Solving ECDLP with Kangaroos: Part 1 + 2 + RCKangaroo

[slammer33]

Wow!

Congrats RetiredCoder!

That was really impressive... 

[satashi_nokamato]

Hello Op, first congratulations on those puzzles you solved, you deserved them without a doubt, used your mind+ hardware and got the prize.

Just wondering if there are any benchmark stats on different GPUs with all your versions? If there is, it would be kind of you to share them on OP, please. Thank you and wish the best for you going for #140. 

[optioncmdPR]

This tool has an error in the . JS regarding calling of a non existant
function  related to u64 causing all zero no memory failure flag when trying to use GPU. Has anyone actually successfully gotten their gpu to work with it?

[nc50lc]

This tool has an error in the . JS regarding calling of a non existant function related to u64 causing all zero no memory failure flag when trying to use GPU.

Are you pertaining to RCKangaroo: github.com/RetiredC/RCKangaroo?

What are the GPUs that you've tried so far; AMD, Nvidia? What are the specific models?
Because it'll only work on select models.

[3dmlib]

Finally I had some time to prepare RCAsm for people who are brave enough to make their CUDA kernels faster
https://github.com/RetiredC/RCAsm

Also check Kernel01 sample for SASS implementation of MulMod256.

Hello.
Can these ASM kernels use in VanitySearch-Bitcrack to make it faster? Program used for brute-force 71puzzle. Or they valid for RCKangaroo only?
Thanks.

[ethanhunt2023]

Mini-puzzle #7 closed.

#135 pk is: 0000000000000000000000000000006D9392A16883F90903D5F78DA57AF07EB2

I’m not judging others' actions, but if I had solved puzzle 135—which is worth around $1 million—I would have made, say, 12 donations of $1,000–$2,000 each. A "puzzle 7" worth $280 is far too little; someone with 0.5 BTC—not 13 BTC—could donate $280, IMHO.

[Leonardo di ser Piero]

Good evening everyone,

I would kindly like to ask if anyone could provide an estimate of the time required to find a target using the new **RCKangaroo V4** with **8× RTX 5090 GPUs**.

In particular, I would be interested in the estimated solving times for the following bit ranges:

- 92 bits
- 96 bits
- 98 bits
- 102 bits
- 104 bits
- 106 bits
- 110 bits

If anyone has already tested this configuration or has benchmark results to share, I would greatly appreciate your feedback.

Thank you very much in advance for your time and help!

[Cricktor]

...

And exactly why don't you try to find it out yourself how performant such a rig is on bitranges you're interested in?

It's not rocket science to construct known sample cases and try to solve them with such a rig, unless you don't have it or don't want to spend money on renting eight of those GPUs.

[Leonardo di ser Piero]

I apologize if my question was inappropriate or bothered you in any way. I was simply hoping to get a rough estimate of the time required, just for reference. If you could share an approximate answer, I would really appreciate it. Thank you again for your time and understanding.

[Cricktor]

Sorry, no, can't help. I don't own or want to afford to rent such expensive gear.

[ABCbits]

Good evening everyone,

I would kindly like to ask if anyone could provide an estimate of the time required to find a target using the new **RCKangaroo V4** with **8× RTX 5090 GPUs**.
--snip--

Sorry, no, can't help. I don't own or want to afford to rent such expensive gear.

I wouldn't want spend any money to rent GPU either. But FYI, there are few online P2P marketplace with short rent duration and reasonable cost. I do not vouch any of them, since who knows if there's backdoor or data theft. But there are few examples,

• vast.ai shows 1x RTX 5090 start from $0.56/hour
• clore.ai shows 1x RTX 5090 start from $0.88/hour

[Leonardo di ser Piero]

Cricktor . Who asked you to rent the GPUs or run any tests for me? I simply asked for an estimated solving time, that's all.

Honestly, your hostility is completely unnecessary. You should take a page out of RetiredCoder's book. He willingly shares his knowledge and results with everyone, out of a genuine desire to help, not with hostility or arrogance.

You could have simply answered my question or chosen not to reply at all. There was no need to turn a simple request for information into an argument.

[WanderingPhilospher]

Good evening everyone,

I would kindly like to ask if anyone could provide an estimate of the time required to find a target using the new **RCKangaroo V4** with **8× RTX 5090 GPUs**.

In particular, I would be interested in the estimated solving times for the following bit ranges:

- 92 bits
- 96 bits
- 98 bits
- 102 bits
- 104 bits
- 106 bits
- 110 bits

If anyone has already tested this configuration or has benchmark results to share, I would greatly appreciate your feedback.

Thank you very much in advance for your time and help!

These are theoretical averages for a normal search at sustained 152 GKey/s (19 GKey/s x 8 each 5090s) with K = 1.15.
92 bits → 91-bit width: 6 minutes 16 seconds
96 bits → 95-bit width: 25 minutes 6 seconds
98 bits → 97-bit width: 50 minutes 12 seconds
102 bits → 101-bit width: 3 hours 21 minutes
104 bits → 103-bit width: 6 hours 42 minutes
106 bits → 105-bit width: 13 hours 23 minutes
110 bits → 109-bit width: 2 days 5 hours 32 minutes

[Leonardo di ser Piero]

Thank you so much, WanderingPhilosopher, for your kind and helpful reply. I truly appreciate your generosity, your humility, and your willingness to share your knowledge and experience with everyone.
People like you, who are happy to help others and share their knowledge without hesitation, make this community a better place for everyone.
Thank you again for taking the time to answer my question and for your kindness. It is greatly appreciated! 🙏

[SleepycatSwap]

Been watching from the sidelines such a interesting puzzle well done to the solver must of been some hardware for all those jump tables!

Keep going! 

[satashi_nokamato]

~

Hello. One question about your fork of this tool;  why there is no option to save work to continue later?

And if it's not off topic and OP kindly tolerate it.
On your keyhunt fork, why setting k to 2048 throws an error and first bloom init is 0 points but takes up 15 GB?

Edit
@OP
I have tried your version with exactly same range and dp in a 80 bit space, it found a 76 bit key in 90 seconds , but with the same setup, it couldn't find a 76 bit key ending with 8 zero at the end of it's k. Hexadecimal. So after 10 minutes I terminated it.  Though speed was exactly as promised.

I thought kangaroo can not skip,  miss any key.

[Anas35]

hello RC,

are you available. i want you to know theres this backdoor from the hashes can you please reply, so i will explain you how its done and you may be able to figure out the steps to match positional fits.
PM me