JAjamesfmilne·Dec 13, 2025
I found some useful clues here. Looks like it uses the regular InfiniBand RDMA APIs.
https://github.com/Anemll/mlx-rdma/commit/a901dbd3f9eeefc628...
In macOS 26.2 enables fast AI clusters with RDMA over ThunderboltAnemll
mlx-rdma
C++2025–2025
Anemll / mlx-rdma
mlx-rdma
experiments with MLX
2Distinct people
2Mentions
0Explicit recommenders
73GitHub stars
What to know before exploring
What it is
experiments with MLX
GitHub
Where it is used
No documented product use has been verified yet.
What people are saying
No recommendation phrase was found in the indexed comments.
No classified warning is attached to this page. Read the discussions before treating silence as approval.
Also in these discussions
No other repository is repeatedly linked in the same discussions.
This week on Hacker News
Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 3:56:07 PM UTC.
No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.
Mentions over time
HN mentions by day2 dated mentions
Observed mentions: 2025-11-25 17:24:03 UTC to 2025-12-13 20:22:15 UTC.
Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.
View mention counts and date ranges
| Period | Start (UTC) | End (exclusive, UTC) | Mentions | Submissions | Comments |
|---|---|---|---|---|---|
| Nov 25 | 2025-11-25 00:00:00 UTC | 2025-11-26 00:00:00 UTC | 1 | 0 | 1 |
| Nov 26 | 2025-11-26 00:00:00 UTC | 2025-11-27 00:00:00 UTC | 0 | 0 | 0 |
| Nov 27 | 2025-11-27 00:00:00 UTC | 2025-11-28 00:00:00 UTC | 0 | 0 | 0 |
| Nov 28 | 2025-11-28 00:00:00 UTC | 2025-11-29 00:00:00 UTC | 0 | 0 | 0 |
| Nov 29 | 2025-11-29 00:00:00 UTC | 2025-11-30 00:00:00 UTC | 0 | 0 | 0 |
| Nov 30 | 2025-11-30 00:00:00 UTC | 2025-12-01 00:00:00 UTC | 0 | 0 | 0 |
| Dec 1 | 2025-12-01 00:00:00 UTC | 2025-12-02 00:00:00 UTC | 0 | 0 | 0 |
| Dec 2 | 2025-12-02 00:00:00 UTC | 2025-12-03 00:00:00 UTC | 0 | 0 | 0 |
| Dec 3 | 2025-12-03 00:00:00 UTC | 2025-12-04 00:00:00 UTC | 0 | 0 | 0 |
| Dec 4 | 2025-12-04 00:00:00 UTC | 2025-12-05 00:00:00 UTC | 0 | 0 | 0 |
| Dec 5 | 2025-12-05 00:00:00 UTC | 2025-12-06 00:00:00 UTC | 0 | 0 | 0 |
| Dec 6 | 2025-12-06 00:00:00 UTC | 2025-12-07 00:00:00 UTC | 0 | 0 | 0 |
| Dec 7 | 2025-12-07 00:00:00 UTC | 2025-12-08 00:00:00 UTC | 0 | 0 | 0 |
| Dec 8 | 2025-12-08 00:00:00 UTC | 2025-12-09 00:00:00 UTC | 0 | 0 | 0 |
| Dec 9 | 2025-12-09 00:00:00 UTC | 2025-12-10 00:00:00 UTC | 0 | 0 | 0 |
| Dec 10 | 2025-12-10 00:00:00 UTC | 2025-12-11 00:00:00 UTC | 0 | 0 | 0 |
| Dec 11 | 2025-12-11 00:00:00 UTC | 2025-12-12 00:00:00 UTC | 0 | 0 | 0 |
| Dec 12 | 2025-12-12 00:00:00 UTC | 2025-12-13 00:00:00 UTC | 0 | 0 | 0 |
| Dec 13 | 2025-12-13 00:00:00 UTC | 2025-12-14 00:00:00 UTC | 1 | 0 | 1 |
What Hacker News says
2 sourcesIn macOS 26.2 (Tahoe) beta, Apple introduced a low-latency Thunderbolt 5 RDMA driver, enabling up to 80 Gb/s bidirectional bandwidth for Mac clustering—ideal for distributed ML on Apple Silicon. It's optimized for low latency, delivering ~14 Gbps throughput at 4K MTU.
My tests (M4 Pro to M3 Ultra): Stock ibv_uc_pingpong achieved ~14 µs round-trip for 4K packets (requires GID index setup). Custom C++ variant hit 6-13 µs/iter: https://x.com/anemll/status/1993192776897642942
Code and details:
https://github.com/Anemll/mlx-rdma/blob/anemll-rdma/ibv_roun...
https://github.com/Anemll/mlx-rdma/blob/anemll-rdma/ibv_roun... (includes steps to enable RDMA in macOS Recovery OS terminal)
Theoretically, this accelerates pipeline parallelism (faster layer handoffs) and tensor parallelism (low-overhead sharding) on GPUs, with potential extensions to ANE
Read the rest
for real-time AI workflows.