On Jan 15, 2010, at 3:01 AM, Gregory Maxwell wrote: > On Thu, Jan 14, 2010 at 7:59 PM, Karsten Nohl <[email protected]> > wrote: > [snip] >> Number of tables * tables height = 2 TB > > The actual raw capacity of a random 32gb SD card here is 32017047552 > bytes (29.8181 GiB). > I wouldn't bother mentioning it, but the difference appears to be an > increase in attack time by 15% in the one GPU case.
Good point; and one dimension in which the Excel model could be further detailed. However, since the points are much less than 128bits each after accounting for the DP bits at the end and an additional optimization we are planing, this 15% hit is more than compensated for. I'll introduce the unit storage size as an additional factor once we have clarity on the exact number for our next table generation. > I also thought it was interesting that adding in 30ms of hdd latency > latency results in rather long solution times. The internet > distributed lookup doesn't sound especially viable if its governed by > the same formula as the disk based calculations. OTOH, the 10ms > lookup of a cheap spinning disk isn't too terrible. What makes USB sticks so much faster than hard disks is their smaller granularity. 32x 32GB USB stick at 2ms access time gives you 160 times as many lookups as 1TB HDD at 10ms. _______________________________________________ A51 mailing list [email protected] http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
