Re: PhoenixMiner 5.7b: fastest Ethereum/Ethash miner with lowest devfee (Win/Linux)

thanks for pointing it out tho i wouldnt mine with 6600's unless i found a lorry full in a field lol. As for Linux it can be challenge but very rewaring to break free of MS tho sometimes it can make one want to smash things up in a rage if its not bad enough learning new HW then throw in an unfamaliar distribution i been on Ubuntu till now having moved to Manjaro to get these 6800s singing without windows such is the passion for that here.

6 x RX6800XT up, undervolted and playing nice under linux.

except the stale shares to sort out....

GPU1: 48C 60% 147W, GPU2: 45C 60% 155W, GPU3: 51C 60% 151W, GPU4: 50C 49% 153W, GPU5: 45C 49% 145W, GPU6: 40C 49% 148W
GPU1: cclock 1200 MHz, cvddc 868 mV, mclock 1000 MHz, Tj 54C, Tmem 66C, 403 kH/J
GPU2: cclock 1200 MHz, cvddc 868 mV, mclock 1000 MHz, Tj 53C, Tmem 74C, 391 kH/J
GPU3: cclock 1200 MHz, cvddc 862 mV, mclock 1000 MHz, Tj 57C, Tmem 70C, 401 kH/J
GPU4: cclock 1200 MHz, cvddc 868 mV, mclock 1000 MHz, Tj 58C, Tmem 64C, 387 kH/J
GPU5: cclock 1200 MHz, cvddc 862 mV, mclock 1000 MHz, Tj 53C, Tmem 62C, 408 kH/J
GPU6: cclock 1200 MHz, cvddc 868 mV, mclock 1000 MHz, Tj 48C, Tmem 56C, 400 kH/J
GPUs power: 899.0 W; total power: 1099.0 W; cost: 6.07 USD/day; 326 kH/J
Eth: New job #9a8fdd86 from eth-eu1.nanopool.org:9999; diff: 10000MH
Eth: New job #beaaa833 from eth-eu1.nanopool.org:9999; diff: 10000MH

*** 0:09 *** 10/3 22:34 **************************************
Eth: Mining ETH on eth-eu1.nanopool.org:9999 for 0:09
Eth speed: 357.727 MH/s, shares: 23/0/0, time: 0:09
GPUs: 1: 60.585 MH/s (6) 2: 59.152 MH/s (4) 3: 59.165 MH/s (4) 4: 59.154 MH/s (1) 5: 60.514 MH/s (3) 6: 59.157 MH/s (5)
Eth: Accepted shares 23 (5 stales), rejected shares 0 (0 stales)
Eth: Incorrect shares 0 (0.00%), est. stales percentage 21.48%
Eth: Maximum difficulty of found share: 772.1 GH (!)
Eth: Average speed (5 min): 355.251 MH/s
Eth: Effective speed: 425.14 MH/s; at pool: 425.14 MH/s

time for a day or two away let the hash crack on...

Have you tried changing the F-state on linux using a script? You can get a stable 62.5 MHs on average for each card.

Check out the link below, it was built for HiveOS (which is an Ubuntu variant) so it might prove useful for you:

https://forum.hiveos.farm/t/rx6800-efficient-overclocking/26448/330

Thanks for the reply i will take a look at the link currently hoovering up info as want to solve the stale share problem which im sure is caused by being stuck on kernel 0 for now.

Actually had a breakthrough late last night just before bed it was a "what if a try this" whilst my inner voice was "dont mess about when your this tired just go to bed man" moment, but after a fk it, it turns out two of the 6800's dont like mclock @ 1075 and drop down to 10Mhs with it. I suspect it it could be throttling as when i drop mclock down to 1050 on those two cards they come back to 63Mhs but with hotter mem temps than the others so the following is how i went to bed and woke with the rig still purring away nice n stable at around 380Mhs. :-)

-clkernel 0
#-clNew

-cvddc -165
-cclock 1200
-mclock 1125,1050,1125,1125,1050,1125

-tmax 60
-tt -60,-70,-60,-50,-50,-50
#-gt 6

GPU1: cclock 1200 MHz, cvddc 868 mV, mclock 1074 MHz, Tj 58C, Tmem 72C, 420 kH/J
GPU2: cclock 1200 MHz, cvddc 862 mV, mclock 1049 MHz, Tj 56C, Tmem 78C, 388 kH/J
GPU3: cclock 1200 MHz, cvddc 868 mV, mclock 1074 MHz, Tj 62C, Tmem 76C, 398 kH/J
GPU4: cclock 1200 MHz, cvddc 862 mV, mclock 1074 MHz, Tj 62C, Tmem 70C, 390 kH/J
GPU5: cclock 1200 MHz, cvddc 868 mV, mclock 1049 MHz, Tj 58C, Tmem 66C, 414 kH/J
GPU6: cclock 1200 MHz, cvddc 868 mV, mclock 1074 MHz, Tj 53C, Tmem 60C, 416 kH/J
GPUs power: 941.0 W; total power: 1141.0 W; cost: 6.30 USD/day; 333 kH/J

Eth speed: 383.928 MH/s, shares: 1862/1/21, time: 14:51
GPUs: 1: 63.194 MH/s (285/2) 2: 63.477 MH/s (330/5) 3: 64.645 MH/s (330/2) 4: 64.643 MH/s (287/2) 5: 63.488 MH/s (320/4) 6: 64.482 MH/s (311/6)

Now the above is all running under Manajaro which is a new experience for me with subtle differences to Ubuntu like no @reboot under crontab lol.

I have a test rig running Ubuntu 20.04.3 with AMDGPU driver 21.20 working nice for one GPU but no more and with no stales as able to use optimized kernels. I would like it very much if PM could release next version to solve the multiGPU issue with latest rocr driver which according to link below is related to rocr driver having PCIE 3.0 atomics limitation and i suppose a skeptic might scowl at AMD for messing with miners in that way unlike Nvidia and there LHR darkness. I haven yet been able to get ubuntu working with multigpu but even if possible its obvious i will be back to driver 20.40 stuck on kernel 0 same as Manjaro... On a side note i did try driver 21.30 just for a test, it crashes after building DAG for single GPU.

https://www.reddit.com/r/gpumining/comments/ptmyjd/ubuntu_20043_amdgpu_2130_opencl_rocr_rocm/

You should be able to get the CVDDC down in the 650 range. I'm running my 3x 6800's in HiveOS clock 1325, mclock 1075 vddc 612, 681,643. This results in 102W,120W,102W. All my cards are Powercolor. The 102W are Red Dragons the 120w is a Fighter. When I ran them in Windows 10, I was getting getting similar power settings after tweaking the powerplay tables with MorePowerTool. I switched to HiveOS because it was easier to try out different ETH miners. Pretty much click and go for that.

interesting thanks for the reply. I have resisted any sort of specialist mining operating system so far but i find this may warrant an exploration of HiveOs given PM is slowly growing obsolete at present although i dont like the registration bit for it. Need to find out who owns Hiveos and therfore who is benefiting from such data harvesting.