I saw some mention of Linux ARM at the end of the video. With the release of NVIDIA Spark, it looks like there's a perfect platform for Karma XPU to run on. Is there anything can share?

Hi chongsong, yes! As mentioned in the keynote, Houdini 22 will ship with a Linux ARM build. Just like our original Apple silicon port, it will be designated as a "Technology Preview" as we address feedback from our users and make sure that it's ready for prime-time.

We've tested Houdini on the following platforms so far:

• AWS Graviton
  
• Ampere Altra/Altra Max
  
• NVIDIA DGX Spark
  

The build targets the Neoverse-n1 instruction set. This corresponds to a minimum of AWS Graviton2 and Ampere Altra. Unfortunately this means that you most likely won't be able to run Houdini on low-end SBCs. You can see if your CPU supports all of the features of neoverse-n1 here [gpages.juszkiewicz.com.pl]

The only other caveat to this build is that there's currently no support for FBX as it isn't available for the platform yet.

This build features not only the command-line utilities, but the entire interactive interface as well, meaning that it's fully supported on Arm workstations from the likes of Nvidia, Ampere and System76.

johnmather
Hi chongsong, yes! As mentioned in the keynote, Houdini 22 will ship with a Linux ARM build. Just like our original Apple silicon port, it will be designated as a "Technology Preview" as we address feedback from our users and make sure that it's ready for prime-time.

We've tested Houdini on the following platforms so far:

• AWS Graviton
  
• Ampere Altra/Altra Max
  
• NVIDIA DGX Spark
  

The build targets the Neoverse-n1 instruction set. This corresponds to a minimum of AWS Graviton2 and Ampere Altra. Unfortunately this means that you most likely won't be able to run Houdini on low-end SBCs. You can see if your CPU supports all of the features of neoverse-n1 here [gpages.juszkiewicz.com.pl]

The only other caveat to this build is that there's currently no support for FBX as it isn't available for the platform yet.

This build features not only the command-line utilities, but the entire interactive interface as well, meaning that it's fully supported on Arm workstations from the likes of Nvidia, Ampere and System76.

Hello, I would like to know more information about running the arm64 version on DGX Spark. Can Karma XPU call the GPU for rendering on it, and what advantages are there for memory allocation under a unified architecture? I am considering getting one as a local LLM platform, and it would be great if it can also serve as a lookdev platform.

chongsong
Can Karma XPU call the GPU for rendering on it, and what advantages are there for memory allocation under a unified architecture?

Yes, Karma XPU is fully supported, however be aware that it currently has no knowledge about unified memory architectures. Because of this, the data in RAM will still be "uploaded" to "VRAM", potentially reducing the amount of available memory. This is something we plan to address however.

johnmather
Yes, Karma XPU is fully supported, however be aware that it currently has no knowledge about unified memory architectures. Because of this, the data in RAM will still be "uploaded" to "VRAM", potentially reducing the amount of available memory. This is something we plan to address however.

Thanks for your reply. Really helpful.