Linux kernel builds approach 10-second mark with AI-assisted patches
Recent AI-assisted patches have reduced a defconfig x86-64 Linux kernel build from over 22 seconds to about 15 seconds on a high-end EPYC workstation. Further gains are expected from upcoming hardware such as Zen 6 processors and PCIe Gen 6 storage. That could soon bring clean kernel compilation times below 10 seconds.
The gains come from v4 kernel patches by Lorenzo Stoakes, a Linux memory-management developer, with help from AI/LLM tools. On Michael Larabel's Phoronix test machine—dual 64-core AMD EPYC 9575F CPUs, 128 cores/256 threads, 1.5TB DDR5-6400, Samsung PCIe Gen5 NVMe, Ubuntu 26.04 LTS—a minimal x86-64 defconfig build fell from above 22 seconds to about 15 seconds, without RAMdisk tuning.
A larger allmodconfig build, including thousands of drivers and subsystems, dropped from 169 seconds to 134 seconds on same setup. Phoronix expects future AMD Zen 6 EPYC 9686F systems with MRDIMMs and PCIe Gen6 storage may push defconfig builds below 10 seconds. Kernel compile times have served as a long-running benchmark.
Faster kernel builds could help Linux developers, distribution maintainers, and CI teams test changes more often and shorten release cycles. Reduced compile time may also lower energy use per build and make experimentation easier on capable machines. However, the headline figures rely on costly, high-core workstations, so smaller contributors and organizations may see less dramatic gains. Hardware vendors may market such performance, while the broader open-source ecosystem could benefit from quicker bug detection and iteration.