AMD CEO Dr. Lisa Su addressed concerns that the rise of agentic AI might cannibalize the GPU market, arguing instead that the demand is largely additive. While GPUs are essential for running foundational models, CPUs play a critical role in orchestration, data movement, and parallel execution required by autonomous agents. This shift could fundamentally change industry-standard CPU-to-GPU ratios, potentially moving from traditional 1:8 configurations toward a more balanced 1:1 ratio as agentic workloads expand.
AMD now supports Google’s Gemma 4 models (2B–31B parameters) across its entire hardware lineup, including Instinct GPUs (datacenters), Radeon GPUs (workstations), and Ryzen AI processors (PCs). The integration is compatible with vLLM, SGLang, llama.cpp, Ollama, and Lemonade Server, aiming to optimize AI performance for both cloud and local deployment.
AMD’s new Strix Halo mobile platform pushes the limits of integrated GPU
design by rethinking the memory subsystem. The article dives into how the
chip balances low‑latency CPU workloads with the bandwidth‑heavy GPU
demands, detailing the architecture of its Infinity Cache, L1/L2 caches,
and LPDDR5X memory. It covers cross‑CCX latency, zero‑copy behavior,
and the impact of shared memory controllers on contention. By
comparing Strix Halo’s performance to desktop and discrete GPUs,
the author illustrates the trade‑offs that enable a powerful iGPU in a
tablet‑form‑factor device like the ROG Flow Z13.
design by rethinking the memory subsystem. The article dives into how the
chip balances low‑latency CPU workloads with the bandwidth‑heavy GPU
demands, detailing the architecture of its Infinity Cache, L1/L2 caches,
and LPDDR5X memory. It covers cross‑CCX latency, zero‑copy behavior,
and the impact of shared memory controllers on contention. By
comparing Strix Halo’s performance to desktop and discrete GPUs,
the author illustrates the trade‑offs that enable a powerful iGPU in a
tablet‑form‑factor device like the ROG Flow Z13.
This article details how to enable full eGPU acceleration on the Raspberry Pi 500+ with a 15-line patch, resulting in significant performance improvements for gaming and local LLMs. It covers the necessary hardware, software modifications, and benchmark results.
Unveiled at CES 2025, the Minisforum AI X1 Pro is a mini PC that offers three 4TB M.2 SSD slots, up to 96GB of DDR5 memory, and features like a dedicated Windows Copilot button. It runs on an AMD Ryzen AI 9 HX 370 processor and supports an eGPU via an OCuLink port. Its compact size includes an integrated PSU and offers a variety of ports.
