Recent Posts - Page 2#
Parallelism for Long-Context Serving: GLM-5.3-Flash on MI300X
Measured throughput of GLM-5.3-Flash on MI300X: how KV-cache capacity shapes TP8 and DP8+EP8 serving, with BF16 and FP8 KV cache.
Introducing the ROCm Handbooks
Discover the ROCm handbooks: book-style HTML and PDF guides that take you from your first HIP kernel to optimizing code for AMD GPUs.
MONAILabel 0.8.5 on AMD Instinct™ GPUs: Interactive Medical-Image Annotation
Learn how to run MONAILabel 0.8.5 for interactive medical-image annotation on AMD Instinct™ GPUs.
PrimusClaw: An Open-Source Managed Agent Runtime for AMD Instinct GPU Clusters
PrimusClaw is an open-source managed agent runtime that hosts durable, sandboxed agent sessions on your own AMD Instinct GPU clusters.
ROCm InferaSim: Cut LLM Serving Trial Costs and Deploy the Best Configuration
Learn how ROCm InferaSim predicts LLM serving performance on AMD Instinct GPUs from a few short runs, so you can deploy the best configuration.
From Operator to Kernel: Profile DeepSeek-V4-Flash in Primus with madengine
Profile an eight-layer training proxy derived from the DeepSeek-V4-Flash Primus configuration on eight MI350X GPUs with madengine and TraceLens, then turn operator-to-kernel insights into faster runs.
Online Quantization: Smooth Path from One Checkpoint to Any Desired Serving Scenario
Learn how AMD Quark's online quantization quantizes LLM weights at load time across vLLM, ATOM, and SGLang, with per-layer mixed FP8/MXFP4 recipes.
Performance Profiling on AMD GPUs - Part 7: AI/PyTorch® Edition
Profile and optimize a PyTorch Evoformer (TinyOpenFold) training workload on AMD Instinct GPUs using rocprofv3, rocprof-compute, and rocprof-sys.
GLM-5.2 AgentX on AMD Instinct MI355X: Boosting Throughput and Interactivity
Learn how cache and kernel optimizations raised GLM-5.2 AgentX throughput and interactivity on AMD Instinct MI355X.
Quark Mixed-Precision Quantization for Wan2.2 Video Generation on AMD Instinct GPUs
Apply AMD Quark mixed-precision MXFP4 quantization to Wan2.2-T2V-A14B video generation with calibration-driven block selection.
FlyDSL Mono Kernels for High-Interactivity LLM Serving in ATOM
FlyDSL mono kernels cut decode latency in ATOM, with MiniMax-M3 and DeepSeek-V4.1 results.
Device Memory Load Issue Stalls and Scheduling Opportunities in Lock-Stepped Kernels on AMD Instinct™ MI300X
Diagnosing and reducing device-memory load issue stalls in lock-stepped kernels on AMD Instinct MI300X by interleaving compute.