Advanced Mac users may wish to manually increase the VRAM allocation on their Apple Silicon Mac for performance reasons when engaging in graphics intensive tasks like running LLMs locally, AI models, or any graphics heavy applications, whether for gaming or video editing. This is possible because Apple Silicon chips offer unified memory architecture, meaning the CPU and GPU share the same memory pool, and while the system will dynamically allocate memory as needed, advanced users can also manually tweak VRAM allocation from RAM on their Mac.

Important: This is genuinely only appropriate for advanced Mac users who have a compelling reason to adjust VRAM themselves, as improperly allocated VRAM may cause significant performance issues, memory problems, or even system crashes. If you’re not familiar with how memory allocation may impact performance, for better or worse, and you’re not comfortable using the command line, this one is not for you.

How to Manually Adjust VRAM Allocation on Apple Silicon Mac

Open the Terminal and use the following command syntax, replacing (size in MB) with the size in megabytes of RAM you’d like to allocate to VRAM:

sudo sysctl iogpu.wired_limit_mb=(size in MB)

For example, to set VRAM to 18GB:

sudo sysctl iogpu.wired_limit_mb=18432

Or, if you have a Mac with a large amount of RAM, say 128GB total, and you want to allocate 120GB to VRAM, you could use the following:

sudo sysctl iogpu.wired_limit_mb=122880

Hit return and authenticate as usual, and the setting will take effect right away.

Considerations when manually adjusting VRAM allocation on Mac with Apple Silicon

You need to be be mindful of how much total RAM you have on the Mac before making these sort of adjustments, and you’ll want to make sure MacOS itself has plenty of RAM available as well.

Generally speaking, the more total RAM you have, the more generous you can be with your VRAM allocation. For example, if a Mac has 128GB total RAM, you could allocate 120GB to VRAM (assuming nothing else is running on the Mac), but if your Mac only has 8GB RAM, you might not want to manually adjust memory allocation at all. What you definitely do not want to do is take a Mac with 16GB RAM and try to allocate all of that to VRAM, which would inevitably cause some serious issues on the Mac as it would immediately run out of memory for anything else.

Making these type of adjustments can potentially improve the performance of software that is dependent on GPU usage and video memory, which includes graphics intensive apps ranging from games to video editing to rendering, to even many popular AI tools whether for creating AI art, or running LLMs locally. Perhaps you’re Llama locally and want to speed it up a bit, or maybe you’re playing around with running DeepSeek locally and want to adjust its performance, or maybe you’re having fun with DiffusionBee and want to give it a boost.

If you are going to proceed with manually adjusting your video RAM allocation on an Apple Silicon Mac, you’ll likely want to optimize performance overall on the Mac for whatever the intended purpose is, like for a game or LLM, so you should quit all other apps, and minimize any processes or tasks running in the background.

And yes this works the same on any Apple Silicon Mac at present, whether it’s an M1, M2, M3, M4, M5, M6, whatever. This will not work on an Intel Mac.

How to return to default VRAM allocation on Apple Silicon Mac

You can simply reboot the Mac to return the VRAM allocation to its default dynamic memory management state.

–

I stumbled into this handy trick on a Github thread for running Llama locally while trying to speed up performance on my M-series Mac, and

Have you manually tweaked VRAM allocation on your Mac before, and for what purpose? Did you find that performance was better with a more generous RAM allocation? Do you have any related interesting tips, advice, or suggestions? Share your experiences and thoughts in the comments.

The above is the detailed content of How to Increase VRAM Allocation on Apple Silicon Mac. For more information, please follow other related articles on the PHP Chinese website!