automatic OC/UC with afterburner

[felixbrucker]

has anybody done that before, or is there a better way?

i have tested the following setup:

1) install afterburner and afterburner remote server
2) configure per miner clocks (and maybe voltages too, idk, i dont use that) somewhere (eg miner-manager)
3) make a http call to the remote server app to update the clocks
4) repeat 3) for each miner being switched

this should allow for automatic OC/UC depending on the miner being run and the gpu it is running on, also it works for nvidia cards too

ideas/suggestions?

cheers

[1715254487]

1715254487

[1715254487]

1715254487

[1715254487]

1715254487

[1715254487]

1715254487

[m1n1ngP4d4w4n]

has anybody done that before, or is there a better way?

i have tested the following setup:

1) install afterburner and afterburner remote server
2) configure per miner clocks (and maybe voltages too, idk, i dont use that) somewhere (eg miner-manager)
3) make a http call to the remote server app to update the clocks
4) repeat 3) for each miner being switched

this should allow for automatic OC/UC depending on the miner being run and the gpu it is running on, also it works for nvidia cards too

ideas/suggestions?

cheers

I use a call to a batfile that use nvidia inspector to do my OC/UC @ the start of the mining script depending on the algo

you can also do it with a combo cygwin/batfile if you want it to be done remotely.

here is a little snippy to get you started

Code:

%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:0,0,150 -setMemoryClockOffset:0,0,595 -setPowerTarget:0,60 -setTempTarget:0,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:1,0,150 -setMemoryClockOffset:1,0,595 -setPowerTarget:1,60 -setTempTarget:1,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:2,0,150 -setMemoryClockOffset:2,0,595 -setPowerTarget:2,60 -setTempTarget:2,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:3,0,150 -setMemoryClockOffset:3,0,595 -setPowerTarget:3,60 -setTempTarget:3,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:4,0,150 -setMemoryClockOffset:4,0,595 -setPowerTarget:4,60 -setTempTarget:4,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:5,0,150 -setMemoryClockOffset:5,0,595 -setPowerTarget:5,60 -setTempTarget:5,0,85
timeout 1

Nvidia Inspector

For remote/recurrent operation, just have a database/datafiles (flat csv or something more fancy) with the list of your miner & type of miner (mining rig 01 - nvidia / mining rig 02 - AMD and so on), another one with algos / type of mining rig (amd or nvidia) and the associated script to start mining, and just do a remote recursive request parsing those little flat file or your fancier database to remotely kill / start miner depending on algos, current profitability, type of rigs & so on.

You would probably also want a tool as a frontend that calculate profitability per type/model of cards to do your switching accordingly, this is about what i do atm, but less complex because i only have one type of rig & cards (12x rigs of 1070 gpus) but the idea is the same, just more data to process to make the correct decision for each rigs

I didn't liked to have my smartness all focused on one frontend, so what i did with my approach was to make each rig smart for it's own mining, and only gather datas remotely both way, basically i have a central server that calculate profitability, there is a tool on each rig that will check this current most profitable algo on the central server and do the algo switching accordingly, also the central server monitor each rig and get current mining status with API's when existing, if the rig is answering in a timely manner both network and/or miner and start some action if it's not (kill / restart miner, hard rig restart and so on), also matching against databases/datafiles to see if the rig is mining @ the proper hashrate more or less a few %. There is alot of ideas  

[felixbrucker]

has anybody done that before, or is there a better way?

i have tested the following setup:

1) install afterburner and afterburner remote server
2) configure per miner clocks (and maybe voltages too, idk, i dont use that) somewhere (eg miner-manager)
3) make a http call to the remote server app to update the clocks
4) repeat 3) for each miner being switched

this should allow for automatic OC/UC depending on the miner being run and the gpu it is running on, also it works for nvidia cards too

ideas/suggestions?

cheers

I use a call to a batfile that use nvidia inspector to do my OC/UC @ the start of the mining script depending on the algo

you can also do it with a combo cygwin/batfile if you want it to be done remotely.

here is a little snippy to get you started

Code:

%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:0,0,150 -setMemoryClockOffset:0,0,595 -setPowerTarget:0,60 -setTempTarget:0,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:1,0,150 -setMemoryClockOffset:1,0,595 -setPowerTarget:1,60 -setTempTarget:1,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:2,0,150 -setMemoryClockOffset:2,0,595 -setPowerTarget:2,60 -setTempTarget:2,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:3,0,150 -setMemoryClockOffset:3,0,595 -setPowerTarget:3,60 -setTempTarget:3,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:4,0,150 -setMemoryClockOffset:4,0,595 -setPowerTarget:4,60 -setTempTarget:4,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:5,0,150 -setMemoryClockOffset:5,0,595 -setPowerTarget:5,60 -setTempTarget:5,0,85
timeout 1

Nvidia Inspector

For remote/recurrent operation, just have a database/datafiles (flat csv or something more fancy) with the list of your miner & type of miner (mining rig 01 - nvidia / mining rig 02 - AMD and so on), another one with algos / type of mining rig (amd or nvidia) and the associated script to start mining, and just do a remote recursive request parsing those little flat file or your fancier database to remotely kill / start miner depending on algos, current profitability, type of rigs & so on.

You would probably also want a tool as a frontend that calculate profitability per type/model of cards to do your switching accordingly, this is about what i do atm, but less complex because i only have one type of rig & cards (12x rigs of 1070 gpus) but the idea is the same, just more data to process to make the correct decision for each rigs

I didn't liked to have my smartness all focused on one frontend, so what i did with my approach was to make each rig smart for it's own mining, and only gather datas remotely both way, basically i have a central server that calculate profitability, there is a tool on each rig that will check this current most profitable algo on the central server and do the algo switching accordingly, also the central server monitor each rig and get current mining status with API's when existing, if the rig is answering in a timely manner both network and/or miner and start some action if it's not (kill / restart miner, hard rig restart and so on), also matching against databases/datafiles to see if the rig is mining @ the proper hashrate more or less a few %. There is alot of ideas  

thanks for the detailed explanation

i was under the impression nvidia inspector only works with nvidia cards (which ruled it out for this purpose), does it also work with amd cards?

cheers

[m1n1ngP4d4w4n]

has anybody done that before, or is there a better way?

i have tested the following setup:

1) install afterburner and afterburner remote server
2) configure per miner clocks (and maybe voltages too, idk, i dont use that) somewhere (eg miner-manager)
3) make a http call to the remote server app to update the clocks
4) repeat 3) for each miner being switched

this should allow for automatic OC/UC depending on the miner being run and the gpu it is running on, also it works for nvidia cards too

ideas/suggestions?

cheers

I use a call to a batfile that use nvidia inspector to do my OC/UC @ the start of the mining script depending on the algo

you can also do it with a combo cygwin/batfile if you want it to be done remotely.

here is a little snippy to get you started

Code:

%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:0,0,150 -setMemoryClockOffset:0,0,595 -setPowerTarget:0,60 -setTempTarget:0,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:1,0,150 -setMemoryClockOffset:1,0,595 -setPowerTarget:1,60 -setTempTarget:1,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:2,0,150 -setMemoryClockOffset:2,0,595 -setPowerTarget:2,60 -setTempTarget:2,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:3,0,150 -setMemoryClockOffset:3,0,595 -setPowerTarget:3,60 -setTempTarget:3,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:4,0,150 -setMemoryClockOffset:4,0,595 -setPowerTarget:4,60 -setTempTarget:4,0,85
timeout 1
%~dp0nvidiaInspector-1.9.7.8\nvidiaInspector.exe -setBaseClockOffset:5,0,150 -setMemoryClockOffset:5,0,595 -setPowerTarget:5,60 -setTempTarget:5,0,85
timeout 1

Nvidia Inspector

For remote/recurrent operation, just have a database/datafiles (flat csv or something more fancy) with the list of your miner & type of miner (mining rig 01 - nvidia / mining rig 02 - AMD and so on), another one with algos / type of mining rig (amd or nvidia) and the associated script to start mining, and just do a remote recursive request parsing those little flat file or your fancier database to remotely kill / start miner depending on algos, current profitability, type of rigs & so on.

You would probably also want a tool as a frontend that calculate profitability per type/model of cards to do your switching accordingly, this is about what i do atm, but less complex because i only have one type of rig & cards (12x rigs of 1070 gpus) but the idea is the same, just more data to process to make the correct decision for each rigs

I didn't liked to have my smartness all focused on one frontend, so what i did with my approach was to make each rig smart for it's own mining, and only gather datas remotely both way, basically i have a central server that calculate profitability, there is a tool on each rig that will check this current most profitable algo on the central server and do the algo switching accordingly, also the central server monitor each rig and get current mining status with API's when existing, if the rig is answering in a timely manner both network and/or miner and start some action if it's not (kill / restart miner, hard rig restart and so on), also matching against databases/datafiles to see if the rig is mining @ the proper hashrate more or less a few %. There is alot of ideas  

thanks for the detailed explanation

i was under the impression nvidia inspector only works with nvidia cards (which ruled it out for this purpose), does it also work with amd cards?

cheers

Oh yes it works only with nvidia cards with such system it wouldn't be an issue, just use another scripted tool for AMD's like wattool if you build it roughtly like i said, you can know when to use one or the other, very easily 

[felixbrucker]

Oh yes it works only with nvidia cards with such system it wouldn't be an issue, just use another scripted tool for AMD's like wattool if you build it roughtly like i said, you can know when to use one or the other, very easily 

wattool only works for rx cards, or am i mistaken? i couldnt find a nice tool to easily set the clocks on, lets say, an amd 7970

if possible i only want one interface to change everything (which afterburner offers)

cheers

ps: i already have a similar structured mining management software with switching, the oc/uc settings would just be executed before starting the miner, everything can be customized (and its not intended for newbies so i dont have to auto select the correct stuff, the user will have to do that once )

[m1n1ngP4d4w4n]

Well im not really knowledgeable about AMD cards for mining since i only got nvidia's, there are other tools im pretty sure of it, maybe if someone pitch-in with the other alternatives but i don't know of a tool in cmdline able to change clock indifferently between nvidia & amd's.