Hi,

Is there anyone running Houdini and Nuke on Mac Studio M2 Max? I know it isn't any good for rendering tasks compared to RTX GPUs, but how is it in general? My Windows PC is giving me hard times popping blue screen and other random crashes even after installing clean OS. I have no idea what to do with it.

Best

Frequent bluescreens generally indicate that you've got either a bad driver that has kernel access (seems unlikely with a clean reinstall), or your hardware has an issue. Bad memory is a frequent culprit. You can test it with one of the many utilities available, or you can pull some of the DIMMs, run for a while and see if it still happens, put some back in & pull others, to narrow it down to the bad module. Swapping out other components is a little more tricky, especially if you don't have spares, or if you have a laptop.

Edit: Clogged heatsinks (dirt, hair) are another source of frequent shutdowns, especially if they occur under load.

JuhaT
Mac Studio M2 Max

If you are looking to buy new, the newer M-series will be out later this year so I would not advise M2 as the orders of magnitude in performance gains will be noticeable since M3 skipped the product line.

I have seen updates to Karma this series of 20.5 that sped it up on an M1 8GB. Yet, Vulkan is not available.

If you need to render all the things all the time every-time, I am not sure that the Metal Spec has been integrated enough yet at an industry level to move adoption of the API in established software.

I think the biggest win with going with a Studio Max in the newer series is that you will be able to run local documentation and scene-graph informed dialogues with Apple Intelligence on device. USD and MaterialX is native on Apple Platforms and XCode versions now on will have AI Code Assist.

Soon -- probably over the next technology cycle of your use case -- rendering will be using Neural Primitives as precomputed latent physics space in a compressed reduced order model that includes the render equation. I mean the future maybe rendering in a Python Panel with with the Houdini Web-server and Browser Logic, device agnostic...

This was 2 years ago and the pace of research is very fast right now.


I would think about the larger picture of the pipeline and identify computational bottlenecks that hinder your work, then ask: Will a precomputed Physics-based AI Render Solution solve this and can I use the software I have to make that DIY?

And lastly: What will hold you back from putting in the time to study with an AI tutor on device trained on documentation and can answer direct questions with examples?