Search papers, labs, and topics across Lattice.
4
0
3
Unlocking the hidden patterns in HPC log data can drastically improve system resilience and predictive maintenance strategies.
Squeezing more performance from power-constrained CPU-GPU clusters is now possible: EcoShift dynamically allocates reclaimed power based on application-specific sensitivity, boosting performance by up to 6%.
EcoSched reveals that optimizing GPU count and co-scheduling can lead to up to 14.8% energy savings in high-performance computing systems.
HPC admins can now visually explore complex system behavior and detect subtle anomalies using a new tool that automatically clusters nodes and highlights performance variations.
