For those of you interested in the performance increase that CUDA is able to provide, I ran some benchmarks to demonstrate. This is with my system at the speeds in my sig.... 3.4ghz on the cpu and (725 / 2800) on my gpu/mem
root@bt:~# pyrit benchmark
Pyrit 0.2.5-dev (C) 2008, 2009 Lukas Lueg http://pyrit.googlecode.com
This code is distributed under the GNU General Public License v3
Connecting to storage... connected
Running benchmark for about 1 seconds... -
Computed 17175.52 PMKs/s total.
#1: 'CUDA-Device #1 'GeForce GTX 285'': 15069.5 PMKs/s (Occ. 99.6%; RTT 2.6)
#2: 'CPU-Core (SSE2)': 671.3 PMKs/s (Occ. 98.1%; RTT 3.2)
#3: 'CPU-Core (SSE2)': 770.2 PMKs/s (Occ. 99.0%; RTT 2.8)
#4: 'CPU-Core (SSE2)': 753.0 PMKs/s (Occ. 98.5%; RTT 3.2)
For those of you wondering too why there's only 3 cpu cores listed. This program will dedicate a cpu core per gpu installed to 'feed' it. A reasonable trade-off I'd say.
So as you can see, it's well over a 15x improvement per gpu vs cpu core. The scaling with CUDA is amazing and really over-looked by most graphics enthusiasts... granted I do agree, video performance should come first but the other uses for these cards are amazing.
It'd be pretty reasonable to assume that GPGPU computing would find its way into drive encryption, and compression programs (winzip, winrar, etc...) I would really love to see CUDA take off into more arenas than it currently occupies