Cache Aware Scheduling Shows Nice Wins for AMD Zen 5 on PostgreSQL, Valkey
rbanffy
26 points
4 comments
May 29, 2026
Related Discussions
Found 5 related stories in 77.5ms across 8,861 title embeddings via pgvector HNSW
- Better JIT for Postgres vladich · 143 pts · March 04, 2026 · 53% similar
- AMD Ryzen AI NPUs Are Finally Useful Under Linux for Running LLMs mikece · 27 pts · March 11, 2026 · 52% similar
- AMD's Ryzen 9 9950X3D2 Dual Edition crams 208MB of cache into a single chip zdw · 91 pts · March 28, 2026 · 51% similar
- AWS engineer reports PostgreSQL perf halved by Linux 7.0, fix may not be easy crcastle · 203 pts · April 05, 2026 · 51% similar
- Linux 6.6 LTS To Linux 7.1 Bechmarks: Performance Up 13% Threadripper Over 3 yrs Bender · 29 pts · May 18, 2026 · 50% similar
Discussion Highlights (3 comments)
ysleepy
I wonder how this behaves with VMs, does this help when the host uses this scheduling, but the VM not? Does both make sense? - the vm threads are probably not pinned to cores. In an case, a great boost for containers on bare metal or maybe core-pinned VMs.
fabian2k
I'm always a bit skeptical about benchmarks like this. There's quite a few knobs you want to adjust for a production PostgreSQL on bigger hardware. And the article doesn't provide any configuration details (like shared buffers, is huge pages enabled). The effect is visible at 500 connections, not at 250. The CPU is a Threadripper with 64 cores and 128 GB memory, which seems like a bit of a mismatch to me. So I wonder how transferable these benchmarks are to different setups.
jauntywundrkind
Beautiful generational leap here, so so so good! I'm also excited to see what folks get up to with sched_ext , this year & ongoingly. Scheduling becoming BPF programmable hopefully shows some other neat wins too.