Hacker News Reader

Comment by jokoon | Hacker News Reader

So can a large FPGA be somehow used to brute force encryption?

I don't really understand electronics to see if a GPU could be faster than a FPGA, but my guess is yes?

It seems that anything that can be programmed is inherently slower than a FPGA equivalent doing the same task.

Does larger enough key size always defeat a FPGA?

I would guess that it becomes power and cost prohibitive for a private company to deliver such possibility, but of course, a large government entity like the NSA might have enough resource to pay for enough FPGA to decrypt most things.

View on HN

> So can a large FPGA be somehow used to brute force encryption?

Yes and no. It is usually possible to use an FPGA to build a specific logic circuit that will run some particular algorithm faster than the algorithm would run on a general-purpose CPU. (By a factor of maybe 2x to 100x. It varies a lot depending on the algorithm.)

It is also the case that if you implement that algorithm on a dedicated custom chip it can always be designed to run faster than the equivalent FPGA.

So if you really want maximum speed, spend the megabucks for dedicated silicon.

That said, modern GPUs and DSP chips might be able to run some algorithms almost as fast as dedicated silicon and faster than FPGAs. So maybe you don't need megabucks. Again, it depends on the algorithm.

In any case you'd need many orders of magnitude of speedup to usefully brute-force a modern crypto algorithm, and the most you'd get even from dedicated silicon is about three. Useful speedup requires quantum computers -- which don't yet exist -- and even quantum computers will only help break asymmetric crypto, not symmetric crypto.

for AES, for example?

AES is symmetric. Asymmetric algorithms are those that have pairs of private and public keys, like RSA.

Even though the FPGA fabric might encode the solution more effectively, there are other important differentiators: clock speed and memory bandwidth. GPUs have higher clock speeds and typically better memory bandwidth (related of course).

With the higher clock speed, GPUs can well outperform FPGAs for many problems.

Reply on news.ycombinator.com