TAtanelpoder·Jan 24, 2026Submission
Modern Memory Bandwidth and Latency Benchmarks
1 points · Open discussionSpareCores
sc-membench
C2026–2026
SpareCores / sc-membench
sc-membench
Modern Memory Bandwidth and Latency Benchmarks
2Distinct people
2Mentions
0Explicit recommenders
19GitHub stars
What to know before exploring
What it is
Modern Memory Bandwidth and Latency Benchmarks
GitHub
Where it is used
No documented product use has been verified yet.
What people are saying
No recommendation phrase was found in the indexed comments.
No classified warning is attached to this page. Read the discussions before treating silence as approval.
Also in these discussions
No other repository is repeatedly linked in the same discussions.
This week on Hacker News
Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 5:04:57 PM UTC.
No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.
Mentions over time
HN mentions by 6-hour interval2 dated mentions
Observed mentions: 2026-01-19 16:02:14 UTC to 2026-01-24 21:08:41 UTC.
Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.
View mention counts and date ranges
| Period | Start (UTC) | End (exclusive, UTC) | Mentions | Submissions | Comments |
|---|---|---|---|---|---|
| Jan 19, 12:00 | 2026-01-19 12:00:00 UTC | 2026-01-19 18:00:00 UTC | 1 | 1 | 0 |
| Jan 19, 18:00 | 2026-01-19 18:00:00 UTC | 2026-01-20 00:00:00 UTC | 0 | 0 | 0 |
| Jan 20, 00:00 | 2026-01-20 00:00:00 UTC | 2026-01-20 06:00:00 UTC | 0 | 0 | 0 |
| Jan 20, 06:00 | 2026-01-20 06:00:00 UTC | 2026-01-20 12:00:00 UTC | 0 | 0 | 0 |
| Jan 20, 12:00 | 2026-01-20 12:00:00 UTC | 2026-01-20 18:00:00 UTC | 0 | 0 | 0 |
| Jan 20, 18:00 | 2026-01-20 18:00:00 UTC | 2026-01-21 00:00:00 UTC | 0 | 0 | 0 |
| Jan 21, 00:00 | 2026-01-21 00:00:00 UTC | 2026-01-21 06:00:00 UTC | 0 | 0 | 0 |
| Jan 21, 06:00 | 2026-01-21 06:00:00 UTC | 2026-01-21 12:00:00 UTC | 0 | 0 | 0 |
| Jan 21, 12:00 | 2026-01-21 12:00:00 UTC | 2026-01-21 18:00:00 UTC | 0 | 0 | 0 |
| Jan 21, 18:00 | 2026-01-21 18:00:00 UTC | 2026-01-22 00:00:00 UTC | 0 | 0 | 0 |
| Jan 22, 00:00 | 2026-01-22 00:00:00 UTC | 2026-01-22 06:00:00 UTC | 0 | 0 | 0 |
| Jan 22, 06:00 | 2026-01-22 06:00:00 UTC | 2026-01-22 12:00:00 UTC | 0 | 0 | 0 |
| Jan 22, 12:00 | 2026-01-22 12:00:00 UTC | 2026-01-22 18:00:00 UTC | 0 | 0 | 0 |
| Jan 22, 18:00 | 2026-01-22 18:00:00 UTC | 2026-01-23 00:00:00 UTC | 0 | 0 | 0 |
| Jan 23, 00:00 | 2026-01-23 00:00:00 UTC | 2026-01-23 06:00:00 UTC | 0 | 0 | 0 |
| Jan 23, 06:00 | 2026-01-23 06:00:00 UTC | 2026-01-23 12:00:00 UTC | 0 | 0 | 0 |
| Jan 23, 12:00 | 2026-01-23 12:00:00 UTC | 2026-01-23 18:00:00 UTC | 0 | 0 | 0 |
| Jan 23, 18:00 | 2026-01-23 18:00:00 UTC | 2026-01-24 00:00:00 UTC | 0 | 0 | 0 |
| Jan 24, 00:00 | 2026-01-24 00:00:00 UTC | 2026-01-24 06:00:00 UTC | 0 | 0 | 0 |
| Jan 24, 06:00 | 2026-01-24 06:00:00 UTC | 2026-01-24 12:00:00 UTC | 0 | 0 | 0 |
| Jan 24, 12:00 | 2026-01-24 12:00:00 UTC | 2026-01-24 18:00:00 UTC | 0 | 0 | 0 |
| Jan 24, 18:00 | 2026-01-24 18:00:00 UTC | 2026-01-25 00:00:00 UTC | 1 | 1 | 0 |
What Hacker News says
2 sourcesShow HN: sc-membench for modern memory bandwidth and latency benchmarks
Hey HN, I'd love to get your feedback on `sc-membench`, a modern, multi-platform, open-source memory bandwidth and latency benchmarking tool designed to work out of the box without manual tuning:
- Comprehensive tests across read, write, copy, and latency (pointer chasing) workloads
- Multi-platform (e.g. evaluated on multiple x86 and arm64 machines), portable (e.g. tested on BSD), and easy to run via Docker
- Efficient multi-threaded measurements via OpenMP
- Optimal thread placement and memory allocation for NUMA systems
- Adaptive test sizes based on CPU cache amounts
- Automatic Transparent Huge Pages
Disclaimer: Yes, used LLMs both for coding and documentation updates -- but we validated the results carefully against existing tools and their shortcomings, the results look super promising so far, already got some encouraging early fee
Read the rest
dback from our direct network, so it's time to ask for scrutiny from the wider community /o\
Motivation: We previously benchmarked 3,000+ cloud server types using `bw_mem` from LMbench, but the results were not always consistent with the detected L1/L2/L3 cache sizes. Debugging identified both cache detection issues (mostly relying on `lscpu`, investigating `lstopo` now), and limitations of `bw_mem` as well, e.g. unexpected slowdowns on servers with 100+ vCPUs. See more details in the "Comparison with lmbench" section of the README.
Why does your feedback matter? We plan to run this across ~5,000 cloud server types, so I'd highly appreciate your feedback on methodology, implementation correctness, example results, and any missing cases before burning through a lot of precious cloud credits :)
Unfortunately, we don't have the resources to implement further metrics (e.g. prefetch-to-load distance, load buffer slots, atomic memory operation latency) in this batch, but I'd be happy to take related ideas on the roadmap.