Bitcoin puzzle transaction ~32 BTC prize to who solves it

[mcdouglasx]

Huge respect for WP, I think his tutorial is great for education purposes.
But a tip for a tutorial on how to sign a tx offline and paste it on a webpage is a little too much to ask Especially since I didn't even see it until a couple days ago.

I don't think anyone had heard of the MARA service if @wp hadn't discovered it. I believe you're undervaluing their valuable contribution. It's important to highlight and give credit to those who deserve it, as their contribution opened new perspectives and solutions.

[benjaniah]

Guys, stop trying to pressure the solver into doing anything. If you're nice, maybe he will stay and you can learn something from him.
We are all grateful for WP's post, sharing a method to securely withdraw a puzzle prize. Nobody is obligated to tip anyone just because you used MARA. You can always show your appreciation for what you've learned by directly sending a tip yourself.

Over the past few months, despite multiple tests of the reliability of MARA's service being successful in safely transferring funds from one wallet to another, people repeatedly continued to cast doubt about MARA. Everything from them having "rogue" employees to hackers having SSH backdoor access to MARA. That was all just a bunch of nonsense. MARA is a publicly traded US company, listed on the NASDAQ stock exchange, with a real board of directors, accountable to their shareholders and clients, and probably with US regulators breathing down their necks, making sure they comply with demands from the SEC, IRS, OFAC, and various law enforcement agencies. There has not been a single instance of a transaction submitted to MARA's Slipstream service being compromised in any way. So give them some credit.

Not too long ago, I used MARA to send my own bitcoin to perhaps the most compromised wallet in the universe (hex private key 0000000000000000000000000000000000000000000000000000000000000001, WIF: KwDiBf89QgGbjEhKnhXJuH7LrciVrZi3qYjgd9M7rFU73sVHnoWn, compressed address 1BgGZ9tcN4rm9KBzDn7KprQz87SZ26SAMH) and then successfully sent it back to myself. I even enabled RBF for both transactions. Any bitcoin that is normally sent to this address gets instantly snatched up by bots. If MARA was compromised in any way, bots would have stolen my coins. But that didn't happen.

Tx from me to address #1: https://mempool.space/tx/bab4b7409ee1955bda0e20e6777ebc9d108a38a987f234e99b97cc0c7c97a019
Tx from address #1 back to me: https://mempool.space/tx/52a62d981c3ca21af5529eaed25599e515c2d5f26a549448e1743b966995df59

[mcdouglasx]

Guys, stop trying to pressure the solver into doing anything. If you're nice, maybe he will stay and you can learn something from him.
We are all grateful for WP's post, sharing a method to securely withdraw a puzzle prize. Nobody is obligated to tip anyone just because you used MARA. You can always show your appreciation for what you've learned by directly sending a tip yourself.

It is not about a donation; that is up to each person if they feel like donating or not. It is more about valuing the contributions and not undermining them. He talks as if it were a 100% original milestone. It is obvious that his solution is just a method already exposed here, and its only advantage was great computing power. So, unless it is something else, the answer is simple: he has nothing to teach us that we do not already know. From what he says, random method with thousands of GPUs are things we already know. I suppose the most he could have added to the software is a database that records already scanned sectors to skip them later.

[benjaniah]

Guys, stop trying to pressure the solver into doing anything. If you're nice, maybe he will stay and you can learn something from him.
We are all grateful for WP's post, sharing a method to securely withdraw a puzzle prize. Nobody is obligated to tip anyone just because you used MARA. You can always show your appreciation for what you've learned by directly sending a tip yourself.

It is not about a donation; that is up to each person if they feel like donating or not. It is more about valuing the contributions and not undermining them. He talks as if it were a 100% original milestone. It is obvious that his solution is just a method already exposed here, and its only advantage was great computing power. So, unless it is something else, the answer is simple: he has nothing to teach us that we do not already know. From what he says, random method with thousands of GPUs are things we already know. I suppose the most he could have added to the software is a database that records already scanned sectors to skip them later.

Did you read his methodology? It wasn't only random scanning and lots of GPU's.

The method was:

Break up puzzle 67's keyspace into 256 sub-ranges each with 2^58 keys
In each sub range, save every private key that generates an address starting with 48 zero's
Statistically there are 1024 proofs in each subrange
If an average of 1024 proofs are found in a sub-range, it statistically guarantees the whole sub-range has been scanned

What software are you talking about? He used custom software. I'm sure he'd like to keep his competitive advantage to himself.

It was found after scanning 57% of the keyspace in 67 days. Which works out to over 7 trillion keys per second. Which, my napkin math says is around 1600 x RTX4090's.

Several thousand

Pretty close. Would need more than 1600 GPU's if they were not all RTX-4090s.

And right now, ~4 days and 2 hours since puzzle 67 was solved, assuming the same scanning speed, he's already scanned almost 2% of puzzle 68.

[mcdouglasx]

Did you read his methodology? It wasn't only random scanning and lots of GPU's.

The method was:

Break up puzzle 67's keyspace into 256 sub-ranges each with 2^58 keys
In each sub range, save every private key that generates an address starting with 48 zero's
Statistically there are 1024 proofs in each subrange
If an average of 1024 proofs are found in a sub-range, it statistically guarantees the whole sub-range has been scanned

You must cite the argument you are referring to here in order to cross-check and validate the information; otherwise, it can be misinterpreted. So according to this, was it a pattern search using bits?.

but...

I don't wan't to be disrespectful to anyone, but most of the theories I see here about patterns look crazy to me.
The ONLY way I could see a pattern between puzzles is if the creator messed up the randomess (unsecure RNG, predictable seed choice, etc...) and I really don't think he would make such a mistake.

I'm not sure if it really refers to prefix search as well. In any case, the only search options would be:

1- scan the entire range.
2- random search.
3- probabilistic search based on prefixes.

All these ideas have been discussed here, so the code is the least important part. Once the idea is created, anyone can replicate it.

[walletrecovery]

https://btcpuzzle.info/puzzle/67
It's hard to believe that Puzzle 67 was opened by simple enthusiasts and not by the creators themselves, since the pool was often attacked by DDOS and 6,69% of the range was already passed, and since the creators of the puzzles know the private key, they can easily monitor how close the pool participants got, so they could easily take their coins themselves under the guise of someone finding them, so this is most likely a scam than some kind of honest competition. Who wants to give strangers more than half a million dollars, it's unrealistic! Let's think about it. Does anyone doubt that this is how things are or do you think everything is honest?

It was open by me, and I'm not the creator of the puzzle.
Let's put those conspiracy theories to rest.

What program did you use for scanning and what equipment do you have, if these are video cards, then which ones and how many? Can you make a transaction to my wallet to prove your words???
What power did you use, how long did it take you to scan the ranges, where did you look for the key, in a pool or solo, and how will you prove that you were the one who opened puzzle 67? Thanks!


I see, one more question how many GPUs did you use to find it?
[/quote]

Several thousand
[/quote]

Show us your video cards, several thousand, you are lying, you are from the team of creators who created and know the private key and you are just mocking us here, it is very funny!

[Bram24732]

Guys, stop trying to pressure the solver into doing anything. If you're nice, maybe he will stay and you can learn something from him.
We are all grateful for WP's post, sharing a method to securely withdraw a puzzle prize. Nobody is obligated to tip anyone just because you used MARA. You can always show your appreciation for what you've learned by directly sending a tip yourself.

It is not about a donation; that is up to each person if they feel like donating or not. It is more about valuing the contributions and not undermining them. He talks as if it were a 100% original milestone. It is obvious that his solution is just a method already exposed here, and its only advantage was great computing power. So, unless it is something else, the answer is simple: he has nothing to teach us that we do not already know. From what he says, random method with thousands of GPUs are things we already know. I suppose the most he could have added to the software is a database that records already scanned sectors to skip them later.

That's spot on. I did not invent anything because there is nothing to invent. None of the pattern theories I've read here are based on math or statistics, just wishful thinking.
The proofs I posted on Github are just a consequence of full random scan, they are not a method. They are the consequence of me scanning the whole range and are only useful to prove that I did so.

I did not come here to teach anyone anything, you have a few members like WP who know their stuff and do that very well already.

To the MARA WP donation crew : Stop thinking this forum is the only place in the world that knows about slipstream and it's some kind of secret weapon. The release of their service was heavily discussed in many news articles and podcasts back then. I don't ask you for a donation each time you make a post on facebook because I made a tutorial to do so on an obscure forum in 2005.

[kTimesG]

Let's recap the main methods:

1. Magic-based

- Adobe Illustrator number horoscopes, misc. font sizes, arrows everywhere
- magic number patterns in hexagons, squares, rectangles;
- adding "hex" numbers left and right, maybe mix in decimals as well, so that we can fit somehow solutions


2. Random methods

- sliders to choose hex digits, maybe we get lucky by clicking the mouse!
- Python scripts that sit idle in background (we all like quiet environment, right?)
- pick random ranges, or random keys - ignore completely that testing 2/3/4/5/6/...N times the same keys or ranges becomes inevitable really, really quick, why care?!

3. Probabilities

- backward statistics: if something is probable, and it happens, than it can still be analyzed as if it was just probable, right? Sure.
- uniform distribution: if something is totally random, then it matters what values it ended up with (and also when), and hence we can predict where & when it's not likely to be repeated
- the keys must not look "not random": very low chances that the solution is key 0x1, or key 32768, 16777216, and so on. Because too many zero or one bits in a row. Even better, do the same with base-10 digits, or upgrade to using hex char strings. Also, invent some alphabet where hex digits are case-sensitive, and let everyone else figure out WTF you smoked.

4. Mystical

- hashes are broken; NSA has the backdoor to SHA256
- there is a hidden relation between secp256k1 formula and RIPEMD160 bits: this makes all sorts of statistical theories to not apply
- neural networks can be trained to predict the keys, because they are really so intelligent that they can predict the future.

5. A shitload of computing power, scanning the ranges just once by splitting the work. Keep some proof of work for each range, like keys for a fixed number of leading zeros. Laugh thinking that some people may interpret this as proof of backward statistics (maybe combined with mystical key-to-hash relationship) as the main reason of why the correct key was eventually found after around 50% checked keys.

[frozenen]

Let's recap the main methods:

1. Magic-based

- Adobe Illustrator number horoscopes, misc. font sizes, arrows everywhere
- magic number patterns in hexagons, squares, rectangles;
- adding "hex" numbers left and right, maybe mix in decimals as well, so that we can fit somehow solutions


2. Random methods

- sliders to choose hex digits, maybe we get lucky by clicking the mouse!
- Python scripts that sit idle in background (we all like quiet environment, right?)
- pick random ranges, or random keys - ignore completely that testing 2/3/4/5/6/...N times the same keys or ranges becomes inevitable really, really quick, why care?!

3. Probabilities

- backward statistics: if something is probable, and it happens, than it can still be analyzed as if it was just probable, right? Sure.
- uniform distribution: if something is totally random, then it matters what values it ended up with (and also when), and hence we can predict where & when it's not likely to be repeated
- the keys must not look "not random": very low chances that the solution is key 0x1, or key 32768, 16777216, and so on. Because too many zero or one bits in a row. Even better, do the same with base-10 digits, or upgrade to using hex char strings. Also, invent some alphabet where hex digits are case-sensitive, and let everyone else figure out WTF you smoked.

4. Mystical

- hashes are broken; NSA has the backdoor to SHA256
- there is a hidden relation between secp256k1 formula and RIPEMD160 bits: this makes all sorts of statistical theories to not apply
- neural networks can be trained to predict the keys, because they are really so intelligent that they can predict the future.

5. A shitload of computing power, scanning the ranges just once by splitting the work. Keep some proof of work for each range, like keys for a fixed number of leading zeros. Laugh thinking that some people may interpret this as proof of backward statistics (maybe combined with mystical key-to-hash relationship) as the main reason of why the correct key was eventually found after around 50% checked keys.

Hey I like Magic-based methods , I still have the wheel of fortune!

[Niekko]

Let's recap the main methods:

1. Magic-based

- Adobe Illustrator number horoscopes, misc. font sizes, arrows everywhere
- magic number patterns in hexagons, squares, rectangles;
- adding "hex" numbers left and right, maybe mix in decimals as well, so that we can fit somehow solutions


2. Random methods

- sliders to choose hex digits, maybe we get lucky by clicking the mouse!
- Python scripts that sit idle in background (we all like quiet environment, right?)
- pick random ranges, or random keys - ignore completely that testing 2/3/4/5/6/...N times the same keys or ranges becomes inevitable really, really quick, why care?!

3. Probabilities

- backward statistics: if something is probable, and it happens, than it can still be analyzed as if it was just probable, right? Sure.
- uniform distribution: if something is totally random, then it matters what values it ended up with (and also when), and hence we can predict where & when it's not likely to be repeated
- the keys must not look "not random": very low chances that the solution is key 0x1, or key 32768, 16777216, and so on. Because too many zero or one bits in a row. Even better, do the same with base-10 digits, or upgrade to using hex char strings. Also, invent some alphabet where hex digits are case-sensitive, and let everyone else figure out WTF you smoked.

4. Mystical

- hashes are broken; NSA has the backdoor to SHA256
- there is a hidden relation between secp256k1 formula and RIPEMD160 bits: this makes all sorts of statistical theories to not apply
- neural networks can be trained to predict the keys, because they are really so intelligent that they can predict the future.

5. A shitload of computing power, scanning the ranges just once by splitting the work. Keep some proof of work for each range, like keys for a fixed number of leading zeros. Laugh thinking that some people may interpret this as proof of backward statistics (maybe combined with mystical key-to-hash relationship) as the main reason of why the correct key was eventually found after around 50% checked keys.

... you forgot the improbable mathematical solutions. 

[7xminer]

Let's recap the main methods:

1. Magic-based

- Adobe Illustrator number horoscopes, misc. font sizes, arrows everywhere
- magic number patterns in hexagons, squares, rectangles;
- adding "hex" numbers left and right, maybe mix in decimals as well, so that we can fit somehow solutions


2. Random methods

- sliders to choose hex digits, maybe we get lucky by clicking the mouse!
- Python scripts that sit idle in background (we all like quiet environment, right?)
- pick random ranges, or random keys - ignore completely that testing 2/3/4/5/6/...N times the same keys or ranges becomes inevitable really, really quick, why care?!

3. Probabilities

- backward statistics: if something is probable, and it happens, than it can still be analyzed as if it was just probable, right? Sure.
- uniform distribution: if something is totally random, then it matters what values it ended up with (and also when), and hence we can predict where & when it's not likely to be repeated
- the keys must not look "not random": very low chances that the solution is key 0x1, or key 32768, 16777216, and so on. Because too many zero or one bits in a row. Even better, do the same with base-10 digits, or upgrade to using hex char strings. Also, invent some alphabet where hex digits are case-sensitive, and let everyone else figure out WTF you smoked.

4. Mystical

- hashes are broken; NSA has the backdoor to SHA256
- there is a hidden relation between secp256k1 formula and RIPEMD160 bits: this makes all sorts of statistical theories to not apply
- neural networks can be trained to predict the keys, because they are really so intelligent that they can predict the future.

5. A shitload of computing power, scanning the ranges just once by splitting the work. Keep some proof of work for each range, like keys for a fixed number of leading zeros. Laugh thinking that some people may interpret this as proof of backward statistics (maybe combined with mystical key-to-hash relationship) as the main reason of why the correct key was eventually found after around 50% checked keys.

I liked the magic-method. LOL 

[zahid888]

puzzle: 66 ?? Possibilities : ??

puzzle: 67 ?? Possibilities : ??

In my previous two posts, a question mark turned into an answer. As I mentioned, there are 5 to 20 possible starting points, and the answer is...

puzzle: 66 2832ed74 Possibilities : 10         puzzle: 67 730fc235 Possibilities : 13

   Seed : 1198740450 KHex : 2832ed74            Seed : 2113081982 KHex : 730fc235
   Seed : 1869248704 KHex : 2832ed74            Seed : 2179848786 KHex : 730fc235
   Seed : 3920977047 KHex : 2832ed74            Seed : 2620256395 KHex : 730fc235
   Seed : 4138865358 KHex : 2832ed74            Seed : 3559516538 KHex : 730fc235
   Seed : 5491054614 KHex : 2832ed74            Seed : 5559894373 KHex : 730fc235
   Seed : 5889540521 KHex : 2832ed74            Seed : 5960477113 KHex : 730fc235
   Seed : 6227837743 KHex : 2832ed74            Seed : 6204436682 KHex : 730fc235
   Seed : 7311367263 KHex : 2832ed74            Seed : 7016671995 KHex : 730fc235
   Seed : 7442940504 KHex : 2832ed74            Seed : 8305603871 KHex : 730fc235
   Seed : 8850149776 KHex : 2832ed74            Seed : 8560029709 KHex : 730fc235
                                                Seed : 8633074902 KHex : 730fc235
                                                Seed : 9737552820 KHex : 730fc235
                                                Seed : 9997208084 KHex : 730fc235

Higher Probability of Success – Brute force requires exact matching, while this method increases chances by working with 5 to 20 potential seeds rather than the entire massive keyspace. still applies on Puzzle 68

@Bram24732...... Belated congratulations!

[dastic]

puzzle: 66 ?? Possibilities : ??

puzzle: 67 ?? Possibilities : ??

In my previous two posts, a question mark turned into an answer. As I mentioned, there are 5 to 20 possible starting points, and the answer is...

puzzle: 66 2832ed74 Possibilities : 10         puzzle: 67 730fc235 Possibilities : 13

   Seed : 1198740450 KHex : 2832ed74            Seed : 2113081982 KHex : 730fc235
   Seed : 1869248704 KHex : 2832ed74            Seed : 2179848786 KHex : 730fc235
   Seed : 3920977047 KHex : 2832ed74            Seed : 2620256395 KHex : 730fc235
   Seed : 4138865358 KHex : 2832ed74            Seed : 3559516538 KHex : 730fc235
   Seed : 5491054614 KHex : 2832ed74            Seed : 5559894373 KHex : 730fc235
   Seed : 5889540521 KHex : 2832ed74            Seed : 5960477113 KHex : 730fc235
   Seed : 6227837743 KHex : 2832ed74            Seed : 6204436682 KHex : 730fc235
   Seed : 7311367263 KHex : 2832ed74            Seed : 7016671995 KHex : 730fc235
   Seed : 7442940504 KHex : 2832ed74            Seed : 8305603871 KHex : 730fc235
   Seed : 8850149776 KHex : 2832ed74            Seed : 8560029709 KHex : 730fc235
                                                Seed : 8633074902 KHex : 730fc235
                                                Seed : 9737552820 KHex : 730fc235
                                                Seed : 9997208084 KHex : 730fc235

Higher Probability of Success – Brute force requires exact matching, while this method increases chances by working with 5 to 20 potential seeds rather than the entire massive keyspace. still applies on Puzzle 68

@Bram24732...... Belated congratulations!

But how did you get the seeds if you don't know the starting hex?

[nomachine]

A shitload of computing power

This is where the fun begins and ends for fools who are full of money.



Spending a fortune on a problem without a second thought, completely indifferent to whether it all goes up in smoke or not..It's hard to understand how anyone could have turned a profit from it.

Was it all just for fun?

The British band The KLF burned one million pounds sterling in cash as part of an art performance.

Maybe this is part of some artistic performance.

[Akito S. M. Hosana]

Maybe this is part of some artistic performance.

It reminds me more of a circus. 

[0x1FFFFFF]

puzzle: 66 ?? Possibilities : ??

puzzle: 67 ?? Possibilities : ??

In my previous two posts, a question mark turned into an answer. As I mentioned, there are 5 to 20 possible starting points, and the answer is...

puzzle: 66 2832ed74 Possibilities : 10         puzzle: 67 730fc235 Possibilities : 13

   Seed : 1198740450 KHex : 2832ed74            Seed : 2113081982 KHex : 730fc235
   Seed : 1869248704 KHex : 2832ed74            Seed : 2179848786 KHex : 730fc235
   Seed : 3920977047 KHex : 2832ed74            Seed : 2620256395 KHex : 730fc235
   Seed : 4138865358 KHex : 2832ed74            Seed : 3559516538 KHex : 730fc235
   Seed : 5491054614 KHex : 2832ed74            Seed : 5559894373 KHex : 730fc235
   Seed : 5889540521 KHex : 2832ed74            Seed : 5960477113 KHex : 730fc235
   Seed : 6227837743 KHex : 2832ed74            Seed : 6204436682 KHex : 730fc235
   Seed : 7311367263 KHex : 2832ed74            Seed : 7016671995 KHex : 730fc235
   Seed : 7442940504 KHex : 2832ed74            Seed : 8305603871 KHex : 730fc235
   Seed : 8850149776 KHex : 2832ed74            Seed : 8560029709 KHex : 730fc235
                                                Seed : 8633074902 KHex : 730fc235
                                                Seed : 9737552820 KHex : 730fc235
                                                Seed : 9997208084 KHex : 730fc235

Hi @zahid888. I am curious about this from last year.. I even write some python script to try to understand this, but still didnt get that :

Code:

import hashlib
from ecdsa import SigningKey, SECP256k1
import binascii
import base58
import random
import time

def set_random_seed(manual_seed=None):
    if manual_seed is not None:
        random.seed(manual_seed)
    else:
        # Auto seed between 1000000000 to 9999999999
        random_seed = random.randint(1000000000, 9999999999)
        random.seed(random_seed)
        print(f"Auto Generated Seed: {random_seed}")

def get_random_private_key(start_range, end_range):
    return random.randint(start_range, end_range)

def private_key_to_public_key(private_key_hex):
    signing_key = SigningKey.from_string(bytes.fromhex(private_key_hex), curve=SECP256k1)
    verifying_key = signing_key.get_verifying_key()
    pub_key = verifying_key.to_string("compressed")
    return pub_key

def public_key_to_address(public_key):
    sha256_hash = hashlib.sha256(public_key).digest()
    ripemd160_hash = hashlib.new('ripemd160')
    ripemd160_hash.update(sha256_hash)
    hash160 = ripemd160_hash.digest()
    version_hash160 = b'\x00' + hash160
    double_sha256 = hashlib.sha256(hashlib.sha256(version_hash160).digest()).digest()
    binary_address = version_hash160 + double_sha256[:4]
    address = base58.b58encode(binary_address).decode('utf-8')
    return address

def generate_address_from_private_key(private_key_int):
    private_key_hex = format(private_key_int, '064x')
    public_key = private_key_to_public_key(private_key_hex)
    address = public_key_to_address(public_key)
    return {
        'private_key': private_key_hex,
        'address': address
    }

def generate_from_bit_range(bit_number):
    start_range = 2 ** (bit_number - 1)
    end_range = (2 ** bit_number) - 1
    private_key_int = get_random_private_key(start_range, end_range)
    result = generate_address_from_private_key(private_key_int)
    
    print(f"\nBit {bit_number}")
    print(f"Range: {hex(start_range)[2:]}:{hex(end_range)[2:]}")
    print(f"Address: {result['address']}")
    print(f"Private Key: {result['private_key']}")
    print("-" * 80)
    return result

def main():
    while True:
        print("\nSelect seed option:")
        print("\n1. Manual seed (input your own number)")
        print("2. Auto seed (random between 1000000000 to 9999999999)")
        print("3. Exit")
        
        choice = input("\nEnter your choice (1-3): ")
        
        if choice == "1":
            try:
                seed = int(input("Enter your seed number: "))
                set_random_seed(seed)
            except ValueError:
                print("Invalid input. Please enter a valid number.")
                continue
        elif choice == "2":
            set_random_seed()
        elif choice == "3":
            print("Exiting program...")
            break
        else:
            print("Invalid choice. Please select 1, 2, or 3.")
            continue
        
        print("\n=== Random Generation from Bit Ranges ===")
        for bit in range(1, 70):
            generate_from_bit_range(bit)

if __name__ == "__main__":
    main()

[Kelvin555]

If you do the calculation, puzzle 67 solver only realized a lot of losses,
67 days x 24 hours x several thousand GPUs = more than puzzle 67 price. I think that's why he is not giving any donations right now.

[Desyationer]

The profitability of Puzzle 67 can be assessed using AI, and it turns out to be quite profitable, maybe even 68.

At a speed of 7 Terrakeys, approximately 1,750 RTX 4090 GPUs with undervolting to 350 W would be required.

🔹 Power Consumption:

Total power: 612.5 kW
Runtime: 67 days (1,608 hours)
Total consumption: 984,900 kWh
🔹 Electricity Costs:

USA (0.19 USD/kWh): 187,131 USD
UK (0.318 USD/kWh): 313,198 USD (≈244,686 GBP)
These costs vary by region and tariff, but the USA and the UK remain among the most expensive places for electricity.

[Kelvin555]

The profitability of Puzzle 67 can be assessed using AI, and it turns out to be quite profitable, maybe even 68.

At a speed of 7 Terrakeys, approximately 1,750 RTX 4090 GPUs with undervolting to 350 W would be required.

🔹 Power Consumption:

Total power: 612.5 kW
Runtime: 67 days (1,608 hours)
Total consumption: 984,900 kWh
🔹 Electricity Costs:

USA (0.19 USD/kWh): 187,131 USD
UK (0.318 USD/kWh): 313,198 USD (≈244,686 GBP)
These costs vary by region and tariff, but the USA and the UK remain among the most expensive places for electricity.

1,750 RTX 4090s rented for 67 days or bought is more than puzzle 67 price.

[Desyationer]

If the GPUs are free for them—let's say it's some AI organization whose computing power the solver has access to—then paid rental is madness.