M365Con 2027 – free online conference · 11–22 Jan 2027 · 120 sessions m365con.net

Graphics processing unit (GPU)

Concept

✦ CONCEPTGraphics processing unit(GPU)M365.FM
Title graphic: Graphics processing unit (GPU) (Concept)

Often discussed together

Episodes about Graphics processing unit (GPU) often also cover Retrieval-augmented generation (RAG).

TypeConcept
WikidataWikidata entry for Graphics processing unit (GPU) Q183484

This short description is not yet available in your language – feel free to add it.

Key points about Graphics processing unit (GPU)

Key points about Graphics processing unit (GPU) from the episodes, each with its source episode:

Episodes about Graphics processing unit (GPU)

Episodes in which Graphics processing unit (GPU) is discussed:

Timeline of episodes about Graphics processing unit (GPU)

Articles about Graphics processing unit (GPU) (4)

Articles in which Graphics processing unit (GPU) is a topic:

Frequently asked questions

What is CPU fallback in ONNX Runtime?

It is when work meant for the GPU quietly runs on the CPU instead. The episode traces how this silent fallback makes throughput collapse and latency unpredictable while dashboards look healthy. · From the episode: Diagnose GPU Underutilization and CPU Fallback in AI Systems

Why is my GPU underutilized and AI latency unstable?

The episode points to the compute pipeline rather than model design. Silent CPU fallback and version mismatches across CUDA, cuDNN, TensorRT and ONNX can leave an apparently adequate GPU idle. · From the episode: Diagnose GPU Underutilization and CPU Fallback in AI Systems

What workload does the episode use as an example?

A Stable Diffusion workload under concurrency with strict P95 latency objectives. On paper the GPU hardware looks perfectly adequate, which makes the collapsing throughput and rising cost so puzzling. · From the episode: Diagnose GPU Underutilization and CPU Fallback in AI Systems

Why do Azure AI Foundry projects fail?

Common reasons are unclear goals, poor data quality, weak governance, technical mistakes and missing observability. Teams also set unrealistic expectations or ignore ongoing care, while complexity, GPU limits and rising costs add pressure. · From the episode: Fix Common Azure AI Foundry Project Failures

How can teams avoid unpredictable Azure AI costs?

Monitor spending closely and simplify the project architecture, which reduces cost and improves manageability. Without close monitoring, costs can rise quickly, especially alongside limited GPU access. · From the episode: Fix Common Azure AI Foundry Project Failures

Community wiki

Short info above, community knowledge below: add practical tips, examples or corrections. New contributions are reviewed first; trusted members publish directly.

Experts and companies for this topic

Nobody has listed this topic yet. Add it to your profile.

Sources

Maintained by Mirko Peters (Microsoft MVP) · Last updated: