Hydra, CUDA partial HEX/WIF/seed/passphrase recovery tool

[Julienbxl]

I present you Hydra : https://github.com/Julienbxl/Hydra

Here some benchmark with a RTX 5060 :

Mode   Target   Throughput
Hex   BTC legacy / SegWit   ~1,200 MK/s
Hex   Bloom BTC   ~1,000 MK/s
Hex   ETH & Bloom ETH   ~700 MK/s
Hex   Bloom (BTC+ETH)   ~550 MK/s
Seed   All modes   ~1,700,000 seeds/s
Passphrase   All modes   ~100,000 pass/s
WIF   All modes   ~900 MK/s


I hope some of you may find it useful

[flapduck]

Nice work, and props for actually posting real throughput numbers.

The big thing I always try to hammer home with these recovery tools is context: this stuff is only useful when you already own the secret and you're narrowing a small search space, like a few unknown words/characters or a fuzzy passphrase you mostly remember. If someone's hoping this turns into a magical button that can crack any wallet, they're gonna have a bad time, and honestly that's how it should be.

[Julienbxl]

I made several updates :

Hex — BTC legacy / SegWit / Bloom BTC   ~1,300 MK/s
Hex — ETH / Bloom ETH   ~800 MK/s
Hex — Bloom BTC+ETH   ~550 MK/s
Hex — BTC or ETH with known pubkey   ~2,500 MK/s
Seed   ~1,700,000 seeds/s
Passphrase   ~100,000 pass/s
WIF   ~2,800 MK/s

Mostly WIF improve speed and a ETH&BTC pubkey mode

There is also a resume mode

[DaveF]

Just tried it and have a small issue *that is probably on my end*.
Have 2 GPUs in the test bed a 3050 and a 4060.
Either one works alone but when both are in the system it dies when I run it. No error no nothing.

Now, it's a test bed so cards / drivers / software have been installed and uninstalled many times over.

Saw the post and had a few minutes to play with what you wrote before I had to head to a customer for a weekend install. Lucky me.

Just wanted to let you know. Will try to check more but with the end of month, end of quarter things going on this week probably not for a while.


-Dave

[Julienbxl]

Did you build it using sources or just downloaded the binary ?

The binary was compiled for sm120
Your generation is I think sm86 if I am not mistaken

I should have precised it in the releases notes

Or some problem with the selection of the GPU what if you force it this way :

nvidia-smi -L
CUDA_VISIBLE_DEVICES=0 ./Hydra <mask> <addr>
CUDA_VISIBLE_DEVICES=1 ./Hydra <mask> <addr>

[Julienbxl]

I don't know why Github suspended my account but it is back !

I made a Windows version available on the repo