CUDACyclone - 6000Mkeys/s RTX4090

[FrozenThroneGuy]

Hello, guys!
Ultra lightweight CUDACyclone is ready, speed is 1.3Gkeys/s on RTX4060.
Key feature - extremely low VRAM usage for rented gpu.
It will work even if Vanity or Keyhunt doesn’t start.
And also this is a good studying sample for your education (why not)? Total 7 small files.
Link: https://github.com/Dookoo2/CUDACyclone

Test start RTX4090:
./CUDACyclone --range 400000000000000000:7fffffffffffffffff --address 1PWo3JeB9jrGwfHDNpdGK54CRas/fsVzXU --grid 512,512
======= PrePhase: GPU Information
Device: NVIDIA GeForce RTX 4090 (compute 8.9)
SM 128
ThreadsPerBlock 256
Blocks 65536
Points batch size: 512
Batches/SM: 512
Memory utilization: 14.5% (3.42 GB / 23.5 GB)
Total threads: 16777216
=== Phase-1: Brooteforce
Time: 30.1 s | Speed: 6038.9 Mkeys/s | Count:
182904267648 | Progress: 0.00 %


Grid: 512 - total threads per SM (max quantity per SM for Ada Lowelace) second 512 - total batches per SM (you can play with this parameter).

[kTimesG]

I don't wanna be an ass but what happened with JLP's copyright and license in your project? Do you understand what an GPL license implies?

[FrozenThroneGuy]

I don't wanna be an ass but what happened with JLP's copyright and license in your project? Do you understand what an GPL license implies?

KtimesG, old buddy! How are you?
I really don’t know anything about license, but link to JLP was added by me from the start of my repo. And FixedPaul too. This is a great guys who really do great code.
If this is not enough, I can paste anything about this guys in my repo, I’m not a stealer:)
And by the way, from JLP it has only part of CUDAMath, and thats all…

[MarangoZ]

Can someone with knowledge check all his code and see if its legit or not
as most of tools that are about puzzle of bitcoin have a backdoor that even if you find the Private Key it shows at you that u have not find it yet while it send the info to the owner of the tool


It looks like you've put a lot of detail into the GitHub repository but the post itself could use more context to help build trust and understanding.

BTW could you share what sets your tool apart from existing ones like KeyHunt and BitCrack?

[FrozenThroneGuy]

Can someone with knowledge check all his code and see if its legit or not
as most of tools that are about puzzle of bitcoin have a backdoor that even if you find the Private Key it shows at you that u have not find it yet while it send the info to the owner of the tool


It looks like you've put a lot of detail into the GitHub repository but the post itself could use more context to help build trust and understanding.

BTW could you share what sets your tool apart from existing ones like KeyHunt and BitCrack?

You can check it by yourself, ask AI for analyse it and find backdoor:)
VanitysearchA-bitcrack from GixedPaul (fastest opensource i saw previously) could give 5.15 Gkeys/ on RTX4090. I achieved 6.2Gkeys/s.
But I have a problem also!
GPU frequency increasing, but speed of work lower and lower after 1 hour of work (4090 from 6.2 to 4.9). THis is typical compute-bound with power throttling.
SM-clock growth → Mkeys/s drop with unchanged mem-clock — classic hit PL/TDP.
But it also have way to solve it, abd i will add it after few days. You need only fix GPU frequency.
This is a part of profiling, which i did and it show a problem:
Every 1.0s: echo '=== NVIDIA GPU Monitoring ==='; nvidia-smi --query-gpu=cloc...
Status 1
=== NVIDIA GPU Monitoring ===
GPU: 1200 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 33.2 s | Speed: 1354.8 Mkeys/s | Count: 44798860192 | Progress: 0.00 %

Status 2
=== NVIDIA GPU Monitoring ===
GPU: 1230 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 763.4 s | Speed: 1293.4 Mkeys/s | Count: 1007654382848 | Progress: 0.00 %

Status 3
=== NVIDIA GPU Monitoring ===
GPU: 1260 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 962.4 s | Speed: 1270.4 Mkeys/s | Count: 1263346131424 | Progress: 0.00 %

Status 4
=== NVIDIA GPU Monitoring ===
GPU: 1275 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 1110.3 s | Speed: 1264.0 Mkeys/s | Count: 1450542514688 | Progress: 0.00 %

Status 5
=== NVIDIA GPU Monitoring ===
GPU: 1335 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 1537.6 s | Speed: 1204.2 Mkeys/s | Count: 1979568127232 | Progress: 0.00 %

Status 6
=== NVIDIA GPU Monitoring ===
GPU: 1365 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 1699.5 s | Speed: 1176.0 Mkeys/s | Count: 2172581213696 | Progress: 0.00 %

Status 7
=== NVIDIA GPU Monitoring ===
GPU: 1470 MHz | Mem: 6250 MHz | Temp:  0°C | Power:   0/  0W | Draw:   0W
Time: 3097.3 s | Speed: 1090.8 Mkeys/s | Count: 3745662943392 | Progress: 0.00 %

[vaccar73]

Hello, guys!
Ultra lightweight CUDACyclone is ready, speed is 1.3Gkeys/s on RTX4060.

Hi FrozenThroneGuy

nice work man, and thank you for sharing it.

I ported it over to Windows exe as a Visual Studio project using Chat GPT
I was not able to retain Linux, so it is just a Windows project.

Anyway I tested  it on a Rtx 3060M Frankenstein card,here is the speed


C:\CUDAWindowsPort_VS2022\x64\Release>CUDAWindowsPort --range 1000:22382FFFFF --address 1HBtApAFA9B2YZw3G2YKSMCtb3dVnjuNe2 --grid 512,256
======== PrePhase: GPU Information ====================
Device               : NVIDIA GeForce RTX 3060 Laptop GPU (compute 8.6)
SM                   : 30
ThreadsPerBlock      : 256
Blocks               : 7680
Points batch size    : 512
Batches/SM           : 256
Memory utilization   : 22.6% (1.35 GB / 6.00 GB)
-------------------------------------------------------
Total threads        : 1966080

======== Phase-1   =========================
Time: 136.8 s | Speed: 997.4 Mkeys/s | Count: 145869780992 | Progress: 99.25 %%

======== FOUND MATCH! =================================
Private Key   : 00000000000000000000000000000000000000000000000000000022382FACD0
Public Key    : 03C060E1E3771CBECCB38E119C2414702F3F5181A89652538851D2E3886BDD70C6

C:\CUDAWindowsPort_VS2022\x64\Release>