Benchmarking DuckDB v2.0-dev on NVIDIA Vera

DuckDB Labs benchmarked a preview build of DuckDB v2.0 (1.6.0.dev379) on NVIDIA's new Vera CPU, which posted a median TPC-H composite score of 3,131,306 QphH@Size at scale factor 1,000 versus 2,120,588.82 for an Intel Xeon 6 x86 baseline. The Vera system ran DuckDB's unmodified arm64 build on a single 88-core socket with Spatial Multithreading and up to 1.2 TB/s of memory bandwidth, while the Xeon 6 baseline used 96 of its 192 threads. Both machines ran Ubuntu 26.04 with 768 GB of memory and a single 3+ TB NVMe SSD, and the benchmark used the duckdb-tpch harness with the full TPC-H workload including update streams.

Benchmarking DuckDB v2.0-dev on NVIDIA Vera TL;DR: We put NVIDIA's new Vera CPU through its paces using the full TPC-H workload on scale factor 1,000. We love running benchmarks on all sorts of new hardware. This often includes somewhat https://duckdb.org/everywhere/raspberry-pi-5/ exotic https://duckdb.org/everywhere/iphone-17-pro/ hardware https://duckdb.org/everywhere/morefine-m700s/ – where we sometimes need to compile a new DuckDB binary. This time, in collaboration with the NVIDIA team, we got our hands on their new Vera chip, aimed at agentic workloads. NVIDIA Vera https://www.nvidia.com/en-us/data-center/vera-cpu/ comes with 88 “Olympus” cores, Spatial Multithreading and up to 1.2 terabytes per second TB/s of memory bandwidth. What's more, Vera runs arm64 binaries – so it runs DuckDB's regular build without any modifications. To see how this new hardware performs, we ran the full TPC-H benchmark not just the read queries on scale factor 1,000. Our setup included two server platforms: an x86 baseline Intel Xeon 6 and NVIDIA Vera. Environment environment System Setup system-setup Both machines ran Ubuntu 26.04. Their hardware included 768 GB of memory and a single NVMe SSD with 3+ TB of storage, formatted as ext4. At the time of writing, DuckDB v2.0 was not released yet https://duckdb.org/2026/09/02/try-duckdb-20-alpha.html , so we used a preview build https://duckdb.org/install/?version=preview , 1.6.0.dev379 . To run the experiments, we used the duckdb-tpch benchmark harness https://github.com/duckdb/duckdb-tpch/ . To re-create our benchmark environment, see our script below. Click here to reveal the setup script. make sure you have the required tools sudo apt install git g++ cmake python3-pip clone the repository at the commit used for the experiments git clone https://github.com/duckdb/duckdb-tpch cd duckdb-tpch git reset --hard cf096cea8a0a054cefe0e695c842ca5bca6ec04d install the required Python dependencies python3 -m venv venv source venv/bin/activate python3 -m pip install -r requirements.txt python3 -m pip install duckdb==1.6.0.dev379 build the dbgen tool cd tpch tools 3.0.1/dbgen/ make cd ../.. generate the dataset, query streams and update streams export SF=1000 ./generate.sh Tip To generate the large-scale TPC-H dataset efficiently, the repository uses the wonderful tpchgen-cli tool https://pypi.org/project/tpchgen-cli/ , a parallel Rust rewrite of the TPC-H data generator. It also uses the old generator to produce update and query streams. This workload is relatively large-scale SF1,000 and highly concurrent during the throughput run. To get the best performance out of DuckDB, we set the following flags: SET block allocator memory = '500G'; SET allocator background threads = true; x86 Baseline: Intel Xeon 6 x86-baseline-intel-xeon-6 The x86 baseline setup used an Intel Xeon 6 CPU. The CPU had 96 cores with hyper-threading https://en.wikipedia.org/wiki/Hyper-threading , yielding 192 threads. However, we found that DuckDB performs best when using 96 threads, so we configured it accordingly: SET threads = 96; NVIDIA Vera nvidia-vera The second machine had the NVIDIA Vera CPU, which packs 88 cores with Spatial Multithreading https://developer.nvidia.com/blog/inside-nvidia-vera-cpu-olympus-cores-built-for-maximum-single-threaded-performance-in-agentic-ai/ . We conducted the benchmark on a dual-socket system but restricted DuckDB to a single socket using the numactl command: numactl --cpunodebind=0 --membind=0 python3 benchmark.py We also pinned the number of threads: SET threads = 88; Benchmark Runs and Results benchmark-runs-and-results We performed an initial warmup run and discarded its result. Then, we conducted three complete runs in sequence and took the median value of the QphH@Size composite scores: | | x86 QphH@Size | Vera QphH@Size | |---|---|---| | Run 1 | 2,120,588.82 | 3,073,292.99 | | Run 2 | 2,022,795.01 | 3,131,306.27 | | Run 3 | 2,109,782.19 | 3,052,102.87 | | Median score | 2,109,782.19 | 3,073,292.99 | The results show that the Vera platform has a 46% performance advantage on TPC-H SF1,000 over the x86 baseline Intel Xeon 6 . In fact, if we were to audit these results through the official TPC-H channels, both would stand a chance to be in the top 3 among the TPC-H Top Performance Results https://www.tpc.org/tpch/results/tpch perf results5.asp?resulttype=noncluster&version=4 on SF1,000. We attribute the performance advantage of Vera to three key factors. First, the high memory bandwidth enables fast scans. Second, combined with low memory latency, it keeps the cores continuously occupied during processing. Finally, the Olympus cores' large caches allow efficient access to the hash tables built during aggregations and joins. We are looking forward to continuing this benchmark by using DuckDB v2.0.0's stable build, running on the full dual-socket setup and incorporating different benchmarks.