These numbers are just crazy. Just owning these cards is very expensive but amazing results. If you are free for CMIYC 2013 we could make you a spot in Insidepro team...
------>>>>>>A place for your advert here<<<<<<------
were you at the cmiyc booth? i think i met you
yes, i was there briefly to socialize
(12-04-2012, 06:52 PM)epixoip Wrote: [ -> ]For those interested that either aren't at Passwords^12 or do not use Twitter, I've posted my Passwords^12 presentation about VCL clustering with Hashcat at https://bindshell.nl/pub/Password_Cracking_HPC.pdf
The presentation looks great. Wish I could I actually see and hear it instead of just reading over it though! Thanks for sharing. Specifically, thanks for sharing your bandwidth measurements and benchmarks.
just got down watching your presentation. good job! looks like SuperCL will eliminate the need for infiniband and we can all use GigE again? if that's true, when will Atom add support for SuperCL?
no, supercl just shifts where you need the high bandwidth/low latency link from between your broker and compute nodes, to between your compute nodes and your storage.
why do you want to eliminate infiniband? it's not a proper cluster unless infiniband is involved.
(12-02-2012, 06:54 PM)epixoip Wrote: [ -> ]10x 7970, 4x 5970, 3x 6990, and a 5870. 4x SDR Infiniband interconnect.
I've been trying multiple 7970 my self with various success. The current setup I've got is running two of them in the same chassis due to heat problems. Would you mind sharing your hardware setup?
/blh