InfoQ Homepage GPU Content on InfoQ
-
From S3 to GPU in One Copy: Rethinking Data Loading for ML Training
Onur Satici discusses Vortex, an open-source columnar file format designed to bypass CPU bottlenecks, enabling ultra-fast S3-to-GPU data streaming and dynamic query pruning at up to 60 Gbps.
/presentations/vortex-columnar-file-format-gpu-streaming/en/mediumimage/onur-satici-medium-1787813397611.jpeg)
-
Running AI at the Edge: Running Real Workloads Directly in the Browser
James Hall explains why engineering teams should shift AI workloads on-device. He shares local inference strategies, WebGPU optimization techniques, and architecture choices for zero-trust privacy.
/presentations/local-ai-browser-inference-privacy/en/mediumimage/james-hall-medium-1787813225372.jpeg)
-
From Fab To Token - The State Of The Market
Jordan Nanos explains how hardware constraints, data center scale, and chip co-design shape modern AI performance and tokenomics from silicon to inference.
/presentations/ai-hardware-tokenomics/en/mediumimage/jordan-nanos-medium-1786538855211.jpg)
-
Chaos Engineering GPU Clusters
Bryan Oliver explains how to apply chaos engineering to massive GPU clusters. Learn how to handle hardware variability, NUMA nodes, and network faults to secure your AI infrastructure.
/presentations/chaos-engineering-gpu/en/mediumimage/bryan-oliver-medium-1782819161258.jpeg)
-
Realtime and Batch Processing of GPU Workloads
Joseph Stein explains how to build a highly available private AI cloud. He shares blueprints on scaling vLLM on enterprise GPUs, implementing gateway guardrails, and optimizing batch workloads.
/presentations/realtime-gpu-workloads/en/mediumimage/medium-1779194310932.jpg)
-
Pitfalls of Unified Memory Models in GPUs
Joe Rowell explores the use of unified memory on modern GPU, the low-level details of how unified memory is realized on an x86-64 system, and some of the tools to understand what's happening on a GPU.
/presentations/gpu-memory-tools/en/mediumimage/Joe-Rowell-medium-image-1727793898391.jpeg)
-
Fuelling the AI Revolution with Gaming
Alison Lowndes talks about the HW & SW that comprise NVIDIA's GPU computing platform for AI, across PC to data center, cloud to edge, training to inference.
/presentations/nvidia-gpu-ai-gaming/en/mediumimage/Alibig-1527954781033.JPG)
-
Taming GPU Threads with F#
Daniel Egloff overviews Alea, an F# alternatives to CUDA C/C++ and OpenCL C++, showing how to write GPU scripts and perform dynamic compilation in F#.
/presentations/gpu-threads-fsharp/en/mediumimage/Danielbig.jpg)
-
Machine Learning at Netflix Scale
Aish Fenton discusses Netflix' machine learning algorithms, including distributed Neural Networks on AWS GPUs, providing insight into offline experimentation and online AB testing.
/presentations/machine-learning-netflix-2014/en/mediumimage/Aishbig.jpg)
-
Extreme Speedups and GPGPU: A Tale of Two Practical Uses of Reified Trees
Olivier Chafik discusses how to make a practical use of reified trees in Scala, with two applications: run-time (re)compilation for extreme speed, and conversion to another language (OpenCL).
/presentations/scala-gpgpu-reified-tree/en/mediumimage/Olibig.jpg)
-
Excel Coding Errors Are Destroying World Economies and F# (with Tsunami) Is Here to Stop Them!
Matthew Moloney discusses using F# and .NET inside Excel, demonstrating doing big data, cloud computing, using GPGPU and compiling F# Excel UDFs.
/presentations/excel-f-sharp-tsunami/en/mediumimage/Matbig.jpg)
-
Accelerating the Web: How GPUs Make Browsers Fast
Jarred Nicholls explains how browsers leverage the GPU to speed up complex web pages by primitive drawing, composing layers and using tiles backing stores.
/presentations/GPU-Browser/en/smallimage/Jar.JPG)