Nvidia GPU’s CANNOT be bios moded, which makes them a bit easier to setup. For them it’s best to only use MSI Afterburner to overclock and limit their power draw. The downside is that you can’t get the maximum out of them, cause we can’t replace the memory timings.
Follow first 4 chapters of AMD Ethereum Mining Guide before continuing.
5. GPU Drivers
Now after Windows is setup properly, download a tool called DDU, That tool will uninstall your current driver and block Windows from automatically installing GPU drivers. That’s important so that Windows does not install an outdated driver! It will ask you to run in “safe mode” but that is not necessary. When you run the program just click on “Clean and restart”.
Now turn off your PC and connect only ONE GPU.
What driver is the best?
We want to have the latest NVIDIA Drivers installed, so go to NVIDIA driver website, and download the latest version for your GPU’s. After you downloaded and installed the driver, shut down your PC and connect all the other GPU’s. Now after you turn the PC back on wait about 10 minutes for the drivers to be automatically installed on the GPU’s. Don’t do anything in the meantime, you can check if all your GPU’s are installed properly if you see them in Device Manager under Display Adapters.
6. GPU Miner
This part is almost identical as for AMD Ethereum Mining Guide, so read it first. You can download Claymore 10.2 NoFee Version.
The only difference is that NVIDIA GPU’s work better in dual mining Sia coins rather than Decred as it’s the case for AMD cards. In dual mining you can get even slightly higher hashrate than just solo mining ethereum. I found out that the best pool for SIA is Nanopool.
For solo ethereum mining, the configuration is:
EthDcrMiner64.exe -epool yourMiningPool -ewal yourEthAddress -epsw x -dcri 6
For dual mining ethereum + sia the optimum would be:
EthDcrMiner64.exe -epool yourMiningPool -ewal yourEthAddress -epsw x -dpool dualCoinMiningPool -dwal dualCoinWalletAddress -dpsw x -dcri 25
If you would be using Nanopool for ethereum and sia (I use that combination) the configuration would look like that:
EthDcrMiner64.exe -epool eth-eu1.nanopool.org:9999 -ewal yourEthAddress.minerName/yourEmail -epsw x -dcoin sia -dpool stratum+tcp://sia-eu1.nanopool.org:7777 -dwal yourSiaAddress/minerName/yourEmail -dpsw x
That would be if you are in Europe as you see from -epool and -dpool addresses. Always chose the pool that is closes to you to reduce the ping/latency. It can be found under “Help” section of Nanopool, all different pool addresses are listed there and chose the one on your Continent.
7. Overclock and power draw
Basically the only thing we can do with NVIDIA GPU’s is to increase memory clock rate as much as possible and limit the power draw (we can’t reduce voltages also).
There are couple of tricks that most people don’t know. On most of GTX 1060 and 1070 GPU’s you can reduce power draw by 50%! That can make GTX 1070 give around 32 MH/s for only 80W from the wall making it the most efficient GPU.
For overclock and limiting power draw we will be using latest version of MSI Afterburner.
Follow the following steps to get the optimal hashrate/power draw for your NVIDIA GPU.
IMPORTANT: Have your integrated GPU as your main during the testing or you will encounter a lot of PC freeze/crash or even GPU driver buging forcing you to reinstall the drivers
Memory clock Overclock
Hashrate is calculated on the memory of the GPU, so the higher the Memory clock rate the higher hashrate we will get. Follow those steps to get the optimum Memory clock rate first.
- STEP – we will first using the MSI Afterburner get the maximum memory overclock for your GPU’s. Check in GPU-Z what memory type your cards are:
- Samsung – the best memory type can be overclocked from 500-1000mhz
- Hynix – the worst memory type, they can be overclocked to 300-600mhz and produce very low hashrate, common in 3gb cards, avoid at all cost
- Micron – common in GTX 1070 cards, it’s good memory type and can be overclocked from 500-1000mhz
- Elpida – it probably won’t handle higher memory overclock it can go from 300-800mhz
- STEP – after you know your memory type start from 300mhz memory overclock, save it and run Claymore for 3-5 minutes
- STEP – if you didn’t get any error increase the memory clock speed by 100mhz and test with Claymore for 3-5 minutes till you get a PC crash/freeze or see this error:
This Error means your memory is overclocked too much.
- STEP – After you found the limit of your Memory clock rate, reduce your overclock setting by 100mhz and that’s your optimal GPU Memory clock rate. For example if you got a crash at 800mhz memory speed or saw that error, reduce the memory to 700mhz and that’s your optimum.
- STEP – Leave the miner for 2 hours to be 100% sure you didn’t crash the rig and look for incorrect shares, if you get incorrect share,freeze,GPU settings reset then you need to reduce the Memory Clock rate by additional 50mhz. For example my Micron GTX 1070 GPU:
Core Clock Overclock
After you found the optimum Memory clock rate, we will need to try and push the GPU Core clock rate to 150 MHz. That’s important because on the last step we will try to reduce power draw of the GPU by heavy amount, and that’s only possible with 150 MHz on the GPU, basically the closer you will get to 150 MHz you will save more power, no need to go over 150 MHz.
- STEP – put your memory clock rate to the max stable overclock settings you found on the previous step
- STEP – now set your Core Clock to +50 MHz. Save settings and test for 3-5 Minutes in Claymore.
- STEP – now increase the Core Clock rate by 20 MHz, save settings and test for 3-5 Minutes
- STEP – repeat that process until you reached 150 MHz or you get a crash/freeze/errors in Claymore . If you got an error before 150 MHz your GPU core is weak and reduce the Core clock rate by 20 MHz of the value that caused the crash.
- STEP – you will probably reach 150 MHz on most GTX 1060/1070 GPU’s and it should look something like that:
Power Limit Reduction
If you followed previous steps, this part can reduce power draw of your NVIDIA GPU’s a lot, by that I mean up to 50% making them run at 70-90W from wall!
- STEP – Set Memory Clock and GPU Clock to maximum ones you got from previous steps
- STEP – Set Power Limit to 70% and test the miner for 2-3 minutes, look for the calculated hashare. On 70% power limit all Nvidia GPU’s should give almost maximum hashrate
- STEP – Try to reduce power draw to 60% and look at the hashrate and test for 3-5 minutes to see if you get crash/freeze/Claymore error, if you do increase power draw to 70% and that’s your optimum
- STEP – This is where the optimum should be, reduce the power draw to only 50% and test for 3-5 minutes. It should give almost full hashrate with massive 50% power draw reduction. If you get crash/freeze/Claymore error increase the power draw to 60% and that’s your optimum. You should get something like this:
This is the final optimum for GTX 1070 8gb Micron cards (Samsung also in my case)
One other example is my GTX 1060 3gb Hynix cards (garbage cards):
Some of my GPU’s can only be reduced to 70% power draw limit, if I go under that limit I lose heavy on ethereum hashrate, so keep an eye on that. In my case it happened because on some GPU’s I can’t get the Core Clock to work above 50 MHz without a crash or freeze. That’s probably because they are low quality cheap GPU’s. On most NVIDIA cards this setting should work properly.