{
  "id": 10758314,
  "title": "AMD Boosting AI/LLM Performance For Radeon iGPUs As Much As 18~23% With Linux 7.4",
  "url": "https://urgent.news/2026/09/29/amd-boosting-ai-llm-performance-for-radeon-igpus-as-much-as-18-23",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-29T18:57:00.000Z",
  "source": {
    "name": "Phoronix",
    "slug": "phoronix",
    "url": "https://www.phoronix.com/review/amd-perfopt"
  },
  "original_language": "en",
  "account": "AMD is enhancing AI/LLM performance for Radeon integrated graphics processors (iGPUs) by up to 18-23% in Linux 7.4. The new PerfOpt feature, introduced in this release, proves particularly beneficial for lower-end hardware. AMD's IOMMU Performance Optimization, part of the AMD I/O Virtualization Technology specification, bypasses the IOMMU when accessing system memory, eliminating associated overhead. This feature is enabled by default for AMD integrated graphics when using Linux 7.4. To test the impact of PerfOpt, a test kernel was built from the IOMMU Git tree, comparing its performance with the feature disabled on an AMD Strix Halo desktop. The Lemonade local AI server with Llama.cpp back-end was utilized for local AI/LLM inferencing. More benchmarks on Linux 7.4's overall performance will follow as the kernel cycle begins in October and the stable kernel release approaches the end of the year, potentially becoming the year's long-term support (LTS) kernel version.",
  "summary": "If you have an AMD Ryzen laptop/desktop and looking to leverage AI/LLM capabilities with the integrated graphics, Linux 7.4 is going to be a real treat especially for lower-end hardware. AMD's new PerfOpt feature being introduced in Linux 7.4 is set to provide some solid gains as shown in these benchmarks today across different AMD Ryzen/Radeon hardware.",
  "key_points": [
    "AMD boosts AI/LLM performance for Radeon iGPUs by 18-23% in Linux 7.4",
    "PerfOpt feature bypasses IOMMU for system memory access, reducing overhead",
    "Test shows PerfOpt improves local AI/LLM inferencing on AMD Strix Halo desktop"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}