Increase your income with Hive. Invite your friends and earn real cryptocurrency!

3080 underperforming

Hi! I have a problem with a 3080 Nvidia FE card, it’s not using 230w and the other ones are.
I’ve tried like 4 different overclocking configs but still giving me like 70-65 Mh/s :frowning:

Could someone help me with that issue?

Main issue is you’re thermal throttling, partially due to poor clocks, partially due to low fan speed. I would start with better overclocks. Don’t use offsets on any 30 series for eth and instead use a locked core clock, 1080mhz core should be plenty for your memory clocks. Do you have autofan on? If so disable it and let it use the static 100% fan. There’s no need for power limits with locked core clocks so you can remove those if you want to as well. It will only use as much power as needed.

Now if after all that it’s still throttling, lower the memory until it stops throttling.

To test for thermal throttling run nvidia-smi -q | egrep "Thermal Slowdown"

Change the settings but still going 70-65Mh/s

I don’t know where do I test the thermal throttling (sorry I’m new with all of this)

1080mhz on the core, not memory.

You still haven’t turned off autofan either.

You can run the command either in the shell on the rig, remotely via hive shell or shellinabox, or the easiest way, the run command box in the menu

Like keaton said, your card is thermal throttling. You are using “Auto Fan” which does not work for 3000 series Nvidia cards. Nvidia’s linux drivers do not include a way to monitor Memory Junction temperature. So, if you are using “Auto Fan” it will only be attached to your Core temperature. Your card will thermal throttle before “Auto Fan” kicks in. That being said, disable it and run your cards at a fixed fan speed of 90 to 100 percent.

Here is my 3080 FE:

Hey! Thanks but I tried with the new config and still getting 70Mh/s look and had >GPU are lost, rebooting, some other errors :frowning:

I turned off the auto fan, and change the MHz on the core, got this when I tried the Thermal Throttling test,

you still havent set your core to 1080mhz on any of the 3080s. like i said, dont use core offsets at all for 30 series on eth. its less efficient than locked core clocks.

Yes, your core is still thermal throttling. That means it is getting too hot and the card is throttling itself to cool down. Try what keaton said, use a locked core clock instead of an core offset. He is talking about the core clock and not the memory clock.

The image I shared above shows my settings.
Core: 1080
Mem: 2400

When you use a plus ‘+’ or minus ‘-’ symbol in the core you are using an offset. Cards use less power and tend to run cooler when you use an absolute core clock.

Hi! Thanks, I tried the settings but still not working :frowning:

Is there something else that I could try? (auto fan is off)

Hi! Sorry, my bad, change the settings but still not working

Run the command again to see if it is still thermal throttling with the new overclocks.
If it is, then the card is still to hot, and no overclocks are going to fix that.
You will have to change out the thermal pads and thermal paste to higher quality ones.

You happen to have a video on how to change the thermal pads on a 3080FE? :confused:

I personally do not have a video, but there are many on youtube.

Here is a random one I found:

Thanks so much! I’m gonna give it a try.

Have a great day.

Hey! And one last question haha

You know why I’m having so many issues with the tempt? The auto fan is off

If you click on the error, you should see the log, and which card is failing.
I’m pretty sure it is GPU 7 and your card is crashing because it is getting too hot.

You can try lowering the mem by 100 at a time until it is stable.
My card still hits 92MH/s with the mem at 1000.

like i mentioned above, lower the memory until it stops thermal throttling, even going negative can have a better result on a throttling card

no memory overclock at all should get you 87mh

Your cards are getting too hot. Please understand that the temperature shown in HiveOS for Nvidia cards is the core temperature. When you are mining ETH, the main thing it uses is your memory. DDR6 memory gets extremely hot and starts to degrade around 100-110c. Cards throttle themselves when they get to hot to try and prevent any damage.

Your best bet is to put each card on a windows machine and dial in the overclocks while you can see the memory temperature. Once you know what clocks each card can use without hitting 100c memory junction temperature, put them back on hiveOS.

I can only estimate but for my cards, when they go above 55c on the core, the memory is basically hitting 100c and on the edge of thermal throttling. You need to adjust your overclocks, mainly the memory clock, 3080s are good with 1080 core, and get the temperature down.