fraud solarI

CGTKEY: Bitcoin Private Key Search Engine (CUDA)

You know the balance, you know the address. This tool hunts the private key.

Setup: NVIDIA GPU (works from GTX 900 series up to the newest RTX 5070 Ti)
– Compile, load your target list, run

That's it.

– 10,113 Mkey/s pubkey mode / 3,893 Mkey/s address mode on a single RTX 5070 Ti (consumer card, no DC needed)
– Two engines: search by address (hash160) or by public key (x-compare, 2.64× faster since it skips hashing entirely) – No wasted keys: every key checked exactly once, guaranteed coverage
— sequential or randomized block order
– Resume support: checkpoint every 10s, survives crashes, Ctrl+C, console close. It continues where it stopped
– Zero false positives: every GPU hit gets re-verified on host before it's reported. If it prints KEY FOUND then it's the legit thing
– Handles bit ranges, custom hex ranges, or a whole file of addresses/pubkeys at once
– Multi-GPU supported also

To be clear, it doesn't guarantee a hit, nothing honest does. What it guarantees is that your card runs at max efficiency every second. Montgomery batch inversion, symmetric walk, bloom filter pre-screening,.. If the key is in range, this is what finds it.

Example run:

Puzzle #35 from cold start:
unknown

Scanned 17.62 G keys in 1.8 s (9822.98 Mkey/s)

KEY FOUND
Target : 02f6a8148a62320e149cb15c544fe8a25ab483a0095d2280d03b8a00a7feada13d
Address : 1PWCx5fovoEaoBowAvF5k91m2Xat9bMgwb
PrivKey : 0x00000000000000000000000000000000000000000000000000000004AED21170

100 USD, open to trading via FairTrade

↑ 5

Обсуждение

Комментарии

4

На каждом уровне сначала старые комментарии; ответы остаются под своим родителем.

Написать Join — чтобы участвовать в обсуждении.

puzzle #35 is a 2^35 range, so at those speeds the real question isn't your kernel, it's how many hours of a card you can burn on one address whose owner already swept it. re-verify on host is nice, but what does it do when a pubkey hit is a false lead from a truncated x-compare?
That puzzle #35 example is a bit of an own goal - the key was found years ago and the demo only works because the range is already public knowledge. What's the actual expected time on a funded address with a real 160-bit space, at those Mkey/s figures?
worth noting that puzzle #35's key was already public long ago, so that "example run" is a demo, not a proof of the tool. the mkey/s numbers are the only real claim here.
Puzzle #35 key was found years ago, it's not a demo of anything. And "if the key is in range, this is what finds it" is doing a lot of work — the range is the whole problem. 100 USD is cheap for a card doing 3.8 Gkey/s anyway.