you need a memory size of 2.4 * 10 ^ 25 bytes. And even applying algorithms that use a trade-off between memory and calculation time - it will still be a huge size.
No it won't. Collisions can be found an an effectively storageless manner with a small constant factor slowdown, I gave google terms upthread.
You're suffering from the same ignorance that caused the collision design flaw in "xthin" where they were claiming that collisions of a 64-bit hash were infeasible to compute due to storage requirements. (which I eventually eventually grew tired of correcting and started responding to all the messages with 64-bit sha2 collisions.)
You want to say that performing a hash according to the scheme:
Hash1 = RIPEMD160(SHA256(Data1))
Hash2 = RIPEMD160(SHA256(Data2))
You can find such Data1 and Data2, in which the first 64 bits in hash1 and hash2 will be the same without using the birthdays attack and bruteforce? And you will not spend much memory?
Can I see the algorithm (GitHub) and proof of work? I want to try it myself. How much time will it take to calculate this?