjustments make it better or worse. Keep moving in the direction of making it better until any change you make makes it worse.
No, just saying a reasonable baseline for the various GPU chipsets is well published already, for example, one of many out there: https://www.hashrate.no
When I got my amd card, I was looking for a starting point. I came to hive hoping to find a reference, and found several in the 6700 thread. Anyone coming here, should know, as keaton said, all cards are different. He also gave us how to tune the amd cards. Still knowing a stable oc is a good start.
Not always that nice. The values fluctuate. Any chance for a tweaking mode that gives you the average value over the last x seconds? Might make tweaking easier
They shouldn’t fluctuate much. If the changes you’re making make no difference you’re either not making big enough changes, or the values are out of the acceptable range for that card.
Here I use the 6600 as an example. 1900mhz is noted as the max memory oc, in hive this number is halved on navi cards, so max is 950 for the memory. If you go over that, it’s ignored. You can also find this limitation by running the miner in shell and making changes and verifying they take.
Wipe out all the overclock setting (click the overclock tab on top) and click clean all the settings.
Let the rig run without any OC for 10-20 minutes to make sure it’s not a hardware issue.
Pick one card and start slowly. First I would raise your volts and go for 800/800/1300. Check one card that you apply to these settings will be stable (like 20-30 minutes at least). If yes - apply to all the cards. If not try to find more conservative core and mem settings.
After that go one by one. First, find something that works, and only after that add it to all other cards. I’ve seen a ton of issues resolved by wiping out OC settings and just taking everything slowly.
P.s. also keep in mind that your AutoFan settings can also create issues like “dead GPU”. I did see a difference between the values you enter to AutoFan and what actually happening with the rig. For example, you put your max fan speed at 5% and on your screen, you see it as 5%. When you look at the rig your fans are not spinning. That brings me to the idea that we do have issues with AutoFan settings and that’s why I keep it this way to make sure we’re not trying to push fans over 100% RPMs and causing issues.
keaton_hiveon could you please try it out on your end and let us know if you have a similar situation? Let’s say AutoFan is allowed at 100% when core above 45 degrees to see what AutoFan will do and if it will crash because it’s trying to spin fans over 100% of allowed RPMs?
@keaton_hiveon I support this request. I also noticed that the Auto-fan function built into HiveOS causes the cards to give an error and go into reboot due to the fact that the coolers cannot cool the device due to the % limit
I got fewer reboots when I put 85% max value for AutoFan. If you put 5% max for the fans you’ll see them completely stopped. That tells me we might have a few % difference and allowing AutoFan to run at 100% would increase chances of crashing since you’re pushing fans over their max limit.
Would be great to test this on other cards like Nvidia. If anyone can help? That would be great!
Here are my updated settings for RX 5700 XT (flashed BIOS), RX 6600, and RX 6600 XT.
Still getting reboots once in 24h but now I do believe this is due to the Wi-Fi card since I’m not getting any “GPU dead” messages and the rig just goes offline.
I have an idea how to test (with another Wi-Fi card). If that’s the case I will share my results and we can tell everyone to stay away from small $10 Tp-link adapters. That’s the feeling I have that I already know what’s the problem
