FPGA is that any good?
#1
hi,
i am taking an old subject back alive. (hope for it at least)
since hashcat 3.0v started to support fpga i wanted to know more about this cards and his performance.
i've read most of the articles about it on the forum, from my understanding back in 2012+- some one asked about it here and here and here Smile , the conclusion in short, atom & epixoip answers was the fpga are good against single algorithim like DES and Bitcoin, bruteforce a single hash and electricity worth. multiple attack modes, multiple hashes, or multiple algorithms and the card was unflexible..
now the question.

why does it fall on multiple hashes,attack,algorithim? (please mention all the detailed.. like really everything)
can u mention other single algorithim beside DES and bitcoin?
what is the different between a single algorithim and others?
why it is not flexibility? what make it so unflexible? what u guys ment by saying it?

xilinx sales fpga and i think the latest of their product is 
http://www.xilinx.com/products/silicon-d...-plus.html
and
http://www.xilinx.com/products/silicon-d...scale.html

i have a benchmark of their LX150 spartan 6 and the result are quite low in some algorithim compare  to gpu. but the spartan is for lowcost and old.

thanks.
#2
(07-06-2016, 01:35 PM)kiara Wrote: since hashcat 3.0v started to support fpga

hashcat 3.00 did not "add support for FPGA"; it added support for any arbitrary OpenCL device on any OpenCL platform. This opens up the possibility of using devices like FPGAs and DSPs with hashcat; however, this is heavily contingent upon the availability, quality, and maturity of an OpenCL SDK for these devices. Right now, there are no competent OpenCL SDKs or runtime libraries available for FPGAs. Altera's OpenCL is garbage, and Xilinx's is alpha at best. 

(07-06-2016, 01:35 PM)kiara Wrote: why does it fall on multiple hashes,attack,algorithim?

Because RAM.
#3
thanks for answering.

can u mention other single algorithim beside DES and bitcoin?
what is the different between a single algorithim and others?
why it is not flexibility? what make it so unflexible? what u guys ment by saying it?
#4
We don't have enough experience with it. Support for all OpenCL devices types just started with hashcat v3.00. Also we don't have any hardware to test on.
#5
so how come u guys said it was unflexible etc.. (on the other posts)
from where the info ?
#6
From what I know about FPGAs, which is really very little, they're somewhere between a CPU (or similar, which seems to include those things in GPUs) and ASICs. CPUs can do anything, they're general-purpose (GPU bits are a bit more specific I think), ASICs are hardware that can do just one thing, really well. FPGAs, as the name suggests, can be sort of reconfigured, but not quickly or easily (I'm guessing here, the Wikipedia page didn't help me much).

So you take your FPGA and mess around with it and it can do a hash - problem is, it's then very good at doing a certain type of calculation only, which means always the same hash, always the same mode. Doing a different type means reconfiguring the thing, which is hard. 

I'm pretty sure such a thing would be far more expensive than a GPU for equivalent performace for at least a number of years, and even then you'd have to know a lot about the hash algorithms you want and how to implement them on the hardware. 

If any of this is wrong, go ahead and shout me down - it's a (barely) educated guess.
#7
Hey guys, late response I know, I was just curious as to whether or not anyone has any experience with OpenCiphers and whether or not these drivers make it possible to use FPGA's for password cracking (WPA/WPA2 in particular)? I'm new hear so forgive me if I've made any mistakes with my post.