#gpu-scaling2 articles

2 articles

AIBullisharXiv – CS AI · Feb 277/106

🧠

veScale-FSDP: Flexible and High-Performance FSDP at Scale

Researchers introduce veScale-FSDP, a redesigned Fully Sharded Data Parallel system that overcomes limitations of current FSDP implementations used for training large-scale AI models. The new system features flexible RaggedShard format and structure-aware planning, achieving 5-66% higher throughput and 16-30% lower memory usage while supporting advanced training methods and scaling to tens of thousands of GPUs.

AINeutralIEEE Spectrum – AI · Dec 277/104

🧠

AI Data Centers Demand More Than Copper Can Deliver

AI data centers are hitting physical limits with copper cables as GPU-to-GPU data rates approach terabit-per-second speeds, requiring thicker, shorter cables that complicate dense connections. Startups Point2 Technology and AttoTude are developing radio-based cable solutions that promise longer reach, lower power consumption, and narrower cables than copper alternatives.

$LINK$MKR