11-29-2024, 04:59 PM
(This post was last modified: 11-29-2024, 05:05 PM by codecuriously.)
I recently saw a post on Reddit that got me thinking:
https://www.reddit.com/r/AsahiLinux/comm...ahi_linux/
It appears as though on Asahi Linux (running on top of Apple Silicon) that hashcat tries to allocate all available VRAM which causes the system to crash. This looks to be because the Apple SoC uses unified memory and with hashcat asking for all of this it causes obvious issues as the system runs out of memory.
The OS/driver developers are working on a fix by setting a hard limit (85%) of memory that can be allocated as VRAM. I believe this is what MacOS currently does to solve this.
My question leads on from that. Is it possible to specify a total or percentage of VRAM that hashcat can use. I think this would have broader use for when you want to use the system during a long running hashcat process.
I have tested "Workload Profiles" (-w1) but this does not work and causes the same issue.
Seems to present the same as a previous issues with Intel iGPUs.
https://github.com/hashcat/hashcat/issues/3426
https://www.reddit.com/r/AsahiLinux/comm...ahi_linux/
It appears as though on Asahi Linux (running on top of Apple Silicon) that hashcat tries to allocate all available VRAM which causes the system to crash. This looks to be because the Apple SoC uses unified memory and with hashcat asking for all of this it causes obvious issues as the system runs out of memory.
The OS/driver developers are working on a fix by setting a hard limit (85%) of memory that can be allocated as VRAM. I believe this is what MacOS currently does to solve this.
My question leads on from that. Is it possible to specify a total or percentage of VRAM that hashcat can use. I think this would have broader use for when you want to use the system during a long running hashcat process.
I have tested "Workload Profiles" (-w1) but this does not work and causes the same issue.
Seems to present the same as a previous issues with Intel iGPUs.
https://github.com/hashcat/hashcat/issues/3426