STAC-M3 Benchmark: L tick database 1.0 on Oxide OxVM w/ 128 vCPUs & 1.25 TiB of RAM

Lobsters Hottest News

Summary

STAC发布了STAC-M3基准测试审计结果:L Labs的L tick database 1.0在Oxide OxVM云虚拟机(128 vCPU、1.25 TiB内存、5块NVMe盘)上运行,在Antuco套件中创下多项纪录,包括最高存储效率(为此前最佳的1.86倍)以及多个查询时间指标领先。

<p>It's using <a href="https://lv1.sh/" rel="ugc">https://lv1.sh/</a></p> <p><a href="https://lobste.rs/s/oh4yfc/stac_m3_benchmark_l_tick_database_1_0_on">Comments</a></p>
Original Article
View Cached Full Text

Cached at: 10/17/26, 06:51 AM

# STAC-M3™ Benchmark Results: L tick database 1.0 on an Oxide OxVM virtual machine with 128 vCPUs and 1.25 TiB of RAM, with 5 x 3.44 TB Oxide Local NVMe Disks (Antuco suite) | STAC - Insight for the Algorithmic Enterprise Source: [https://docs.stacresearch.com/M3-20261005a](https://docs.stacresearch.com/M3-20261005a) [![STAC](https://docs.stacresearch.com/sites/all/themes/stac/logo.png)](https://docs.stacresearch.com/) Type: Audited Specs: STAC\-M3™ Benchmarks \(Antuco Suite\) STAC recently performed STAC\-M3™ benchmark audits on a stack featuring the L tick database 1\.0 from L Labs PBC running on a single Oxide OxVM cloud virtual machine with 128 vCPUs and 1\.25 TiB of RAM, with 5 x 3\.44 TB Oxide Local NVMe disks\. STAC\-M3 is the industry standard enterprise tick\-analytics benchmark for database and hardware stacks that manage large time series of market data\. It provides a vendor\-neutral framework for evaluating the performance, efficiency, and scalability of time\-series databases used in financial analytics and backtesting\. This report covers the baseline suite \(Antuco\), for performance under fixed conditions\. Results for the Shasta suite are available[here](https://docs.stacresearch.com/M3-20261005s)\. In these latest tests, the combination of the L tick database and the Oxide OxVM cloud virtual machine achieved the following: - This solution achieved the highest storage efficiency \(STAC\-M3\.v1\.1\.STORAGE\.EFF\) of any published STAC\-M3 Antuco result, 364%- 1\.86x the efficiency of the previous best result \(SHK211203\) - This solution outperformed all publicly disclosed results for STAC\-M3\.β1\.1T\.YRHIBID\-2\.TIME- 1\.81x faster than the previous best result \(KDB211210\) - Compared to SUTs using the q language, this solution achieved the fastest mean\-response STAC\-M3\.β1\.1T\.NBBO\.TIME- 1\.58x faster than the previous best result \(KDB230707\) - Was faster in 11 of 17 mean\-response time benchmarks than all other publicly disclosed results using a cloud\-based deployment architecture - Outperformed a similar SUT using the q language, NAND flash storage / NVMe SSD, and the same file system \(KDB250929\) in 9 of 17 mean\-response time benchmarks, including:- 8\.4x speed for STAC\-M3\.β1\.10T\.THEOPL\.TIME - Over 1\.7x speedup in STAC\-M3\.β1\.1T\.STATS\-UI\.TIME The Stack Under Test included: - Database- L tick database 1\.0 from L Labs PBC - Database node- 1 x Oxide OxVM cloud virtual machine, with: - 1 x 128 vCPU Oxide Virtual Turin\-like Processor - 1,280 GiB VM memory \(1,259 GiB visible to the OS\) - 5 x 3\.44 TB \(3,200 GiB\) Oxide virtual NVMe disks - Operating system- Ubuntu 24\.04\.4 LTS The report provides detailed performance metrics, methodology, and visualizations for the Antuco suite\. The full benchmark report is available for download below\. Insights Subscribers can also access detailed system configurations and the code used in this project to reproduce the results in their own environments\. To learn more about subscription options, please[contact us](https://docs.stacresearch.com/cdn-cgi/l/email-protection#b1d8dfd7def1e2e5f0f2c3d4c2d4d0c3d2d99fd2dedc)\. ## User login

Similar Articles

@MinLiBuilds: https://x.com/MinLiBuilds/status/2107988055101685781

X AI KOLs Timeline

The author shares a complete guide for deploying Qwen3.8-Flash-Next (125B-A6B) on 12GB VRAM (RTX 5070) with Strata, achieving nearly 100 tok/s, and lets Ant Group's newly released Ling-3.1-flash Agent handle the entire workflow from scratch — machine check, download, installation, troubleshooting, and benchmarking.

@lxfater: With 64GB RAM, the 100-billion-parameter model is already deployed. We've got the hammer — now where's the nail?

X AI KOLs Timeline

A 12GB GPU with 64GB of system RAM is enough to run a 100-billion-parameter large model: the open-source project Strata can run quantized versions of Qwen3.8-Flash-Next — about 94 tokens/s with the Q2_0 quantization and about 53 tokens/s with IQ3_S on an RTX 5070. The author remarks that the hammer is ready; the question is where the nail (i.e., the application scenario) is.

@nicebabycat: https://x.com/nicebabycat/status/2091726637155103126

X AI KOLs Following

This article provides a detailed test of the local deployment and performance of the Ling-3.0-tiny model on an Apple M5 chip Mac, demonstrating the feasibility of running a 7.9B parameter model at 47 tokens per second without a discrete GPU.

Janas: Run LLMs larger than RAM by streaming experts from SSD

Hacker News Top

Janas 是一个开源工具,通过从 SSD 流式加载专家模型(mixture-of-experts),让内存小于模型体积的笔记本也能运行 LLM。在 32GB 内存的笔记本上,Qwen3-Next-80B(48GB)推理速度达 37 tok/s,显著超越 llama.cpp 的 2.8 tok/s。