Mac M1 - Bcrypt speed
#1
Long time listener first time caller. So I've been playing around with 6.1.1. on M1 and want to brain storm if anyone has any ideas.

I've been able to top out at 250MH/s after getting to this command:

hashcat -m 3200 -a 3 -w 4 -i --increment-min 6 /Users/blahblah/crack.txt -a??a?a?a?a?a?a?a

Before I was sitting at around 50-80 h/s with this:

hashcat -m 3200 -a 3  -i --increment-min 6 /Users/blahblah/crack.txt -a??a?a?a?a?a?a?a

Also touched around 70-80 h/s using a file word list:

hashcat -m 3200 -a 0   /Users/blahblah/crack.txt /Users/blahblah/weakpass


Two things:
1)Anyone know a path to reach faster? My CPU usage and device memory wasn't effected at all, actually was better when using -w 4 (83% idle available)
2)Hopefully anyone else using a M1 searching for how to optimize comes through this post. -w 4 does wonders...


Terminal readout:
OpenCL API (OpenCL 1.2 (Oct 29 2020 19:50:08)) - Platform #1 [Apple]
====================================================================
* Device #1: Apple M1, 10858/10922 MB (1024 MB allocatable), 8MCU

Minimum password length supported by kernel: 0
Maximum password length supported by kernel: 72

Hashes: 7 digests; 7 unique digests, 7 unique salts
Bitmaps: 16 bits, 65536 entries, 0x0000ffff mask, 262144 bytes, 5/13 rotates

Applicable optimizers applied:
* Zero-Byte
* Brute-Force

Watchdog: Hardware monitoring interface not found on your system.
Watchdog: Temperature abort trigger disabled.

Host memory required for this attack: 81 MB
Reply
#2
This sounds like you may be hitting a few issues:

For one, your command is wrong. "-a??a?a?a?a?a?a?a" is a broken mask, look at the front.

Second, 250MH/s in bcrypt is impossible, doesnt matter the hardware. 250H/s would even be impressive for such a weak device. I would double check that first number.

Third, CPU util will, of course, be almost flat. It's not using the CPU at all, it's using the GPU. You are already going about as fast as that chip can handle for your bcrypt cost, which appears to be about 80H/s. You might gain some speed with the bank conflict optimizations but those aren't in hashcat yet.
Reply
#3
(04-28-2021, 07:20 PM)Chick3nman Wrote: This sounds like you may be hitting a few issues:

For one, your command is wrong. "-a??a?a?a?a?a?a?a" is a broken mask, look at the front.

Second, 250MH/s in bcrypt is impossible, doesnt matter the hardware. 250H/s would even be impressive for such a weak device. I would double check that first number.

Third, CPU util will, of course, be almost flat. It's not using the CPU at all, it's using the GPU. You are already going about as fast as that chip can handle for your bcrypt cost, which appears to be about 80H/s. You might gain some speed with the bank conflict optimizations but those aren't in hashcat yet.

Typo on that mask...

Attaching a screenshot of a current crack I'm working on... Showing 284MH/s - maybe I'm crazy, maybe code behind is wrong...


Attached Files
.png   Screen Shot 2021-04-28 at 1.22.19 PM.png (Size: 326.51 KB / Downloads: 9)
Reply
#4
Something is seriously broken with your runtime. I wouldn't trust that it has actually tested any of the passwords correctly at all. 250MH/s for 10 cost bcrypt would be the same as about 58,000 RTX 3090's.
Reply
#5
maybe double checking with known plain "hashcat" from examples

cost 5 and 10

$2a$05$LhayLxezLhK1LhWvKxCyLOj0j1u.Kj0jZ0pEmm134uzrQlFvQJLF6
$2a$10$coQsOmjEnDpQgrsCphDM8./ssZKRPRnHA.aciAK7NaYL.EBdqIFf6

wasnt there news that apple wants to drop opencl support?
Reply