Incident Library

Organisation

IBM

IBM · ibm

2 cases · 2025

Root cause Hardware failure1unknown1
Component PCIe link1Server power supply and power capping1
Contributing factor –
Symptom Throughput degradation2

Cases

  1. IL-0419 A Granite-20B training job on IBM's Vela ran almost 3x slower because one node's power card fault throttled its GPUsIBM · 2025

    Root cause Hardware failure

  2. IL-0420 PCIe links on IBM's Vela GPU nodes dropped to a lower generation's speed and slowed training jobs by up to 2xIBM · 2025

    Root cause unknown