AI Blogs#
MoE Training Optimization with Primus
Learn how to speed up MoE training on AMD Instinct GPUs with Primus: a fused MoE megakernel, FP8/MXFP8 expert GEMMs, and training-at-scale tuning.
Micro-World 2: Streaming Real-time Interactive World Model
This blog demonstrates how Micro-World 2 combines precisely annotated action data, four-step diffusion distillation, and ROCm-optimized inference to deliver controllable, long-horizon, real-time world generation on a single AMD Instinct GPU through a fully open-source pipeline.
Optimizing World Action Models on AMD GPUs
Speed up World Action Models on AMD GPUs with GEAK kernels, fewer denoising steps, and latent actions, cutting FastWAM latency 25× at about 97 percent success.
hipCIM 26.06: GPU-Accelerated Whole-Slide Image I/O on AMD Instinct™ GPUs
In this blog you will see how hipCIM 26.06 reads whole-slide images on AMD Instinct™ GPUs, and how fast and faithfully it decodes them.
Bringing Ultra-Fast, Interactive LLM Inference to AMD Instinct™ MI355X GPUs
Accelerate single-request LLM inference on AMD Instinct MI355X with megakernels, speculative decoding, and MXFP4 for 1.9-3.3x faster decode.
Introducing rocOpt: GPU-Accelerated Decision Optimization on AMD ROCm
Run large-scale MILP, LP, QP, and vehicle routing problems on AMD Instinct GPUs - introducing rocopt, optimization built for ROCm.
Long Context, One Node: Serving GLM-5.3-FP8 on MI300X with DP8+EP8
Learn how to serve GLM-5.3-FP8 on one AMD Instinct™ MI300X node with DP8+EP8 and vLLM v0.28 for long-context workloads, and compare it with TP=8.
AMD Instinct MI455X vs MI355X: A Technical Look at the Advancing AI 2026 Inference Numbers
Explore the MI455X vs MI355X inference numbers from Advancing AI 2026, and learn how each benchmark was measured and what it really means.
DI Series: Serving Kimi-K3 MXFP4 Wide-EP Disaggregated on AMD Instinct MI300X and MI325X
Learn how to serve a 1.4 TiB Kimi-K3 MXFP4 mixture-of-experts model disaggregated across AMD Instinct MI300X and MI325X GPUs.
Pre-Training Hybrid LLMs from Scratch in Primus on AMD Instinct™ GPUs
Pre-train hybrid LLMs mixing MLA attention with Gated DeltaNet, Kimi Delta Attention and Mamba2 mixers on AMD Instinct MI355X GPUs, with open recipes.
Completing the GPU Performance Picture: Understanding TAF Alongside Peak FLOPs and MAF
Learn how Typical Attained FLOPs complements Peak FLOPs and Max-Achievable FLOPs, including the methodology and MI325X results
ROCm 10.1: Breaking the Data-Movement Bottleneck
ROCm 10.1 targets the data-movement bottleneck with AMD Infinity Storage, NUMA-aware memory, and a modernized stack.