# geekbench

> Geekbench benchmark workloads: Geekbench: Asset Compression, Geekbench: Background Blur, Geekbench: Clang, Geekbench: File Compression and 13 more.

Geekbench

## Geekbench: Asset Compression {#geekbench-asset_compression}

Compresses 16 texture images and geometry files using ASTC, BC7, DXTC, and Draco.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:asset_compression</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>334 – 78.5K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Asset Compression](/benchmarks/histograms/geekbench-asset_compression.svg)

## Geekbench: Background Blur {#geekbench-background_blur}

Separates and blurs the background of 10 frames in a 1080p video, using DeepLabV3+.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:background_blur</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>219 – 25.6K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Background Blur](/benchmarks/histograms/geekbench-background_blur.svg)

## Geekbench: Clang {#geekbench-clang}

Compiles files (8 in single-core mode, 96 in multi-core mode) of the Lua interpreter using Clang and the musl libc as the C standard library for the compiled files.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:clang</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>402 – 119.5K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Clang](/benchmarks/histograms/geekbench-clang.svg)

## Geekbench: File Compression {#geekbench-file_compression}

Compresses and decompresses the Ruby 3.1.2 source archive (a 75 MB archive with 9841 files) using LZ4 and ZSTD on an in-memory encrypted file system. It also verifies the files using SHA1.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:file_compression</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>298 – 19.8K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: File Compression](/benchmarks/histograms/geekbench-file_compression.svg)

## Geekbench: HDR {#geekbench-hdr}

Blends six 16MP SDR photos to create a single HDR photo, using a recovery process and radiance map construction (Debevec and Malik 1997), and a tone mapping algorithm (Reinhard and Devlin 2005).The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:hdr</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>375 – 34.7K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: HDR](/benchmarks/histograms/geekbench-hdr.svg)

## Geekbench: Horizon Detection {#geekbench-horizon_detection}

Detects and straightens uneven or crooked horizon lines in a 48MP photo to make it look more realistic, using the Canny edge detector and the Hough transform.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:horizon_detection</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>435 – 57.7K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Horizon Detection](/benchmarks/histograms/geekbench-horizon_detection.svg)

## Geekbench: HTML5 Browser {#geekbench-html5_browser}

Opens and renders web pages (8 in single-core mode, 32 in multi-core mode) using a headless web browser.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:html5_browser</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>370 – 47.3K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: HTML5 Browser](/benchmarks/histograms/geekbench-html5_browser.svg)

## Geekbench: Navigation {#geekbench-navigation}

Generates 24 different routes between a sequence of locations on two OpenStreetMap maps (one for a small city, one for a large city) using Dijkstra's algorithm.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:navigation</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>358 – 29K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Navigation](/benchmarks/histograms/geekbench-navigation.svg)

## Geekbench: Object Detection {#geekbench-object_detection}

Detects and classifies objects in 300x300 pixel photos (16 in single-core mode, 64 in multi-core mode) using the MobileNet v1 SSD convolutional neural network.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:object_detection</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>50 – 24.7K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Object Detection](/benchmarks/histograms/geekbench-object_detection.svg)

## Geekbench: Object Remover {#geekbench-object_remover}

Removes an object (using a mask) from a 3MP photo, and fills in the gap left behind using the iterative PatchMatch Inpainting approach (Barnes et al. 2009).The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:object_remover</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>247 – 32.6K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Object Remover](/benchmarks/histograms/geekbench-object_remover.svg)

## Geekbench: PDF Renderer {#geekbench-pdf_renderer}

Opens complex PDF documents (4 in single-core mode, 16 in multi-core mode) of park maps from the American National Park Service (sizes from 897 kB to 1.5 MB) with large vector images, lines and text.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:pdf_renderer</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>364 – 30.2K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: PDF Renderer](/benchmarks/histograms/geekbench-pdf_renderer.svg)

## Geekbench: Photo Filter {#geekbench-photo_filter}

Applies colour and blur filters, level adjustments, cropping, scaling, and image compositing filters to 10 photos that range in size from 3 MP to 15 MP.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:photo_filter</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>246 – 29.5K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 32 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Photo Filter](/benchmarks/histograms/geekbench-photo_filter.svg)

## Geekbench: Photo Library {#geekbench-photo_library}

Categorizes and tags photos (16 in single-core mode, 64 in multi-core mode) based on the objects that they contain. The workload performs JPEG decompression, thumbnail generation, image transformations, image classification (using MobileNet 1.0), and storing data in SQLite.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:photo_library</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>204 – 62.1K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Photo Library](/benchmarks/histograms/geekbench-photo_library.svg)

## Geekbench: Ray Tracer {#geekbench-ray_tracer}

Renders the Blender BMW scene using a custom ray tracer built with the Intel Embree ray tracing library.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:ray_tracer</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>349 – 269.4K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Ray Tracer](/benchmarks/histograms/geekbench-ray_tracer.svg)

## Geekbench: Score {#geekbench-score}

Composite score using the weighted arithmetic mean of the subsection scores, which are computed using the geometric mean of the related scores.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:score</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>334 – 33.1K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark does not scale well on instances with more than 64 vCPUs. Empirical data shows measured scores plateau and may even degrade as vCPU count increases, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Score](/benchmarks/histograms/geekbench-score.svg)

## Geekbench: Structure from Motion {#geekbench-structure_from_motion}

Generates 3D geometry by constructing the coordinates of the points that are visible in nine 2D images of the same scene.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:structure_from_motion</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>240 – 65.8K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Structure from Motion](/benchmarks/histograms/geekbench-structure_from_motion.svg)

## Geekbench: Text Processing {#geekbench-text_processing}

Loads 190 markdown files, parses the contents using regular expressions, stores metadata in a SQLite database, and exports the content to a different format on an in-memory encrypted file system, using a mix of C++ and Python.The score is calibrated against a baseline score of 2,500 (Dell Precision 3460 with a Core i7-12700 processor) as per the Geekbench 6 Benchmark Internals.

Benchmark details<dl><dt>Benchmark ID</dt><dd><code>geekbench:text_processing</code></dd><dt>Unit of measurement</dt><dd>—</dd><dt>Server types evaluated</dt><dd>3,379</dd><dt>Total measurements</dt><dd>6,758</dd><dt>Range of measurements</dt><dd>346 – 4.8K</dd><dt>Better</dt><dd>↑ Higher is better</dd></dl>

Configuration options

| Field | Description | Example values |
| --- | --- | --- |
| `cores` | Single-Core or Multi-Core performance tests. | `multi` `single` |

:::warning

This benchmark uses a fixed workload size and does not scale with vCPU count beyond 4 vCPUs, so it is not recommended for evaluating larger instance sizes.

:::

Score distribution <small>(6,758 measurements)</small>

![Score distribution of Geekbench: Text Processing](/benchmarks/histograms/geekbench-text_processing.svg)
