May 29, 2026
Cache me outside, benchmark war
Cache Aware Scheduling Shows Nice Wins for AMD Zen 5 on PostgreSQL, Valkey
Linux may get a speed boost, but the comments are already fighting over whether the test is legit
TLDR: A coming Linux 7.2 feature appears to make some AMD systems faster in database and network tasks by keeping related work closer together on the chip. Commenters are split between excitement over easy speed gains and skepticism that the benchmark setup reflects real-world server use.
Linux fans are eyeing a potentially juicy upgrade: a new feature called Cache Aware Scheduling could land in Linux 7.2 and it seems to help some workloads on AMD’s latest chips, especially PostgreSQL databases, Valkey-style caching jobs, web serving, and network tests. In plain English, it tries to keep related work closer together inside the processor so less time is wasted fetching data. The benchmark numbers looked promising on a monster AMD Threadripper machine, and that was enough to get the crowd buzzing.
But the real show, of course, was in the replies. One camp basically said, “Nice, but what happens in the real world?” User fabian2k came in with classic comment-section side-eye, questioning whether the PostgreSQL setup was even realistic and pointing out that the gains only really showed up under very heavy connection counts. Translation: some readers think this is exciting, others think it smells a little like “lab win, production maybe.”
Then came the practical crowd. ysleepy immediately wanted to know how this plays with virtual machines and containers — because if the boost disappears once you’re running cloud-style setups, the hype changes fast. That turned the thread into a mini drama between benchmark believers and deployment skeptics. The vibe was less “wow, free speed!” and more “show us the receipts.” It’s peak Linux community energy: celebrate the patch, then interrogate every last setting like it’s a courtroom drama.
Key Points
- •Phoronix says Cache Aware Scheduling is close to being merged and is tentatively expected for Linux 7.2.
- •The feature aims to improve performance on CPUs with multiple cache domains by keeping related tasks within the same last level cache domain.
- •Benchmarking was conducted on a System76 Thelio Major with an AMD Ryzen Threadripper 9980X, 128GB of DDR5, a 1TB NVMe SSD, Radeon AI PRO R9700 graphics, Ubuntu 26.04, and a custom Linux 7.1-rc2-based kernel.
- •The article reports lower latency and higher throughput in Microsoft Ethr, slight gains in Stargate, and stronger PostgreSQL benefits under higher client concurrency.
- •The article says measurable gains were also seen in TigerBeetle, Nginx, and some other workloads, while CPU AI workloads, Blender, and other common workstation tasks showed no noticeable improvement.