News for hackers. Stuff that matters.

A comment on Apple's MLX adding CUDA support. Parent comment

Re:Apple's MLX adding CUDA support by zcbenz on Monday July 14, 2025 @07:20PM (#44566412)
In the absence of hardware unified memory, CUDA will automatically copy data between CPU/GPU when there are page faults.
[ Parent | on HN ]
Re:Apple's MLX adding CUDA support by fenced_load on Monday July 14, 2025 @08:08PM (#44566633)
There is also NVLink c2c support between Nvidia's CPUs and GPUs that doesn't require any copy, CPUs and GPUs directly access each other's memory over a coherent bus. IIRC, they have 4 CPU + 4 GPU servers already available.
[ Parent | on HN ]
Re:Apple's MLX adding CUDA support by benreesman on Monday July 14, 2025 @08:35PM (#44566751)
Yeah NCCL is a whole world and it's not even the only thing involved, but IIRC that's the difference between 8xH100 PCI and 8xH100 SXM2.
[ Parent | on HN ]
Re:Apple's MLX adding CUDA support by nickysielicki on Monday July 14, 2025 @09:21PM (#44566987)
See also: https://www.kernel.org/doc/html/v5.0/vm/hmm.html
[ Parent | on HN ]
Re:Apple's MLX adding CUDA support by saagarjha on Monday July 14, 2025 @10:02PM (#44567184)
This seems like it would be slow…
[ Parent | on HN ]
Re:Apple's MLX adding CUDA support by freeone3000 on Monday July 14, 2025 @10:28PM (#44567307)
Matches my experience. It’s memory stalls all over the place, aggravated (on 12.3 at least) there wasn’t even a prefetcher.
[ Parent | on HN ]