The Definitive Guide toAI Data Centers
Ask the GuideAboutAccount
Guide › Glossary › GPUDirect Storage

GPUDirect Storage · GDS

NVIDIA technology moving data directly from storage into GPU memory, bypassing the CPU bounce buffer.

Current numbers

~7x / ~10xDynamo + wide-EP MoE throughput on GB200 NVL72 vs B200 (Dynamo 1.0 GA at GTC 2026); NIXL+GPUDirect Storage prefill speedup for long contextas of 2026 · register ↗

Related terms: CPU · DALI · GPU

Where the guide uses this

9.3
NVMe Tiers, GPUDirect Storage & the CPU-Bypass Data Path
14.4
Reliability Engineering for Training (Operational)
9.7
Inference & KV-Cache Storage: The New Memory Hierarchy
1.7
The Requirements-and-Consequences Matrix
1.2
Training Data Centers: Synchronous, Dense, Checkpointable
8.3
Network Silicon: Switch ASICs, NICs & DPUs

← All terms