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

Absolute Core Clock Thread (Turing and Ampere)

There is another topic with further discussion about it:

hey dude… it this you Hashrate stable ? because i bought this card as new and its memory temperature was hitting 110 in result the hashrate was dropping to 90s.
my 3090 was MSI Aurus.
i tried is on Windows on Nicehash. i didnt try 3090 in HiveOS Rig

@octet

Sadly, doesn’t seem to work on my 3060 Ti, 3070 and 3080 with two different rigs. I tried with the recommended range, and either it doesn’t do anything (if I use the higher range of absolute core clock), or it crashes the entire system with a hard reset with the only solution whenever I use lower clock values.

It must be because my OC are fairly agressive when it comes to memory clock and already as low as it can be in power usage. I really thought I could make it work, but it’s a lot of work and hard crashes to find the perfect settings to save 2-3 watts over what already seems to be good numbers.

Can you share a screenshot of your rig’s HiveOS worker dashboard?

Anyone have an issue where the Absolute Core Clock setting is unavailable? Regardless of the value I enter it stays as an offset. I am running latest OS (0.6-203@210519).

Just created a thread about it.

I guess hiveos is not supporting it ( mean the TI variant)?
it’s not mentioned in the examples: System Changelog and News - #41 by hiveos
you could try locking it manually using nvidia-smi
Syntax should be
nvidia-smi -i 0 -lgc 1000
This should set card id 0 to 1000mhz locked. Try it out. Please set core clocks to zero (auto) in hiveos before attempting to be sure.

Thanks, I’ll give this a try. I also don’t have the option to lock the core on my 1660 Supers either. Not just with the Ti variant.

Has anyone tried to use this on 1080 Ti’s? I was curious if it would work for them as well?

Does anyone know perfect OC for 1660s evga and 3060ti eagle? When i try to set core clock with nvidia-smi -lgc command but it reset and using old core…

old hiveos version (0.6-203@210526) with 3060ti with 2700 mem it was okay and currently I’m using 0.6-203@210531 version now i have some invalid shares on 3060ti (i didn’t take screenshot) but:


here is my 0.6-203@210526 version settings.

Is it safe to overclock any core clock setting past 500? I put my rtx 2070s at 1000 core clock and 1800 mem and power consumption dropped over 100W but is this safe to do to the card?

yes, above 500 it changes the way it functions and start acting as absolute core clock, thats why your power consumption droped

Right, I understand that part but what is actually happening to the GPU when “absolute” core clock is enabled versus just core clock?

“Just core clock” is an offset value and can fluctuate. Absolute core clock is exactly what it says it is. This is also better for the card’s longevity as it lessens power draw + temperature throttling considerably (all things being equal).

Absolue core clock. killed my hash rate on my 3090 and 3080 and now they are not going back to previous.
I recently changed the settings to absolute core clock. The power has dropped substancially which I see is the cause of the hash rate drop. So I tried to revert them back to normal settings ie core clock and only at the very start of mining do they start off well, but pretty soon they revert back to power saving and I end up losing hash rate. My 3080 is sitting at 78 hash and my 3090 is sitting at 90 hash. my 3080 used to be at 90+ hash rate and my 3090 was at 118 hash.
initially I set them to 1150 absolue core clock to test them out, and when I saw they went to shite. I tried to revert them by going back to -100 core clock, but unfortunately the power has not increased.
I have tried to remove them from mining, and adding them back. No success. what else should I be considering?
I did take my absolute core clock all the way up to 1500 on my 3090 from the initial 1150, I tried to zero PL, but none of this has worked

For the 3080 use 1070 core and 2150 mem.
I get 99 MH at about 205 to 215 depending on gpu silcone lottery, and NB miner

1 Like

Adding a Gigabyte 2070 (non super) with Samsung memory:

Results in: 44.25 MHs and 111 watts on NBminer 39.6

This card was ok (43 MHs and about 110 watts) but things started to get really interesting after locking the core clock. Haven’t confirmed the wattage number yet but if this holds, then the 2070 Super with Samsung memory might well be on the top shelf of really efficient cards.

The card above is the Asus TUF Gaming RTX 3080 (non-LHR). If anyone has better settings for this, please feel free to share it.