Ethereum update: [Guide] Overclocking and Problem solving for AMD RX 570/580
Bored, so I’d figure I’d write a guide on how i got my 13 RX580’s stable with overclocking on Windows 10 and just some general information that newbies to mining may have overlooked. This guide focus on Windows OS solely because Windows 10 offers better control, better power savings, and overclocking tools for AMD cards than Linux. Linux is simple and easy to use I agree; however if you want the maximum hash from your GPUs and lowest power consumption you MUST use windows. This guide assumes you know the basics of computers, getting miner programs such as claymore/Phoenix miner and have already BIOs flashed the cards.
First off windows 10 (v1709 to be specific) is your best friend for AMD cards mining Etherum. This is strictly for 9+ GPU setups as windows Fall creators update allow 13 GPUs in a system. if you are only using 8 or less GPUs I highly suggest you look at windows 10 LTSB. I suggest using a 120GB SSDs and having a single 120GB SSD as cloner which you can always revert back too. use something like this;
just an FYI you can legally use windows 10 without activating(https://www.windowscentral.com/you-do-not-need-activate-windows-10) With this, you have fallback. Should something get corrupted you can quick clone and install AMD drivers and be up and running again in no time.
I recommend AwesomeMiner for monitoring rigs, you get up three free rigs! no charge! http://www.awesomeminer.com/
#[Section 1: Getting Started]
Setup your rig with the EXACT motherboard/CPU/RAM etc you are going to use. I highly suggest you always use exact same specification for multiple rigs. This cuts down on potential problems that might stem(mostly from the motherboard). however it will work regardless of variations of setup.
Install Win10 and create your starting point. What I normally do is, turn off UAC, Put 60GB of Virtual memory, Disable Windows updates, disable Auto, lock memory in pages, install drivers, put power settings to Maximum. Most of this should be easy enough for users, but to disable “Auto install drivers”
1.Right click the Start button and select Control Panel.
2.Make your way to System and Security.
4.Click Advanced system settings from the left sidebar.
5.Select the Hardware tab.
6.Press the Device Installation Settings button.
7.Choose No, and then press the Save Changes button.
For windows update you cannot actually stop updates but you can defer updates for 30 days and major updates (Like 1803) for 1 year. Simply go to windows Update settings, then advanced, and you should see the options there. there is also a program called winstop which does pretty much all that for you. check it out here (http://www.novirusthanks.org/products/win-update-stop/)
Locking pages in memory is not required for Etherum mining nor 60GB of VM, but it is required for XMR mining. So rather than coming back later if you like to mine XMR we’ll just do this now.
To lock pages in memory:
On the Start menu, click Run. In the Open box, type gpedit.msc.
1.On the Local Group Policy Editor console, expand Computer Configuration, and then expand Windows Settings.
2.Expand Security Settings, and then expand Local Policies.
3.Select the User Rights Assignment folder.
4.The policies will be displayed in the details pane.
5.In the pane, double-click Lock pages in memory.
6.In the Local Security Setting – Lock pages in memory dialog box, click Add User or Group(Local Computer name)
Now, add overdrive-n-tool, Lastest AMD crimson driver, and HWinfo. Clone to another SSD, setup your rig(If you havent already), start windows. Now with windows I suggest you use the onboard video.
Note: HWinfo with 12+ GPUs can take up to 5 minutes to load all the sensors it can also cause BSODs or crashes! Try disabling SWI, you can find this under C:ProgamfilesHWinfo and the edit config.ini file. Change it from 1 to 0. I still experience slow loading and sometimes crashes
First and foremost go into Device manager and check that all GPUs are reporting. Simply count the number of “VGA DISPLAY ADAPTERS” should be the exact number of you GPUs you have + 1 for the integrated graphics. If you do not see all devices listed start checking power cables and PCI-E risers. I suggest “Unplugging” the USB from the back of the GPUs PCI-E Riser until you find the one that is not reporting. Replace the riser and see if it reports properly now. once all GPUs are confirmed install the AMD driver.
For miners, I suggest Phoenix miner as I have gained extra hash and find it far more stable than Claymore’s also Dual-mining isn’t as profitable right now but that entirely up to the user. In Claymore 11.7 or Pheonix 2.9d launch the miner then when all the cards load hit “Y” to enable Compute mode for all cards, stop the miner and restart the computer.
#[Section 2: Overlocking]
Disclaimer:Overclocking depends heavily on the manufacture, card type, memory type, etc. For overclocking AMD RX580/570’s open up overdrive-n-tool. Now in Overdrive click on the “P1 – P6” states for core and “P1 – P2” states for memory. You should see them grey out. This simply mean the card will only use P7/P3 state for both core and memory! which is how you get the maximum power savings. You should only adjust P7 and P3 states. Hit apply then save as “GPU1” do this for every GPU in your system naming them as you go “GPU2” “GPU3” etc. I can only reference my overclocks with 580 8Gbs and 4Gbs. There are two ways of doing this the 100% fool proof method or the “luck” quick method.
When applying overclocks using overdrive and tool, the first time you hit “apply” the clocks are applied. If you start the miner, then forget to apply a clock on a GPU, you have to restart the computer. Additionally any adjustments you make in Overdrive and tool only get applied after restarting the computer. This is simply a AMD driver limitation.
“Fool” Proof Method
Depending on the card, baseline for memory OC will be 1750 for 4GB and 2000 for 8GB. Start there. Let the miner run for 1 day. If it successfully completes 24hrs with no crashes or restarts you’re good to start OCing! If you experience crashing start checking PCI-E Risers. I personally recommend having a test PC setup so you can individually test card /w riser. I understand this is a painstaking process but if you want near 100% uptime it pays to do this now rather than constantly fighting with crashes. find out what part of the rig is failing, replace/repair and continue testing.
Now start adjusting Memory clocks by 25Mhz, using HWinfo checking for Memory errors. Keep moving the memory up in steps of 25mhz let it run for 4 hours and check. No memory errors then keep moving up. once you find the maximum start tweaking by 5mhz and 10mhz until you only get a few memory errors. I found for RX 580 8GB Samsung range anywhere from 2100-2200Mhz Memory and RX580 4GBs Elpida from 1950-1990Mhz.
[Note on BIOs mods]: If you used the POLARIS one click timing mod most cards will not clock far beyond 2000mhz! this is because the memory straps are too tight(1500 straps) won’t allow higher overclock. However I found that the one-click mod is far better for dual mining. Again it’s up to the users but I prefer 1625/1750 Straps allowing for higher memory clocks.
You’ll know when you reached the limit for the memory, as soon as you start mining HWinfo will start reporting memory errors on GPUs racking them up consistently. The best you can achieve is 0 memory errors at worst a few thousand over a couple of hours. I personally set my limits to about 100,000 Memory error over a 4 day period. This is by no means an exact science! I’ve had GPUs run for days with 1 million/100,000+ Memory errors! Once you find that sweet spot for your memory start tweaking Voltages, 950mv is standard, depending on the cards anywhere from 800mv to 900mv. Again using HWinfo as your guide. You may find that Higher OC require higher Mv. This entirely depends on you. You can leave it at 950mv and stay very safe or tweak for lower wattage.
######[Special: Micron Memory]
Some of the new 580’s 8GBs and 570 8GBs are rolling out with micron Memory. This is the best memory for AMD forget Samsung. Micron memory can do 2200 to 2250Mhz memory no sweat. You can easily achieve 33Mhz with this type of memory, but it requires a high core clock around 1200-1300mhz. It also uses more wattage/Mv.
For Core clocks it’s much easier. I recommend 1200/1150Mhz for 8GB/4GB and 850Mv~900Mv. You can go for lower core clocks like 1100/800mv if you’re goal to save as much power you’d only lose 0.xx hash.
Basically take known overclocks from others, 1150/2150 for 8GB and 1150/1900 for 4GBs and just start from there making adjustments as needed. Obviously this method will cut down on time but if start encountering issues which may be a riser, you’ll never know. If you are 100% certain that all your risers are good then you can start here. You’ll probably encounter far more crashing and putting out fires, but you may get lucky and all your cards can handle those clocks
#[Section 3:Monitor and tweaks]
OverDrive-n-tool is what makes Windows 10 so powerful Individually configuring each card for there maximum overclocks! No guess work like Linux! I have Some RX580 8GB Samsung’s hitting 2200 while others can barely do 2100. It varies heavily.
Now if your using AwesomeMiner, you’ll get a bonus that if a GPU crashes, the miner will auto-restart and that GPU will go back to default. meaning you don’t lose out on precious mining time because a single GPU had a little bit too high OC. To find the GPU that crashed simply look at hash-rates or go to overdrive and find the one where it’s values have been reset. Every system has it quirks, I have XFX 580 4GBs that will BSOD as soon as you launch the miner with applied OC. Restart and apply those same clocks, then it will hash away for day. Sometimes I get BSOD after an hour of mining, then restart it and again back to mining for days. I had a single XFX 580 8GB run for 6+ days then randomly crash out even though had zero memory errors on the GPU. These are just examples.Once you find your sweet spot clocks, let them run don’t touch or tweak see how many days you can go without a crash or restart. If you’re experience crashes keep on tweaking depending on Voltage or memory clocks for GPUs that may be crashing.
If you want to switch to XMR mining, literally all you need is Claymore’s XMR miner or XMR-stak no need to do anymore tweaks your ETH clocks should directly apply to your XMR, cards should be doing about 900H/s for 8GBs and 800~ for 4GB’s. If you get any errors make sure you have 60+ VM and lock memory in pages was applied.
I have a “Hobbiest” Farm of 170+ GPUs Mix of Nvidia and AMD. There is plenty i’ve learned about heat, power, etc that is out of the scope of this guide. I’ve been running 76 RX 580’s 8GBs MSI and XFX, along with 4GB’s. I use them in 13 GPU setup(H110 Pro BTC+) and is stable for days pending the bad storms and losing power. It would be my honest recommendation that most people stick to 8 GPUs after my experience with 13 GPU rigs they are too much to handle and the software isn’t quite up for snuff. However running 13 GPUs has allowed me to condense rigs and save on equipment such as PSUs, motherboards, ram, SSDs, etc. So it’s all a trade off. It all takes time and patience. If you run multiple rigs, setup one tweak only 1 at a time, nothing more stressful then having your entire farm offline while your sleeping or at work! Hope this guide will helps newbies and experts alike.
Ethereum is a decentralized platform that runs smart contracts: applications that run exactly as programmed without any possibility of downtime, censorship, fraud or third-party interference.
Don’t forget to share the post if you love it !