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

RTX 3060 getting worse hashrate with overclocking when mining ERGO

try to use nvidia-smi -lgc 1305 on the terminal

Using Nanominer? Nanominer runing well :wink:

Use nvidia 465.31 !

Try this setting on cards

-490 Core
2600 Memory
112w PL

that means your overclocks are not stable

I see people here overclocking memory to 2400-2600 to get 110 Mh @ 100-110w. But there are better settings. I am doing 121 MH @ 2200 memory clocks. Here are my settings:

150w is too high my dude, get it down to 100-115w

yes there are better settingsā€¦

I could push it over 125.x but less stable.

Btw: My gigabytes perform best, although all say they are shitty.

gigabyes are the coolest I found. Also, how are you only drawing 103-113w when PL @ 140? when I set high PL it draws that much, eg. if I set PL 140w then it will draw 140w.

Absolute core clock ist the magic for power consumption.

my other bigger rig:

how do I set absolute core clocks?

Just set my settingsā€¦ you will see that hive shows ā€œabsolute core clockā€ then when entering 1550 above itā€¦

You have to play a bit around what is for your cards bestā€¦

I would start with core 1400-1600, memory 2300-2600. power always 130-140.

great, thanks! I will try it out tomorrow.

I have the V2. itā€™s a bit trickier, My settings are 77 -242 2600 111 and then I lock the core with nvidia-smi -lgc 1305 to make it stable at 119.3 Mh/s 100W and my MUtl is just 38.
What MUtl are you getting?

if it works for your v2 cards its ok.

  • you first set relative core clock
  • then you set absolute core clock with nvidia-smi -lgc 1305

I set it directly to 1550 with a 50 second delay. No need for me to set relative and then manually absolute clock.

[11:19:09] INFO - ======================== Summary 2021-07-14 11:19:09 ========================

[11:19:09] INFO - |ID|Device|Hashrate|Accept|Reject|Inv|Powr|Temp|Fan|CClk|GMClk|MUtl|Eff/Watt|
[11:19:09] INFO - | 0|Nvidia| 123.8 M| 449| 4| 0| 102| 61|100|1552| 8575| 40| 1.214 M|
[11:19:09] INFO - | 1|Nvidia| 123.5 M| 471| 0| 0| 97| 49| 85|1552| 8575| 40| 1.274 M|
[11:19:09] INFO - | 2|Nvidia| 123.7 M| 456| 1| 0| 113| 53| 85|1552| 8575| 40| 1.095 M|
[11:19:09] INFO - | 3|Nvidia| 123.6 M| 451| 0| 0| 105| 57| 85|1552| 8575| 40| 1.177 M|
[11:19:09] INFO - |------------------Ā±-----Ā±-----Ā±ā€“Ā±ā€”Ā±--------------------------------|
[11:19:09] INFO - | Total: 494.6 M| 1827| 5| 0| 417| Uptime: 0D 10:24:41 CPU: 1% |
[11:19:09] INFO - =============================================================================

Uptime so low cause i tried your settingsā€¦ lower hashrate. under 100 cause my cards are V1 3060s.

Also I need a dummy HMDI, IGPU graphics on, GPU main display on BIOS and boot with the HMDI connected to the monitor to the GPU. The verbose stacks at boot so I unplug the monitor from the GPU and then I do that:

Iā€™ve just repeated the procedure to be sure as I found this yesterday testing BIOS settings.

Summary

what happens if you reduce memory a bit so it shows 8550-8575ā€¦ and set
nvidia-smi -lgc 1450 ?

Up to MUlt 40 hit the limiter and halves

Maybe doing both of them from the terminal ?

Anyone knows the script?

You should change to LolMiner, you will get greater performance in combination with Drivers 460.39 and:

Lowest Fee than other miners only 1.5% vs 2.0% or 2.5%

You will have real hash, not like you have now. LolMiner is only of the two miners that report real hash in ERGO, because they count the no mining time of the Data Ergo Set. You are not seeing drops now and that is not true. During the Data Ergo Set you are not able to mineā€¦ how can you still have Mhsā€¦ so thatā€™s not honest from that miners

Iā€™ve been a few hours running it on lolminer for that exact reason and to see if I have different results on the pool.

ATM Iā€™m not having rejects as I was getting with NBMiner.

Regarding the driver, 460.39 does not recognise the new 3060ā€™s.

1 Like

You are right, my fault. I thought you have a v1 and not a v2. If is v1 it can still work with 460.39

At the moment you are seeing reality the average including the no mining time of the Data Ergo Set is real hash. And you are in a lucky moment 2% over the average speed :slight_smile:

1 Like

Have you tried on P0 state? I was able to squeeze a bit while watts didnā€™t go mad.

1 Like