Quantcast
Channel: hashcat Forum - All Forums
Viewing all articles
Browse latest Browse all 7673

Scrypt and GPUs

$
0
0
Here is a quote from github on Scypt's massive memory requirements.  By Jsteube.

  • "GPU's will not be able to crack this because of the memory requirement. Math behind: Mem per candidate: size_scrypt = (128 * scrypt_r) * scrypt_N = 256MB. But to make use of the GPU power, we need to spawn at least 1280 parallel computations, so we end up with a memory requirement of 320GB ram per GPU. It would require a minimum of a 256 divisions TMTO to get it working on a GTX1080, which drops the performance < 1 H/s."
Since Scrypt has such a huge memory requirement to make it GPUable, would it not be possible to provide the GPU's with "unlimited" memory through the use of CUDA unified memory features? This feature was introduced in CUDA 4 in 2014 and provides a single memory space directly accessible by all GPUs and CPUs in the system, with automatic page migration for data locality. 

Viewing all articles
Browse latest Browse all 7673

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>