LUXCoin Config File Tuning with AMD RX580 Cards

If you’ve not started mining luxcoin yet, go read my guide on getting started first.

Luxcoin is the coin to be mining right now. Rather luckily for those of us in the know, has not listed the phi algorithm on it’s homepage. What that means is the general public who have not heard about mining lux on their own channels don’t know about it. That means less miners, which means less difficulty, which means more profit for those of us with an ear to the ground.

There is one drawback to a coin with a lower volume of miners- there are less people figuring out how to mine it, and how to mine it well. And so it appears the task has fallen to me.

Please read the whole article before making any changes to your config file, as I found some settings actually stopped some rigs working even when the changes were rolled back.

I spent many hours figuring this out. Tips small or large are welcome, my address for LUX is LMMLQvz3ZZo6R29vZM7gCgBQUsvyGYChjh.

Note that this information is for RX580 8GB cards. I did however see good improvements using these settings on a rig with RX480 4GB cards. I therefore think this information is valid for all AMD RX series cards.

The speeds i acquired with these changes are:
RX580 8GB – 17Mhs
RX480 8GB – 16Mhs (@ Stock, probably can go faster with overclock)
RX480 4GB – 15.5Mhs  (@ Stock, probably can go faster with overclock)

This information is for SG Miner 5.6.1 and may not still be valid with new miners in the future.

Drivers & System Specs:

Drivers seems like less of an issue than with most other miners. I had been using the AMD blockchain drivers, but noticed that using AMDs latest drivers instead  (17.12.1 at time of writing) made no difference at all. Therefore I think you can use whatever the latest standard package is without drawback. Edit: With the new drivers on some rigs i’m having problems GPU 0 running slow. It may be best to stick with the blockchain drivers. 

I also noticed that in general system RAM quantity is important when mining LUX. Systems with more RAM perform better.

Config File Variables:

So, without further ado, let’s look at the commands one can add to the config file. I’ll detail all the commands I tried, and their results.

Note that all tests were done at -i20. I did test alternative I settings, but they resulted in lowered performance averages.

––shaders <value> Screwing with this made no notable difference. 

-nfactor <value> The default value of 10 seems about right. I tried various settings, which all just seemed to make the hashrates jump around a lot. Leave this alone or set it to 10. 

––worksize <value> Changing this does make some difference. It also seemed to make the values much more dynamic over time. I settled on a value of 128 as that seemed to give a stable and high result. 

––thread-concurrency <value> Now we are into territory that makes a difference. A RX580 card has 2304 shaders, so you need to use a multiple of this number. I tried most values, but eventually settled on 4608 has giving the best result over a longer period of time. 

––gpu-threads <value> This value is the big daddy, the one that will really make a difference. However, setting it too high will result in an error or a crash to desktop. As far as I can tell the value  you can use is linked to your GPU & system RAM. More RAM means more threads, but i’m not totally certain what the link is yet. I found on some of my rigs with just 2GB of RAM, that after putting this setting in, they would crash to desktop and continue to crash to desktop even after the setting was removed. This was fixed by creating a new user in windows and trying again with a roll-backed config. At least one of my rigs with 2GB of system RAM did allow 4 threads on 4GB cards, but not 8GB cards. So try various values and see what the highest value you can use is. I found 8 is the maximum and gave the best performance. The higher number you can use, the better performance will be.

The Ideal Config

You can download a BAT file with this config typed in for you here.

Given all of the above, the best config I have found is (If you copy and paste this, check the dashes stayed as short dashes):


sgminer.exe -k phi -o stratum+tcp:// -u LMMLQvz3ZZo6R29vZM7gCgBQUsvyGYChjh -p c=LUX -I 20 ––gpu-platform 2 >log.txt -w 128 ––thread-concurrency 4608 ––gpu-threads 8 ––nfactor 10


Overclocking with LUX mining will let you squeeze out those last few hashes. Lux is much more sensitive to core speeds, and I found I could push my RX580s up to 1475MHz without getting unstable. Memory speeds seem to make less difference, but it did slow down a little with them reduced so i’ve left them at 2200MHz. I’ve not tried BIOS modifications at all, as I like my warranties. However i’m sure with some modification it may be possible to squeeze out more.

That’s all. Good luck! Please leave your results and system specs in the comments below, as more information over more rig variables will help me to perfect this further. 


  1. Hey, thank you for your guide. It help a lot. Just a little question though, I have 8gb of ram with windows 10 and 7 RX580 8GB. The only gpu-threads I can use is 3. Above that it give me a lot of error. Do you have any idea? I changed my virtual memory to 56000, but even without that I have the same prob.
    So atm I can get 16 but not more than that.

    Thank you

    • I’m starting to wonder if it’s linked to how many cards are in the system. The only system I got 8 threads in has 2 cards. The systems I have with 4 and 5 cards will run 4 but no more. You’re saying a 7 card system goes to 3. I need to experiment more, it could be a red herring.

  2. What if i have a 570 and a 480 on the same rig, shaders are not the same numbers, anyway to config them seperatly?

      • use comma separated values. better yet, write out a config file from the running miner (press ‘s’, then ‘w’ if i recall) and you’ll see how it’s done for at least some of the settings

  3. Awesome guide and can’t wait to try it out.

    Just wanted to ask i was mining fine for a few months on my RX480 rig, and now i cant get it to work. keeps saying that SGMINER is not found, im 100% sure its the antivirus or windows defender and have tried turning it all off and allowing exclusions but still not luck . Windows defender keeps saying TROJAN:WIN32/Bitrep.A removed

    Any thoughts or ideas

  4. Faboo005, I am wondering if you have the same thing than me, do something your hashrate drop? Like by exemple when I monitor my rig with ethmonitoring I can see that my hashrate time to time will drop to 12.5 for a couple of minutes and then go back to 16. It happen many time a day, not sure if it could be my bios mod or something else. I use the latest blockchain driver. Do you all have the same prob?

    • Hashrates do go up and down, generally it’s best just to watch the average figures for a better idea whats happening. If it happens enough to drag your average down, then maybe the card is getting too hot causing the clock speed to go down.

      • The temp of my cards is around 60 Celsius (which is very good) and my clock speed is 1366 and memory is 2000, I have more prob if I change the speed core, maybe I should put back my bios to the default bios, could it be my bios mod? And I want to know how much rejected do you have? I am between 2 – 3% of rejected share,what about you?

        • I generally only get rejects on rigs that are being taking up and down while i work on them. Rejects on working rigs are rare. Did you fix the clock speed in the bios, or is it from overclocking software?

          • Nah I didn’t touch to the core clock ans memory clock, only changed the timing in the bios, maybe thats my prob, I have an other rig with Nvidia and I get almost no rejected shares. Do you think I should pull back my bios by default?

  5. Also, I think you may have the same GPU than me, I have 7 RX580 SAPPHIRE pulse OC with Samsung brand. I saw in a previous post that you had that model. What bios mod are you using? Stock bios?

    • Yup stock BIOS! Mining LUX only really benefits from a higher core speed, which I can do with afterburner, so there is no real need to mess with BIOS type stuff like memory timings.

      • Do you have that exam model of gpu?, did you change the virtual memory of windows? What tweaks of windows are you using and how much ram do you have?

    • I’m afraid i’ve not been able to get hold of any Vega cards to test. It’s probably mostly the same settings, just with a different shader count.

    • This is my config for 2 RX580s, 1 Vega64, and 1 Vega56 (flashed to 64). Devices 0 & 2 are the 580s, 3 is the 64, and 4 is the 56. I have a 5th card (HD7950) on this rig which would be device 1, but I don’t use it for LUX mining as it’s not very performant – therefore the second value in all comma separated quintuples can be disregarded (i.e. first and third values for 580s, fourth for vega 64, fifth for vega56.) This is used w/ v5.6.1.1 of the miner, rather than the latest, as it performed much better in my tests.

      “pools”: [
      “url”: “stratum+tcp://”,
      “user”: “LYqFGn2UE4jAkrJWxpmUujtPsTpMCf94gK”,
      “pass”: “c=LUX”
      “profiles”: [],
      “failover-only”: true,
      “algorithm”: “phi”,
      “device”: “0,2,3,4”,
      “intensity”: “20,10,20,23,22”,
      “gpu-threads”: “3,2,3,2,1”,
      “thread-concurrency”: “4608,1792,4608,32768,28672”,
      “worksize”: “128,128,128,128,64”,
      “temp-cutoff”: “75”,
      “temp-overheat”: “65”,
      “temp-target”: “40,50,40,50,50”,
      “gpu-memdiff”: “0,0,0,0,0”,
      “shares”: “0”,
      “kernel-path”: “/usr/x86_64-w64-mingw32/bin”,
      “api-mcast-port”: “4028”,
      “api-port”: “4028”,
      “debug-log”: true,
      “expiry”: “28”,
      “failover-switch-delay”: “60”,
      “gpu-dyninterval”: “7”,
      “gpu-platform”: “0”,
      “hamsi-expand-big”: “4”,
      “keccak-unroll”: “0”,
      “log”: “5”,
      “no-pool-disable”: true,
      “no-client-reconnect”: true,
      “queue”: “1”,
      “scan-time”: “7”,
      “tcp-keepalive”: “30”,
      “temp-hysteresis”: “3”,
      “vote”: “0”,
      “watchpool-refresh”: “30”

      I manage gpu settings through OverdriveNTool rather than with the miner. 580s are clocked around 1375 @ ~985mV for core, and 1250MHz mem. Core frequency and/or voltage can vary based on ASIC quality and/or vrm manufacturer. These settings seem to be pretty reliable around 77-80%. Vegas are clocked around 1425MHz @ 925mV for core, and 1000MHz mem (this can prob go even lower to 900 or less.)

      Results are
      RX580: 17MH/s @ 140W
      Vega56: 24.5MH/s @ 180W
      Vega64: 29.25MH/s @ 200W

  6. My Sgminer cannot work! Please check:
    [09:19:41] Waiting on sem in miner thread
    [09:19:41] Starting device 5 mining thread 4…
    [09:19:41] [THR31] Popping work from get queue to get work
    [09:19:41] Pushing sem post to thread 34
    [09:19:41] Waiting on sem in miner thread
    [09:19:41] Starting device 5 mining thread 5…
    [09:19:41] [THR32] Popping work from get queue to get work
    [09:19:41] Pushing sem post to thread 35
    [09:19:41] Waiting on sem in miner thread
    [09:19:41] [THR33] Popping work from get queue to get work
    [09:19:41] Waiting on sem in miner thread
    [09:19:41] [THR34] Popping work from get queue to get work
    [09:19:41] Waiting on sem in miner thread
    [09:19:41] [THR35] Popping work from get queue to get work
    [09:19:41] Selecting for work
    [09:19:41] Selecting for work
    [09:19:41] API not running – API will not be available
    [09:19:42] [THR0] gen_stratum_work() – algorithm = phi
    [09:19:42] [THR0] Generated stratum merkle 43b34c77824e0e9e4279c9a3f3de2646a742e48328daec88a8137c06dec4444d
    [09:19:42] [THR0] Generated stratum header 00000007164c22d11c49b4c40ae8d9c99b71abe437b84fb4391151373369d589492e650543b34c77824e0e9e4279c9a3f3de2646a742e48328daec88a8137c06dec4444d5a669bb11b02909700000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
    [09:19:42] [THR0] Work job_id 52 nonce2 0 ntime 5a669bb1
    [09:19:42] data 00000007164c22d11c49b4c40ae8d9c99b71abe437b84fb4391151373369d589492e650543b34c77824e0e9e4279c9a3f3de2646a742e48328daec88a8137c06dec4444d5a669bb11b02909700000000 midstate 4fc5120d909386239149aa7532fec82d5b9f62f256d884f3c0b9c6fa42d7fd25b1ca0532976ad8310e7318816ab9ef20c912ec1cecb6da1e32b95699c1ac2b62
    [09:19:42] [THR0] Generated target 000000000000000000000000000000000000000000000000000068aa42000000
    [09:19:42] New block: 492e65053369d5893911513737b84fb49b71abe40ae8d9c91c49b4c4164c22d1… diff 25.6K
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184
    [09:19:43] WTF MUTEX ERROR ON LOCK! errno=0 in ../adl.c lock_adl():184

  7. After trying these I think im getting about 0.7 more MH out of my rx 570 4gb ITX. Do you think if I double my RAM on my machine to 8gb I will see an improvement? Also the threads at 4 crashed but on second attempt it worked ok?

    • I got 24 gb of ram on my system, an 8gb rx570 and i cant get more then 13 mh/s with this sucker. I really hate this fucking msi armor rx570 it is a real peace of shit and thermal are really bad when i OC it very little i get like +10-15°C. My Rx480 works like a charm and i get about 14.7-15mh/s. Think i might sell my rx570 and try to find an other rx480!

      • 32gb that’s overkill for a rig, isn’t it? I take it that it’s your gaming pc that you are using? Im not 100% happy with my current setup as Windows dropped WIFI connection last night leaving me not mining for a number of hours. I think I will need to look into home plugs to ensure a better connection. Unless Linux will do a better job for me?

        • Just an old pc that i had laying around, bought 2 gpu and a psu.

          By the way guys put your hashing power toward futurespool it is by far the most paying atm i get like 71% more then wattomine tells me

  8. I’d really like to thank you for the tutorial. Hard to find respect for the AMD. My 3x RX560’s are already thanking you. I didn’t speed up much (5%) but my temperatures dropped to that 60C range!

          • Try to OC to 1450-1475mhz corespeed and undvolt to -25 or -42 or until your corespeed is stable and dont jump all around the place, it was the same for me but the OC got my 570 to about 14.7 mh/s and my 480 to 15.8mh/s

          • i keep getting the same thing. I am using sgminer-phi-5.6.1-bitbandi-3-windows-amd64 as well and when i change the ––gpu-platform 2,1,0 the bat file just crashes

            GPU_MAX_ALLOC_PERCENT 70

            sgminer.exe -k phi -o stratum+tcp:// -u MYADDRESS -p c=LUX -I 19

            is what i have and it at least opens up to the pic doneqamili’s shared.

            I have tried different variations of the suggested Bat configs from the top of the article with no success. I am running 4 ADM RX570 Sapphire and 1 ASUS ROG STRIX 570. All cards are detected i can switch to a claymore miner fine. any suggestions would be greatly appreciated

          • Check if windows has deleted your sgminer.exe file. Windows defender has a habit of quietly deleting it without telling you, and the result is your bat file just crashing with no error.

  9. Hi i would like to make a question, i can only get to the next message:

    ” started sgminer 5.6.1-bitbandi-3
    *Using Jansson 2.7 ”

    And after that it only continues showing me that without further moves or showing anything else, as if crashed but the command tick appearing and dissappearing… any hope for me? RX580 and 8gb of ram

    • You probably just need to wait. It might take a little while to build the files it needs. Ive seen rigs take as long as 20 minutes to start the first time.

  10. Dude! First let me say that my rig was down for a few days and I was battling with the wallet update to 4.0. Finally getting that all set, I started up my bat file only to get a whopping 500kh/s per card! Argh! So, on to research a solution. I stumbled upon your article here. I took your advice on the updated sgminer and downloaded your batch file, tweaked it a bit and BOOM!!! Rig runs great now. Here’s my config: I’m running Win 7, 4 Gb RAM, 4 MSI RX470s, 4 Gb. I’m getting about 12.50 MH/s per card. It’s about 1.25 MH/s more per card using your bat settings. So far, so good!! You’re definitely having some LUX come your way bro!

      • Great tutorial Thank You!!. Using 18.2.1 driver since my system just wont run my AMD card using the blockchain driver.

        Hash rates varies around 16 MH/s but have seen 20 Mh/s. Set my gpu threads to 12 with no crash but I wonder if that’s what giving me fluctuation.
        Have another 4 Nvidia gpus running something else on this same rig.
        But I get fluctuation even running this alone.

  11. mem speeds can be reduced significantly, at least for RX580s and Vegas. I’ve dropped my mem to 1250 (there seems to be a threshold somewhere around there,) and still get 17MH/s. Also, as I mentioned on your initial guide, v5.6.1.1 of the miner seems much better (in raw performance, efficiency, and stability) than v5.6.1.3

    • Looks like w/ the older version, increasing intensity from 20 -> 22 produces additional gains of a couple percent (power is increased by the same or lower percentage, so it’s worthwhile.) Also, increasing thread-concurrency, while not producing noticeable gains, does stabilize the hashrate. Most threads talk about 8 x shaders being the magic number. I found 4x (e.g. 9216 for RX580) was enough to give me the desired effect.

      Finally, dropping gpu-threads from 8 to 3 had zero effect on a single card from my tests, and is actually necessary on my 8 gpu rig.

      • Yeah i’ve been testing on various rigs since you suggested it, and i’m finding very similar stuff. I do get a faster hash rate with 2 threads than I do with 1, but after that adding more does not seem to help. The interaction of the version with afterburner is weird- as soon as I open afterburner at all the hash rate drops, even with totally stock settings. Any overclocking attempts make it worse. I find if I just run the rig with no overclock it sits happily at 16.5mhs per card. I’m trying to find a way of forcing an overclock, as i’m pretty sure 17.5-18mhz is attainable, but i want to get there without increasing the core voltage.

        • Got some Nitros running around 17.4 @ 135W – that’s around 1375-1400 @ 965-985mV. They def could go higher, but that’s about where I like to keep them for heat and efficiency purposes – I generally find that over 1400, voltages needs increase dramatically.

          Try OverdriveNTool instead of afterburner – I don’t have any probs w/ it. Although in general, I run it via bat script before the miner, to set all the OC profiles programmatically.

          The only weirdness I’ve come across is that I generally like to disable/re-enable my cards before starting the miner (also as part of the startup script,) and doesn’t seem to like that. It complains about not being able to load the kernel or something.

1 Trackback / Pingback

  1. How to mine LUX coin with AMD – BlockChurn

Leave a Reply

Your email address will not be published.