Multi GPU on Multi Server Question

[fammari]

Hello Everyone,
Is it possible to combine the power of multi GPU on multiple servers? let us say i want to rent 5 instances from Vast.ai whereby each instance has 2x or 4x GPUs (1080 or 2080 etc..), is there a way to connect them all whereby BTCrecover or hashcat or any other software can see them as one server to use the total power?

Appreciate your help.

[mocacinno]

For hashcat, there should be sollutions, eventough i've never tried them, and they're probably not a one-click-sollution (but rather a manual setup);

https://hashcat.net/wiki/doku.php?id=frequently_asked_questions#how_can_i_distribute_the_work_on_different_computersnodes

Like i said: never tried this myself... But it should work...

[fammari]

For hashcat, there should be sollutions, eventough i've never tried them, and they're probably not a one-click-sollution (but rather a manual setup);

https://hashcat.net/wiki/doku.php?id=frequently_asked_questions#how_can_i_distribute_the_work_on_different_computersnodes

Like i said: never tried this myself... But it should work...

Brilliant, Thank you so much for your help.

[o_e_l_e_o]

I assume it is a wallet password you are trying to break?

btcrecover is able to split any work across multiple devices/CPUs/GPUs, and even has a tutorial page specifically for doing this with Vast.ai, available here: https://btcrecover.readthedocs.io/en/latest/Usage_Examples/2020-10-06_Multi-GPU_with_vastai/Example_Multi-GPU_with_vastai/

If you are trying to brute force a wallet password, you'll want to create a wallet extract first, so whichever server (hopefully) finds your password is still unable to access your coins. Tutorial page for this available here: https://btcrecover.readthedocs.io/en/latest/Extract_Scripts/

[fammari]

You are right, i am trying to recover my wallet password.

Thanks for the info. i guess you can use multiple GPUs using mentioned article but you can't combine the power of all distributed servers in one single process. what i want is to have the processing power of all rented servers ( let us say 5 servers whereby each server has like 5 GPUs), so total processing power will be (5 GPUs power X 5 Servers).

will BTCRecover distributed solution do that? as far i know Hashcat will do it using my desired way right?

Having wallet extract is an excellent idea, i will see how to do it then try out the distributed solution.

Cheers

[khaled0111]

but you can't combine the power of all distributed servers in one single process. what i want is to have the processing power of all rented servers ( let us say 5 servers whereby each server has like 5 GPUs), so total processing power will be (5 GPUs power X 5 Servers).

Actually you can. You are not going to combine the power of all thisr servers in a single process, rather, you will split the work among them.
The "--worker" command on BTCRecover is what allows you to do this.
Those servers don't even have to connect to each other.
You can read the documentation about the "--worker" command and how to spread the work across multiple servers here: https://docs.btcrecover.org/en/latest/Usage_Examples/2020-05-23_multi_device_descrambling_12_word_seed_with_extras/Example_multi_device_descrambling_12_word_seed_with_extras/

[o_e_l_e_o]

will BTCRecover distributed solution do that?

Yeah, the command --worker as suggested by khaled0111 should achieve this. You would run the same command on each server, with the exception of --worker 1/5 appended to the command for the first server, --worker 2/5 for the second server, and so on. Each server will then check every 5th password. btcrecover should automatically make use of all the GPUs available on each server to achieve optimal speeds.

If you end up renting uneven servers, for example one server with double the GPU power of the others, then you can assign it double the work compared to the others by using the command --worker 1,2/5