Benchmark data request
#1
Lightbulb 
Given there's a possiblity of confusing, could the tester provide like a SKU or other unique product identifier besides the chipset ?

I have not seen such thus far.
Reply
#2
The GPU chip itself is what matters for cracking performance, and all cards from the same model family use literally the same GPU chips. We also strongly urge users to only purchase reference design cards, which are identical regardless of manufacturer. Where OEM cards differ is in terms of build quality, level of R&D, and cooling solution. None of which impacts benchmark speeds.
Reply
#3
A few questions arose.

... what on earth are reference design cards, how do i identify them ?
... i assumed buying a motherboard + videocard from the same vendor would be most optimal ?

I'm aware the chipset is what makes a GPU tick. What made me drop the initial post was i saw mention of AMD R9 290X but there was no mention of the amount of RAM installed, if the system and/or card had been overclocked etc.

Right now the benchmark content is littered across the forums i pressume ? Or is there like a central repository which may provide an interface for comparison.

... in addition, given the use of AVX, SSE and other CPU extensions and for example the i7-5820K offering a huge amount of L3 cache, would there be any decent performance to obtain for hash cracking ?

Thanks
Reply
#4
... what on earth are reference design cards, how do i identify them ?
... i assumed buying a motherboard + videocard from the same vendor would be most optimal ?

With one word, no. As of right now the quality of motherboards and gpus have nothing in common. At the moment Asus builds arguably the best motherboards but beware of their "tripple-dipple-super-edition" GPUs.
What epix is trying to say is that for a professional cracking rig we strongly advise buying cards that have not been tinkered with by oems. If you have problems to spot the difference, usually vendors first release reference cards and follow up with their own altered models 1-3 month after launch.

Don't try to mix CPU and GPU. There are clear lines for which both hardware performs best. For example you CPUs clearly are more efficient at bcrypt where as GPUs clearly outperform at CPUs in MD5 and other non-salted algos.

Do you want the most bang for the buck or just the most performance? If latter go for double xeon 2660, if first check out the AMD FX 8k series but your i7 is pretty good middleground.
sch0.org
Reply
#5
(03-24-2015, 11:30 AM)Jarth Wrote: ... what on earth are reference design cards, how do i identify them ?

You can read more about reference design cards here: http://hashcat.net/forum/thread-3949-pos...l#pid22844


(03-24-2015, 11:30 AM)Jarth Wrote: ... i assumed buying a motherboard + videocard from the same vendor would be most optimal ?

No, it literally does not matter one bit.


(03-24-2015, 11:30 AM)Jarth Wrote: I'm aware the chipset is what makes a GPU tick. What made me drop the initial post was i saw mention of AMD R9 290X but there was no mention of the amount of RAM installed, if the system and/or card had been overclocked etc.

That's because motherboard, RAM, and CPU overclocking are all largely irrelevant, or at most inconsequental, for GPU cracking performance.

Benchmarks should mention if the GPU is overclocked, however. If the post does not state, you can safely assume stock clocks were used.


(03-24-2015, 11:30 AM)Jarth Wrote: Right now the benchmark content is littered across the forums i pressume ? Or is there like a central repository which may provide an interface for comparison.

Correct, it's not well-organized. You can use the search feature on the forums.


(03-24-2015, 11:30 AM)Jarth Wrote: ... in addition, given the use of AVX, SSE and other CPU extensions and for example the i7-5820K offering a huge amount of L3 cache, would there be any decent performance to obtain for hash cracking ?

i7-5820K has support for AVX2, which hashcat-cliAVX2 is able to take advantage of for some algorithms. But this only matters for CPU cracking, not GPU cracking. The amount of L3 cache isn't really that important for CPU cracking. Most algorithms are register-based, or stay in L1 cache.
Reply
#6
Thanks. It is clear i still have much to learn about cracking, if i'm able to catch up at all that is. I assumed pre-loading hash lists into Graphics card RAM would greatly enhance any cracking involved.
Reply