hashcat Forum

Hi, guys!
I have a pc with Intel Xeon + RAM 16GB + Windows 10 x64 and I use the latest version of hashcat 6.1.1.

My config for GPUs:

Quote:CUDA API (CUDA 11.0)
====================
* Device #1: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #2: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #3: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #4: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #5: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #6: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #7: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU
* Device #8: GeForce RTX 2080 Ti, 9248/11264 MB, 68MCU

But when I try to use hashcat at full speed with 8 GPUs I have this error:

Quote:Insufficient memory available
cuMemAlloc(): out of memory

I did several tries and discovered a very strange RAM limitation:
If I use 1 or 2 or 3 GPUs - everything is good and working ok:
1 GPU - Host memory required for this attack: 1164 MB
2 GPUs - Host memory required for this attack: 2329 MB
3 GPUs - Host memory required for this attack: 3494 MB
4 GPUs - cuMemAlloc(): out of memory
...
8 GPUs - cuMemAlloc(): out of memory

But If I try to use more than 3 GPUs i.e. 4 or 8 simultaneously
I always receive error "out of memory". And the same situation for both CUDA and OpenCL.

As I understand it looks like x32 limitation of max RAM 4G.
But my OS is x64 and I have more than 4GB of RAM.

Is there a way to run hashcat in x64 compatible mode?
Early it were `hashcat64` and `hashcat32` but now it has only one binary.

Please help me to understand what I'm doing wrong and how to fix it.

You should have either equal to or more than the amount of VRAM you have as system RAM. So with 8x 2080 TIs, you need at least 88gb of system RAM, not 16gb.

.... (what chick3nman said) and it of course has nothing to do with 32-bit binaries... hashcat doesn't ship with 32-bit executables anymore (you could still compile them yourself if somebody really needs 32-bit support).
It's definitely not a 32-bit problem, because hashcat only has 64-bit binaries since several releases.

you just don't have enough RAM, 16 GB is way too little, it's enough for 1 GPU, but definitely not for that many GPUs.

Thank you guys for your help!