Windows/6900xt Slow cracking speeds?
#1
I've been playing around with my kids 6900xt and hashcat to see what kind of performance I can get out of it. Benchmarking gives similar results as to what are posted on this forum. Specifically looking at NTLM, the benchmark shows about 95000Mh/s. However, when I am actually cracking passwords, I get less than 3000Mh/s. What gives? I'm currently running it with a wordlist and ruleset, which I assume will slow it down some, but I didn't expect that much!

I'm running Hashcat 6.2.5 (with the OpenCL API) and have the latest AMD drivers installed for the 6900xt. Should I revert to older versions?

I also have my personal computer that runs an AMD 5600xt and an RTX 3050 with the latest drivers and Hashcat version and BOTH cards individual get better performance than the 6900xt.

I also apologize if this has been brought up before....I tried to search for a similar thread but didn't find one.
Reply
#2
Command hashcat -I and paste here resualt
Reply
#3
(07-24-2022, 09:38 AM)marc1n Wrote: Command hashcat -I and paste here resualt

hashcat (v6.2.5) starting in backend information mode

hipDeviceGetAttribute(): 1

HIP Info:
=========

HIP.Version.: 5.0.20451

Backend Device ID #1
  Name...........: AMD Radeon RX 6900 XT
  Processor(s)...: 0
  Clock..........: 0
  Memory.Total...: 0 MB
  Memory.Free....: 0 MB
  PCI.Addr.BDFe..: 0000:00:00.0

OpenCL Info:
============

OpenCL Platform ID #1
  Vendor..: Advanced Micro Devices, Inc.
  Name....: AMD Accelerated Parallel Processing
  Version.: OpenCL 2.2 AMD-APP (3417.0)

  Backend Device ID #2
    Type...........: GPU
    Vendor.ID......: 1
    Vendor.........: Advanced Micro Devices, Inc.
    Name...........: AMD Radeon RX 6900 XT
    Version........: OpenCL 2.0 AMD-APP (3417.0)
    Processor(s)...: 40
    Clock..........: 2135
    Memory.Total...: 16368 MB (limited to 13695 MB allocatable in one block)
    Memory.Free....: 16256 MB
    OpenCL.Version.: OpenCL C 2.0
    Driver.Version.: 3417.0 (PAL,LC)
    PCI.Addr.BDF...: 2d:00.0

I'm running this on Windows 11.
Reply
#4
Bumping to see if anyone has any ideas...
Reply
#5
If you're seeing performance loss, there could many factors. Please post the console of your attack first. There could many factors though such as your drivers, hardware, temperatures, overclocks if you have them, or it could be as simple as appending -O to your attack.
Reply
#6
maybe your wordlist and rulessets are to small

as rules are applied on the GPU/MOD-loop this can be one possiblitity
Reply
#7
Its not an issue of wordlist/rulesets being too small. I'm running a wordlist that is 120GB and a ruleset with 65000+ rules.

As for drivers, running the latest drivers for the 6900xt as managed by the latest AMD Adrenalin software. Hardware shouldn't be an issue....I can list those if needed, but the computer was built in the last few years with new parts. Temperatures are all fine, no overclocks, and I do have -O appended on my attack.

Current cracking speeds for NTLMv1 (mode 1000):
RTX 3050 (in Computer 1): 6270.7 MH/s
Radeon 5600xt (in Computer 1): 4144.5 MH/s
Radeon 6900xt (in Computer 2): 3538.2 MH/s

While Computer 1 has a better motherboard and CPU, I can't imagine that is a significant factor causing such a discrepancy between the two. Maybe I'm wrong.....
Reply
#8
(08-26-2022, 08:57 PM)cyberomak Wrote: Its not an issue of wordlist/rulesets being too small. I'm running a wordlist that is 120GB and a ruleset with 65000+ rules.

As for drivers, running the latest drivers for the 6900xt as managed by the latest AMD Adrenalin software. Hardware shouldn't be an issue....I can list those if needed, but the computer was built in the last few years with new parts. Temperatures are all fine, no overclocks, and I do have -O appended on my attack.

Current cracking speeds for NTLMv1 (mode 1000):
RTX 3050 (in Computer 1): 6270.7 MH/s
Radeon 5600xt (in Computer 1): 4144.5 MH/s
Radeon 6900xt (in Computer 2): 3538.2 MH/s

While Computer 1 has a better motherboard and CPU, I can't imagine that is a significant factor causing such a discrepancy between the two. Maybe I'm wrong.....

im not quite sure which, but there is a HIP version currently not supported by hashcat, when starting a simple cracking task is there any message about unsupported HIP version and falling back to opencl?

try to compare hahscat -I output from both computers (take a look at HIP version and stats for your devices), see your fist post, HIP is recognized, but the device shows 0 ram, 0 processors and i think hashcat skips this device as it seems incapable of running an attack, maybe a downgrade of HIP fixes this
Reply