<- Return to getpimp.org

Miner Farm question about HW errors



  • When viewing the stats for the GPU in Miner Farm I see a lot of HW errors. What does that mean? Should I be concerned?


  • Staff

    Great question!

    On AMD especially, ETH generates a lot of them, they should not be a problem. If you are getting more HW than Accepted shares, that may be an issue. Try providing us a screenshot or paste if you still need more feedback. Thanks!



  • I am consistently getting about twice as many HW than Accepted shares. That is the best I can do - what are the HW's and why do they occur and how can we control them? I get them on stock Bios or Modded Bios to varying degrees depending on how much I played with them.



  • It seems to me like you're pushing your GPUs too hard? Perhaps you can OC them less and adjust your miner worksize, intensity, gpu-threads settings to more conservative and see if there is a change in behaviour?



  • Paz,

    I have no idea where to edit those worksize, intensity, gpu-threads settings and what values are valid for an RX580 Hynix mining Ethereum. Can you help me with that? I am relatively new at mining but comfortable with linux.



  • I reset my GPU's to original "Asus RX580 Dual OC 8GB Hynix Ram" bios and played with the /root/gpu-config.json file but every setting creates about the same rate of HW errors and drew a lot more current (1350 watts) than my modded bios the runs a 6 GPU rig at 1000 watts. If I went with your default 1200 and 1800 settings same HW error rate and high power consumption.

    So .... I am back to my modded bios because I won't burn out the EVGA 1300G2 power supply running at 1000 watts according to my watt meter on the power cord.

    I do have a spare power supply that I could use to reduce the stress on the EVGA 1300G2 power supply by taking 3 of the GPUs power from the spare power supply to test out theories but it wouldn't solve anything regarding the HW error rate that I am reporting.

    I really want to start with an original non-modded BIOS and be able to adjust the core voltage and me voltage using the /root/gpu-config.json file but I don't know the parameters to add to make that happen. Do you have any reference document with those core and mem voltage variables so that I can try it?

    I am down to thinking it is simply the AMD driver that is driving me nuts.



  • right now you cannot set core voltage from gpu-config. the team is investigating linux software to see if something can do that. but it will take some time.

    for right now, there is OhGodAnOffset which is a paid undervolting tool for linux which you can find on google.



  • guys,

    can you help to tell if this HW acceptable?
    GPU0: 55C 2054RPM | 21.93M/21.91Mh/s | R: 0% HW:689 WU:20.923/m xI:2048
    GPU1: 56C 2081RPM | 21.93M/21.91Mh/s | R: 0% HW:629 WU:21.276/m xI:2048

    pimp 2.4.2
    2 x Sapphire RX580 +nitro but no OC, just run it out of the box w/o any tweak
    pool ethermine.org (asia server)
    is xI in the correct value for RX 580?
    up and running for 5 days now with avg 0.00339 daily earning (2 GPU)

    any suggestion or comment?



  • My RX580s run at 30.14 after using Polaris Bios Editor one click timing patch. I'm also using Claymore instead of sgminer, it feels like its way more stable to me.



  • Paz, how to use Polaris Bios Editor through pimp?

    My rx580 hash is no diff than rx570, i think something is not right ....
    undervoltage perhaps? any idea how to do it?



  • you can use polaris bios editor from the desktop of pimp. connect a monitor to your rig, or use VNC Viewer to connect remotely (or MS remote viewer). as for how polaris works, there are lots of great youtube videos. i prefer not to use a downloaded bios. I use atiflash to save the stock rom, and then polaris to use the one click patch which is literally clicking one button and then saving the bios. you then reflash it using atiflash again. but i suggest you read up on it full and watch several videos before you undertake this. you are voiding the warranty and if something goes wrong, your gpu could be bricked permanently with no replacement.



  • googling polaris bios editor, and i dont think i could use it, dont understand how to fill the numbers

    anyway, i changed 2 RX 580 4GB with 4 RX 570 8GB, and meddling with intensity and thread, got very low HW and stable hash at 23.64
    think that should do it for now till pimp update with undervoltage (maybe?)

    building new rig with 4 RX 580 now (2 4gb + 2 8gb) and hope achieve better than 570's



  • I have been using /opt/confs/sgminer.eth.0_1518492883194_fixed hw errors2.jpg conf
    I was able to adjust the xintensity and the gpu-threads.
    I use vi, edit is for wimps :-P

    I have 4 RS580 cards as of now, I adjusted to 4 threads and ran with xintensity of 13.

    The threads can be less than the number of cards, that's just what I decided to try out.

    No HW errors occurred so I increased xintensity to 26. Very few errors, better hashrate. It's a slightly lower hashrate than the original setting (which was ridiculously high)...but no HW errors. I know the errors shouldn't cause a problem...and they did seem to count all the same with the pool...but I don't like seeing errors.

    I will keep trying to optimize the settings further as there are various options that can be tweaked. There was enough information within the forums to be able to reach out and see the options. Just those two settings seemed to resolve it for me.


 



Want 10% more hash from your rigs?



We promise to keep your email safe and never spam you.



Copyright (c) 2017 PiMP LLC. All rights Reserved.

Looks like your connection to PiMP Forum was lost, please wait while we try to reconnect.