Slurm Learns Topology — Rack-Scale Scheduling Becomes a Hard Constraint
NVIDIA's GB200 NVL72 has turned GPU cluster scheduling into a topology-aware discipline, and the Slurm block plugin is how the software stack catches up to the hardware. Plus Cloudflare cuts twenty percent of its workforce citing a six hundred percent jump in internal AI usage, and Q-CTRL just clocked a three-thousand-times quantum speedup on real hardware without fault tolerance.
Welcome to Amaze Networks for Thursday, May eighth. Okay, I need to lead with the quantum story today because it is genuinely too good to bury.
Please do, because I've been sitting on this since I read it this morning and it deserves the full setup.
So Q-CTRL and IBM just published results showing a three-thousand-times speedup in simulating the Fermi-Hubbard model on one hundred twenty qubits. Three thousand times faster than the best classical methods for the same problem.
New episodes, every weekday.
Amaze Networks drops at 4 AM CT, Monday through Friday. Spotify and Apple Podcasts submissions in progress.