You will never have a commercially made ASIC miner for SHA encryption!
Until every organisation on the internet and off the internet changes the way they encrypt their information, you are out of luck.
Reason being:
This ASIC hardware could too easily be reverse engineered to decrypt SHA. And with it's small size and power consumption they could be ran almost anywhere barely undetected constantly attacking SHA at an incredible rate. Essentially these devices would put too much power into public hands.
If you are going to continue to pursue this Bitcoin mining game you better load up on FPGAs and GPUs!
Now someone tell me I am wrong.
Thanks.
Let's put bitcoin aside, and focus completely on security here, since you've made that a serious concern. Here's the thing with security, it's exponentially more difficult to break, for every small difficulty increase in encrypting. This means it'll always be much easier to increase difficulty, than it will to break the harder difficulty. As technology advances and we get fast machines to break 256 bit encryption, 512 comes out, and after some time, we can then break 512, but it's already been upgraded to 1024, and so on.
Let's look at some numbers just to check this out:
http://www.cryptopp.com/benchmarks.htmlUsing the first algorithm type, let's assume our arbitrary encryption uses 17.2 cycles per byte of the key, for some arbitrary length data (making a lot of assumptions, but they're made fairly).
So, our 8-bit cypher takes 17.2 CPU cycles to encrypt. With 8-bit security, there are 256 possible outcomes (2^8). All outcomes can be found in 4403.2 cycles (256*17.2). I'm going to go crazy here and assume absolutely no collisions for these examples. In reality, however, you may need to encrypt/hash 500+ different items to cover all 256 possibilities, or you may never find all 256, but for these examples, we'll assume no collisions, and everything plays nice.
Now, let's assume we get some fancy new computers available, and now we need to upgrade our security. Let's use 16-bit encryption. That'll take 34.4 cycles to encrypt (2*17.2), giving us 65536 combinations possible. That'll take 1.127 million cycles to calculate all possible values. With a 3GHz processor (3 billion cycles per second), you're still talking about 1/1000 of a second to calculate all possible values.
Let's up our security to 24-bit. Encryption = 51.6 cycles. 16.8m possibilities. 288m cycles for all possibilities (1/30th of a second).
32-bit security. Encrypt = 68.8 cycles. 4.3 billion possibilities. 73.9 billion cycles for all combinations. 24.62 seconds.
40-bit security. Encrypt = 86 cycles. 1.1 trillion possibilities. 18.9 trillion cycles for all combinations. 1 hour, 45 minutes.
48-bit security. Encrypt = 103.2 cycles. 281 trillion possibilities. 4.84 quadrillion cycles for all possibilities. 18 days, 16 hours.
So you see, while it take so incredibly little time to encrypt, the decryption time grows exponentially larger with each little bit of security you add. So, would it be possible for someone to create some ASICs that are designed to break security in use today? Sure, but as soon as that was figured out, everyone would just increase what security strength they use, and all the ASICs become completely useless then. It would be interesting if someone started working on ASICs for 2048 bit encryption of some popular types, then waited for everyone to upgrade (creating rainbow tables or whatever in the meantime), then they could attack several companies at a single time before everyone noes up to 4096. It'd be a bit surprising to jump up in security, only to find it less secure than what everyone was just on.
