Posts: 3
Threads: 1
Joined: Apr 2017
Hi guys,
I was wondering if anyone has tried any of the cloud GPU systems by either google, amazon or softlayer?
With softlayer for example you get 24gb of DDR5 memory, 4992 cuda cores and 480gb/sec of bandwidth through put for $5.30 per hour.
Has anyone tried any of these and if so what kind of performance did you get?
https://postimg.org/image/hrqdb3chf/
Posts: 2,267
Threads: 16
Joined: Feb 2013
Teslas are slower than Maxwell cards when it comes to hash cracking.
Maybe you can find a "cloud provider" with 970, 980, 980ti, 1080, 1080ti etc; probably not, because they do not focus on hash cracking but on other types of computations (which include/focus on floating point instructions etc).
They might not be (too) slow (e.g. compared to even older GPUs or CPUs etc), but as said they could be much slower than a very cheap consumer GPU that you get for around 150$.
Posts: 3
Threads: 1
Joined: Apr 2017
I'm not fully knowledgeable but doesn't hashcat work best with cuda cores? does the memory or bandwidth have much to do with the performance?
Posts: 2,267
Threads: 16
Joined: Feb 2013
Of course more cores means more performance (in general), but here we are talking about some very different architectures and instruction sets. That is what makes the real difference for hash cracking.
Maxwell is currently much better when it comes to hash cracking than Kepler (Teslas etc)
Posts: 3
Threads: 1
Joined: Apr 2017
Ah okay it seems that all the cloud companies use teslas well all the ones i've looked at so far have them as their only option.
If only google could setup The-Distribution-Which-Does-Not-Handle-OpenCL-Well (Kali) linux for me and install some really powerful maxwells
