cheap hardware for 200.0 kH/s WPA2 with single GPU
#1
hashcats

we have a super low budget, like 500 eur for a complete server

we aim for a speed around ....

Hashmode: 2500 - WPA/WPA2 ----> Speed.Dev.#1.....:   200.0 kH/s
Hashmode: 1700 - SHA-512 ----> Speed.Dev.#1.....:   500.0 MH/s

we consider to use one of these GPUs ....

wpa2 eur gpu
285000 300 GeForce GTX 1070
224000 160 Radeon RX 580
205800 200 GeForce GTX 1060 6GB, 1536/6144 MB
185000 250 Radeon RX 480

here is a gaming laptop for 1000 eur
MSI GL63 8RE-811 (0016P5-811)
GPU = GeForce GTX 1060 6GB, 1536/6144 MB

can we go cheaper than that? (1000 eur)
we only need a hashcat server,
so we dont need the display, for example.

what would be a minimal hardware setup .... ?
to run one of the above GPUs
and to avoid most bottlenecks

we guess a raspberry pi 3 is too weak .... why?
what are the limiting factors here?

greetings from a poor mans cave
(yes we have electricity)

edit, here is our current wishlist
GPU
190eur MSI GeForce GTX 1060 6GT OCV1, 6GB GDDR5, DVI, HDMI, DP (V809-2234R), 150Watt peak power consumption
board + cpu + ram with Intel Socket 1151
055eur ASRock H110M-G/M.2 (90-MXB5H0-A0UAYZ), MicroATX, 3xPCIe
055eur Intel Celeron G3930, 2x 2.90GHz, boxed (BX80677G3930)
037eur Crucial DIMM 8GB, DDR4-2400, CL17 (CT8G4DFS824A)
case
026eur Inter-Tech IT-5908 (88881237), ATX, 2.3kg, 34Liter
psu
080eur be quiet! Pure Power 10 700W ATX 2.4 (BN275), 30dB
= 443eur
makes sense?
Reply
#2
A hashcat brain can run on any simple computer, doesn't even need to have a GPU.
Reply
#3
I am following this post with interest. None of my many computers will run the latest version of Hashcat. I am hoping that I can stick a later model GPU into one and solve the problem in that way. Waiting for knowledgeable replies ...
Reply
#4
I have a 1060 3GB, and definitely I would like to move to a 1070 or better.. So my advice is to buy maybe like a 1070 or 1070ti second hand, as hash/watt would be more beneficial than the 1060.
Keep in mind that applying rules could make the hash rate drop a lot.
And buy FOUNDER/REFERENCE design. I made the mistake of buying a dual fan model and I`ll try to solve it as soon as possible.

That´s a chart I ve done for the propose of comparing different models, assuming buying used stuff.



cheers.
Reply
#5
> A hashcat brain can run on any simple computer

maybe, but we want speed around 200 kH/s for WPA2
edit, aah, you mean GPU is the largest bottleneck .... sounds good



> I have a 1060 3GB

please post your benchmarks, we can only find marks for the 6GB models



> my advice is to buy maybe like a 1070 or 1070ti second hand

GTX 1070 Ti .... used for 260 eur .... 370 kHps per 170 watt .... not bad



> Keep in mind that applying rules could make the hash rate drop a lot.

we only need hybrid mode, with a random dictionary and incremental bruteforce.
this way, we expect similar speeds as in benchmarks.



> And buy FOUNDER/REFERENCE design.
> I made the mistake of buying a dual fan model and I`ll try to solve it as soon as possible.

why founder edition? why choose nVidia as card manufacturer?
why not dual fan?


as we understand, manufacturers only variate details,
like power source (PCIe or external 8-pin / 6-pin)
or cooling system (open air, blower, water cooling)
or card length
or RAM size
or warranty terms
or electric component quality
or frequency / voltage defaults ....

[ we plan to use "power limiting" or undervolting, to reduce hardware stress ]



we also found an "alternative mainboard" in this thread: Laptop choices

they use an "ExpressCard to PCIe adapter" to connect laptop and GPU.
also called eGPU or "video card dock".

forumhero Wrote:#7
03-11-2015

About a year ago I explored the possibility of a DIY external GPU setup. I had an ATI HD 5770, a 1000 watt PSU and a 7 year old Dell laptop all collecting dust and thought it would be a fun little project.

Long story short it was extremely easy to setup and performance was the same as having the card in a desktop. All I needed was a ExpressCard Bus to PCIe-x1 cable (part# PE4L-EC060A), install the GPU drivers (my case the AMD drivers) and had hashcat up and running in no time.

I did have reliability issues where the driver would crash on certain hash types forcing a restart of the laptop. Will have to fire it up again to see if reliability has improved with newer drivers and hashcat but for the most part it worked really good.

seems we can also connect multiple video cards to one ExpressCard slot
using a "PCIe port multiplier", either integrated into the EC--PCIe adapter board,
or modular, plugged into the adapter.

bottleneck?
+ PCIe 3.0 1x < 1 GByte/s
+ ExpressCard < 0.3 GByte/s
= key size < 0.3 GByte/s / 400 kH/s = 750 Byte
--> in theory, no bottleneck



since we already have a laptop with expresscard slot,
that would reduce our wishlist to ....



target = 200 kH/s WPA2
150eur GeForce GTX 1060 6GB, used --> 120 watt TDP, 140 watt peak
040eur ExpressCard to PCIe adapter
040eur be quiet! Pure Power 11 300W ATX 2.4 (BN290)
= 230eur total
= 1.15 eur/kHps + 0.6 watt/kHps
= 870 kHps/eur + 1670 kHps/watt



target = 370 kH/s WPA2
260eur GeForce GTX 1070 Ti 8GB, used --> 180 watt TDP, 290 watt peak
040eur ExpressCard to PCIe adapter
050eur be quiet! Pure Power 11 400W ATX 2.4 (BN292)
= 350eur total
= 0.95 eur/kHps + 0.5 watt/kHps
= 1060 kHps/eur + 2050 kHps/watt



.... to confirm your estimate that the GTX 1070 Ti is cheaper in the long run
= after 9 months of use at 0.30eur/kWh

the old question, pay now or pay later?

but also, pay how much power?
120 watt --> 25 eur per month
180 watt --> 40 eur per month



numbers porn!
Reply
#6
Have you considered the newer GTX1660? It should deliver ~280kh/s@120W TDP. Prices start at 200€. Kind of in the middle between the other two.

In long term running costs consider also the rest of the hardware, which shifts the calculation towards higher performing cards.
Reply
#7
Have anyone seen real 1660 cards benchmarks?
I'm wondering if they will outperform the 1070.
Reply
#8
The GTX1660Ti and the 1070 should be on the same level. The non-Ti 1660 is performing slower. I only estimated the ~280kh/s in WPA2 based on other "Turing"-Benchmarks, which is usually still pretty accurate.
Reply