Mining with 1x 6GPU's or 6x 1GPU's

kruisdraadkruisdraad Member Posts: 61
Anyone has experience with mining per card and per group of cards? What i mean here is:

per card = starting a ethminer with explicit a specific GPU and that 6 times
per group = starting a single ethminer with 6 cards

my own experience shows me that if you have 6 cards and 1 misbehaves (whatever the reason) then the entire group will go down. Using 6 threads gives a little less MH/s (but thats prolly the openCL part, for some reason with -U/Cuda you cannot specify cards @Genoil ? ) but if a thread dies you;d still have 5 running.

thoughts / idea's anyone?


  • Dr_House86Dr_House86 Member Posts: 566 ✭✭✭
    Claymore miner restart when one card fails.
    Works fine.
    Dunno why but gpu3 hangs sometimes, the miner restart and all run as it should.
  • GenoilGenoil 0xeb9310b185455f863f526dab3d245809f6854b4dMember Posts: 769 ✭✭✭
    edited June 2016
    --cuda-devices \
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    Yes, I have quite a bit of experience with that. :) I found, as you did, that a single card hanging, when running a single, threaded miner, hangs the whole rig. Sometimes (infrequently) a reboot won't even clear it; I have to power cycle. I run Ubuntu, and I also found you can look at 'top' output and see the hung ethminer, when you run one ethminer per card, because it will have, not surprisingly, 0 CPU utilization all the time. That allowed me to write a simple script and set up a root cronjob that detects a hung ethminer/GPU and automatically reboot the rig. The reboot always clears the problem; I never need to power cycle. I've ported @Genoil 's "auto-DAG" feature to stock ethminer, so the reboot is super quick. I run the root cronjob every 2 minutes.
  • kruisdraadkruisdraad Member Posts: 61
    @dlehenky a reboot does not reset the GPU, but you can force a reboot with the SMI (even without rebooting Ubuntu).

    But you mine with 6 cards now or just 6x 1?

    Also another tip: in the latests drivers you get updates if problems are there.

    @Genoil grrrrr i really looked over that one a gazillion times, sorry
  • dlehenkydlehenky Member Posts: 2,249 ✭✭✭✭
    @kruisdraad I'm really not sure what your saying. I now mine with one ethminer per GPU, 6 GPUs per system. If one GPU gets hung, and I kill the process, the process turns into a zombie, and that process can only be cleared with a reboot (linux). So, whether a reboot "resets the GPU" or not, it solves the problem and everything comes back up running correctly. I should know, they're my systems, and I have observed this outcome a number of times. If your tip on the latest drivers is based on Windows, it doesn't apply to linux, since the linux AMD drivers are updated *much* more slowly that the AMD Windows drivers. Besides, I don't have a problem to solve. :)
  • scootyscooty Member Posts: 14
    >:) >:) >:) >:) watch out with this guy, he try to pull me off with $100 and not sharing any information how he setup my server. what a SCAM!!!! pretty sure he would put some coin stealer in the system once start mining. >:) >:) >:) >:)
  • workwork Member Posts: 2,084 ✭✭✭✭
    @scooty who are you talking about?
  • kruisdraadkruisdraad Member Posts: 61
    @work he is trying to crosspost and add imflamatory remarks. Just ignore him
  • kruisdraadkruisdraad Member Posts: 61
    @dlehenky if you reboot your hanging proces in linux is gone. It hangs because it cannot 'clear' all the jobs at the GPU because its gone wrong. My experiance is if you reset the card using SMI its responding again and you process will die (because it starts getting replies again). That said, my tooling will only try twice to resolve it else its automaticlly rebooting the server. I've added special stuff and created a installation (Ubuntu custom) so that a boot is done in 2 seconds ... so why just dont reboot?
Sign In or Register to comment.