benchmarking Archives

Tech Optimizer

January 8, 2026

Faster Is Not Always Better: Choosing the Right PostgreSQL Insert Strategy in Python (+Benchmarks)

Inserting 2 million records per second into Postgres is achievable. The analysis explores five methods for inserting data into Postgres using Python, focusing on trade-offs in abstraction, safety, convenience, and performance rather than just speed. High-volume insert workloads are common in scenarios like loading records, syncing data, backfilling analytics tables, and ingesting events. Minor inefficiencies can lead to significant performance impacts. To interact with Postgres, the psycopg3 driver is used alongside SQLAlchemy, which provides two layers: Core and ORM. Psycopg3 is a low-level driver requiring manual SQL management, while SQLAlchemy Core offers a SQL abstraction, and the ORM maps Python classes to database tables, enhancing productivity but introducing overhead. Benchmarking involves measuring only the time spent transferring data from Python to Postgres, ensuring a fair comparison among methods. The fastest method may not always be the best due to maintenance costs, correctness guarantees, and cognitive load. The right insertion strategy depends on the existing data structure rather than just row count. The ORM is suited for CRUD-heavy applications, Core for data ingestion and analytics, and the Driver for maximum throughput in extensive writes. Performance issues can arise from mismatching abstractions, and reverting to a lower level may enhance performance. A guideline for choosing methods is: - Use ORM for applications prioritizing correctness and productivity. - Use Core for data movement or transformation balancing safety and speed. - Use Driver for pushing performance limits with raw power and full responsibility.

Tech Optimizer

January 6, 2026

How to speed up mass data inserts in PostgreSQL when using Spring

A common task in enterprise systems involves loading large volumes of data into PostgreSQL, which can be slow if using a loop in Java that calls the save() method for each record. Initial tests show that a basic insert operation takes over 10 minutes. Hibernate retains every object in the session-level cache until the transaction concludes, leading to high memory consumption. Enabling batching with the jdbc.batch_size setting improves performance by about 23%, but high memory usage persists. Application-side caching of IDs is faster than database-side caching, achieving a 30% performance gain. Using saveAll() results in minimal improvement as it still calls save() in a loop. Enabling reWriteBatchedInserts in the JDBC driver allows for multi-row inserts, reducing operation time to 4 minutes and 37 seconds. Clearing the session-level cache with EntityManager further optimizes memory usage, resulting in an 18-second time gain. Setting order_inserts=true enhances batching efficiency. With all optimizations, insertion time is reduced to 4 minutes and 19 seconds. A custom data insertion layer can be created to generate INSERT statements, with reflection being a straightforward but not optimal method. Prepared statements improve performance and security, with an efficient batch size of 5,000 records. PostgreSQL's COPY command facilitates faster bulk data transfers than multi-row inserts. Integrating COPY with Spring transactions enhances performance, and parallelizing the insertion process with a ThreadPoolExecutor significantly improves speed and memory usage.

Winsage

January 5, 2026

Windows 11 finishes an embarrassing last in speed test of six Windows generations — but there’s a big catch

A YouTuber named TrigrZolt conducted tests comparing six versions of Windows (XP, Vista, 7, 8.1, 10, and 11) using six Lenovo ThinkPad X220 notebooks with Intel Core i5-2520M CPUs, 8GB of RAM, and 256GB HDDs, which do not meet the minimum requirements for Windows 11. Windows 11 ranked last in several metrics, including boot-up speed, memory consumption, and battery life. It was slower than all other versions in opening applications, particularly Paint and File Explorer. However, it ranked fourth in storage efficiency and performed reasonably well in one web-page loading test. The testing conditions, including outdated hardware and the choice of benchmarking applications, may have skewed the results against Windows 11. Users expressed concerns about Windows 11 being bloated and criticized its performance issues.

Winsage

January 3, 2026

Benchmarking Windows Against Itself, From Windows XP To Windows 11

TrigrZolt conducted benchmarks on multiple Windows versions (XP, Vista, 7, 8.1, 10, and 11) using a Lenovo ThinkPad X220. Windows 8.1 had the fastest startup time due to its Fast Boot feature, while Windows 11 was notably slow, struggling to load the taskbar. Windows XP had the smallest installation size and least RAM usage at 800 MB, compared to Windows 11's 3.3 GB. In memory management tests with the Supermium browser, Windows XP and Windows 11 performed similarly under heavy usage, but Windows 7 and 8.1 managed to keep over 200 tabs open before hitting the 5 GB RAM limit. Windows 11 also performed poorly in battery life, video rendering, and application launch times, with users experiencing significant delays. The poor performance of Windows 11 is attributed to extensive code rewrites since Windows 7 and the default storage encryption with BitLocker, which slows performance. Microsoft has implemented workarounds to improve File Explorer's responsiveness. Overall, the trend indicates that increasing resource demands can lead to inefficiencies in software performance.

AppWizard

December 31, 2025

Why Modern PC Games Are CPU-Bound Again

Modern PC gaming faces a challenge where gamers experience underutilized GPUs at around 60% capacity, leading to faltering frame rates despite investing in high-end graphics cards. This issue arises not from CPU inadequacy but from gaming demands evolving faster than how games utilize CPU power. Players upgrading to more powerful GPUs like the RTX 4080 or RX 7900 XT report minimal frame rate improvements, particularly at 1080p and 1440p resolutions, with CPU cores being pushed to their limits while GPU usage declines. This trend is evident across various game genres, and even DirectX 12, which offers improved multithreading, struggles to scale effectively beyond six to eight cores. Developers recognize these CPU limitations but find it challenging to address them. Modern games simulate numerous elements continuously, rely on real-time asset streaming, and face difficulties in distributing workloads effectively across CPU cores. Additionally, console hardware influences game design, often leading to CPU-bound scenarios on high-end PCs. Performance varies between Intel and AMD CPUs depending on game design, with some games demonstrating effective CPU scaling while others do not. CPU bottlenecks are expected to persist as GPU advancements outpace CPU improvements, emphasizing the need for gamers to consider CPU capabilities alongside GPU upgrades.

Winsage

December 31, 2025

Unexpected Surprise: Windows 11 Outperforming Linux On An Intel Arrow Lake H Laptop

Recent benchmarks on the Lenovo ThinkPad P1 Gen 8 show that Microsoft Windows 11 outperformed Ubuntu Linux in several demanding workloads, including multi-threaded tasks and CPU-based rendering applications like Blender and V-RAY. The device features an Intel Core Ultra 7 255H processor with 16 cores and was tested with 64GB of LPDDR5-7467 memory and NVIDIA RTX Pro 1000 graphics. The performance metrics were consistent with expectations for the hardware, and the benchmarks used official binaries for both operating systems. Lenovo and Intel are investigating the results further, and the findings may indicate a broader trend in operating system performance. The original benchmarks were conducted using Windows 11 Pro and Ubuntu 24.04.3 LTS.

AppWizard

December 27, 2025

How to Properly Benchmark your PC Games Using CapFrameX

Average Frames Per Second (FPS) is a common metric in PC gaming performance, but consistency in gameplay experience is equally important, highlighted by metrics such as 1% low and 0.1% low average FPS. CapFrameX is a tool used for capturing and analyzing detailed performance data, including frametimes, which provide a more accurate assessment of gaming performance than basic FPS counters. Key metrics generated by CapFrameX include: - Average FPS: Overall framerate averaged across the capture session. - 1% low average FPS: Average of the worst 1% of framerates, indicating sustained performance. - 0.1% low average FPS: Average of the worst 0.1% of framerates, highlighting rare but significant performance dips. To ensure accurate benchmarking results with CapFrameX, users should update their UEFI BIOS, operating system, and drivers, clear unnecessary applications, configure power settings for maximum performance, monitor temperatures, and conduct multiple benchmark runs under consistent conditions. The setup process for CapFrameX includes downloading the software, configuring capture options, and verifying the setup through test captures. After capturing benchmark runs, users can analyze the data using the Analysis and Comparison tabs in CapFrameX to evaluate performance metrics and identify potential issues. Common pitfalls in benchmarking include inconsistent scenes, overlooking frametime variance, and not preparing the system properly.

Tech Optimizer

December 22, 2025

Biscuit PostgreSQL Extension Accelerates LIKE Queries with Roaring Bitmaps

Biscuit is an open-source project developed by CrystallineCore, available on GitHub, that introduces a specialized index access method for PostgreSQL to enhance query speed for pattern matching in LIKE queries. It addresses limitations of traditional indexing methods like B-tree and GIN indexes, particularly in complex LIKE or ILIKE operations and multi-column searches. Biscuit uses in-memory bitmap structures to reduce query times by precomputing bitmap representations of string patterns, allowing for rapid filtering without scanning entire rows. The project has a vibrant community contributing to its optimization and performance improvements, especially for ARM architectures. Independent tests show that Biscuit can outperform PostgreSQL's pg_trgm by factors of 5-10x on wildcard-heavy workloads. Real-world applications include significant reductions in query latencies for e-commerce and log processing pipelines, with a reported 40% reduction in ETL job times for error pattern detection in server logs. Biscuit faces challenges such as the initial time required to build indexes on large tables and limited compatibility with older PostgreSQL versions, but the maintainers are actively addressing these issues. The design of Biscuit encourages integration with other open-source tools, and community feedback is shaping its future development.

BetaBeacon

December 19, 2025

I tried to melt my OnePlus 15 with some real games

The OnePlus 15 was tested for gaming performance using games like Call of Duty: Mobile and Genshin Impact. The phone performed well under pressure, dissipating heat effectively. Emulation tests for PS3 and Windows games were also conducted, with the phone handling them smoothly. The phone did not overheat during testing, indicating that OnePlus has fixed previous heating issues. The reviewer recommends the OnePlus 15 for potential buyers due to its stable performance and recent fixes for heating problems.

Tech Optimizer

December 18, 2025

Improve Aurora PostgreSQL throughput by up to 165% and price-performance ratio by up to 120% using Optimized Reads on AWS Graviton4-based R8gd instances

AWS has introduced Graviton4-based R8gd instances for Amazon Aurora, which offer significant performance enhancements over Graviton2-based db.r6g instances. Users can expect up to 165% higher throughput, a 120% improvement in price-performance ratio, and an 80% enhancement in application response time. The integration of an Optimized Reads-enabled tiered cache allows for expanded database caching capacity, minimizing network storage access and improving query response times. Organizations like Mindbody and Claroty have reported substantial performance improvements and cost savings after implementing these features. Aurora PostgreSQL 17 has introduced smarter storage batching algorithms and optimized writes, further enhancing performance for I/O-intensive applications. The db.r8gd instances allow for scaling up to 48xl, providing 192 vCPUs, 50 Gigabits of network bandwidth, and 10.4 TiB of local NVMe capacity. The benchmark testing using HammerDB demonstrated significant improvements in throughput, price-performance ratio, and response times when upgrading from db.r6g to db.r8gd instances. The Aurora Optimized Reads-enabled tiered cache is available for Aurora PostgreSQL versions 17.4 and higher, among others.