GPU

In the Ultra Transcenders books

AZ-305AZ-900AI-200DP-750AZ-400AZ-802MD-102ALZ

Each book explains GPU in context, with comparison tables and the common traps.

Terms in this definition

• LIKE

  Compares strings with a pattern that can contain the % and _ wildcards. Because it only understands character patterns, searching big volumes of text this way is much slower than using full-text search.

Related terms

• AI Runtime

  Microsoft Learn steers GPU deep learning towards this rather than Databricks Runtime ML. It provides serverless GPUs, currently in Public Preview, for fine-tuning and bespoke deep learning models.

• Compute cluster

  Azure Machine Learning compute of several managed CPU or GPU nodes that scales from min_instances to max_instances with each job; set the minimum to 0 and an idle scale-down time so you pay only when jobs are running.

• Compute metrics

  Hardware, Spark and GPU figures sampled once a minute and kept for 30 days, shown on the Metrics tab of classic all-purpose and jobs compute. Query insights replace it on serverless compute.

• Dedicated access mode

  Assigns classic compute to one person or one group. Some capabilities work only in this mode, among them R, the RDD API, GPU instances and the ML flavour of Databricks Runtime.

• GPU partitioning

  A Hyper-V capability that lets several virtual machines share one physical graphics card through SR-IOV. Windows Server 2025 added live migration for virtual machines with GPU partitions.

• Managed compute

  Way of deploying models in Foundry on dedicated GPU VMs that Foundry manages, exposed through a REST endpoint and charged by compute hour. It suits open-source, partner and custom models, not Azure OpenAI ones.

• NC-series

  VM family with NVIDIA GPUs used for GPU workload profiles in Container Apps, including Consumption-GPU-NC8as-T4 and Dedicated NC24-A100.