Software Development Banner Showing Programming Code on a Monitor

Puget Bench Development Program Resources

This page is intended as a one-stop location for information related to the Puget Bench Development Program. If you have any questions, please email your Puget Bench representative or [email protected]

Quickly Jump To: Latest Dev Builds • Common Resources • Puget Bench Roadmap • Offline Activation • Result Filtering • GitHub Access

Latest Public Release

Puget Bench for Creators 1.4.50 – download here

Latest Dev Builds

Updated 10/28/2026

Application

Puget Bench for Creators N/A

Installers
  • Windows (x68)
  • Windows (ARM)
  • macOS
Update Notes (compared to current public release)
  • N/A
Known Issues
  • N/A

Benchmarks

Puget Bench for DaVinci Resolve 3.0.0-dev.3
Benchmark Scripts

Place in the “com.puget.benchmark\benchmarks” folder

Assets

Place in the “com.puget.benchmark\assets” folder

  • Assets automatically downloaded by the benchmark launcher
Update Notes
  • From 3.0.0-dev.2
    • Added Photography Tests
      • Added new Photography tests using API functionality introduced in DaVinci Resolve Studio 21.1.2/
      • Added JPG and TIFF export tests using 100 photos from Canon, Nikon, Sony, and Panasonic cameras.
      • Added an AI UltraNR test to evaluate AI-powered noise reduction performance.
      • Added an Open FX test combining Adjustments, Film Look Creator, Dead Pixel Fixer, Sharpen, Glow, and Noise Reduction (Better).
      • Added a Fusion effects test (available only through the Custom preset). This test is not included in the normal presets due to VRAM issues that causes Resolve to crash. This issue has been reported to Blackmagic.
    • Updated Benchmark Categories and Scoring
      • Reworked test categories in preparation for a planned overhaul of the benchmark presets, moving away from the current Basic, Standard, and Extended structure toward workflow-specific presets.
      • Split AI tests into the following categories: AI Editing Effects, AI Visual Effects, AI Audio, and AI Media Intelligence.
      • Split GPU Effects tests into the following categories: Editing Effects, Visual Effects.
      • Added placeholder scoring modifiers for the new category scores.
      • Category scores currently contribute equally to the primary scores, though weighting may be introduced in the future.
    • Benchmark Preset Changes (WIP)
      • Began transitioning toward workflow-specific presets: Baseline, Video Editing, VFX, AI, Photography, and Comprehensive.
      • All presets will include baseline tests to allow for consistent CPU and GPU Baseline scores across all presets.
      • The new preset structure is still in development and is not yet implemented in this build.
      • Current presets currently work as follows:
      • Basic: Runs the tests used for the CPU and GPU Baseline scores.
      • Standard: Runs all tests except those requiring more than 12GB of VRAM or additional package downloads.
      • Extended: Runs all tests.
      • See Dr_Preset_Overhaul_Results.xlsx for the planned test assignments for each preset.
  • From 3.0.0-dev.1
    • The DaVinci Resolve Studio 21.1.1 update made several AI Runtime APIs asynchronous, causing some tests to fail because the benchmark assumed they were synchronous. Runtime tests now poll for a speed result instead of querying the speed function only once.
  • From 0.0.9-dev
    • Trialing new GPU Effects tests
      • Optical Flow – 50% Standard Faster
      • Film Look Creator
    • Trialing new AI Render tests
      • Voice Isolation (10 Tracks)
      • Blemish Removal
      • Beauty
      • Face Reshaper
      • Face Age Transformer
      • CineFocus (Faster)
      • UltraSharpen
      • Speed Warp 25% (Faster)
      • Speed Warp 25% (Better)
      • UltraNR v2
    • Trialing new AI Runtime Tests
      • Audio Transcription (Speaker Detection)
      • Audio Classification
      • Motion Deblur
      • Speech Generator*
      • IntelliSearch (Faster)*
      • IntelliSearch (Better)*
      • Slate Analysis*
    • * tests require additional packages to be manually downloaded through the “Extras Download Manager”
Known Issues
  • Many of the new AI tests take SIGNIFICANT VRAM and can’t be run on all systems

Sample Results

These results are included to give you an idea of the expected performance with various hardware. Our main Puget Bench QA systems are:

  • Intel Core i9 13900K + NVIDIA GeForce RTX 3080
  • AMD Ryzen 9 5950X + AMD Radeon RX 7900 XTX
  • Apple MacBook Pro M5 Max

However, we often also include results for other hardware combinations depending on platform availability.

Common Resources

Public webpages:

Non-public resources:

Puget Bench Roadmap

Current roadmaps and development tasks are available here:

Offline Activation & Upload/Download Disabling

Offline activation can be achieved by placing an offline.json file in the application folder. This file can also be used to disable the ability to enable result uploading the GUI (uploading can still be performed via CLI with the –upload_results argument) and to disable the downloading of assets over the internet.

  1. Go to the “com.puget.benchmark” folder
    • The easiest way is to launch the Creators application, go to settings, and click “Reveal Logs”
    • Actual location is:
      Windows: C:\Users[USERNAME]\AppData\Local\com.puget.benchmark
      MacOS: ~/Library/Application Support/com.puget.benchmark
  2. In this folder, create a new “offline.json” file, or download this file and place in the “com.puget.benchmark” folder
    • Within this file, enter:
      {
      “license”: “LICENSE_KEY”,
      “allowUpload”: true,
      “allowDownload”: true
      }
  3. Update LICENSE_KEY with your license key
  4. For “allowUpload” and “allowDownload,” change the value to false to disable the GUI’s result uploading and benchmark/downloading options

Benchmark Database Result Filtering

To help prevent NDA and pre-launch results from showing up in our public benchmark result database, we offer a robust system for filtering results. We currently allow for two methods:

  • Whitelist – allowing only results that match
  • Backlist – blocking any result that matches

Filters are triggered based on two fields that are logged for all gathered system specs:

  • Manufacturer (GenuineIntel, AuthenticAMD, NVIDIA, etc.)
  • Name (Intel Core i9-14900K, AMD Ryzen 9 7900X 12-Core Processor, NVIDIA GeForce RTX 3070, etc.)

Both the whitelist and blacklist allow for single- (_) and multi-character (%) wildcards. Examples:

  • “Puget T5_0 Laptop” would trigger for
    • “Puget T510 Laptop”
    • “Puget T520 Laptop”
  • “Puget % T5_0 Laptop” would trigger for
    • “Puget T510 Laptop”
    • “Puget Pro T520 Laptop”
    • “Puget Entry T500 Laptop”

Filters can be either permanent (for common pre-launch naming schemes), or temporary and removed after a product launch. To request a filter, please email your Puget Bench Development Program representative.

Client-Side Result Upload Blocking

In addition to the server-side result filtering, we also have basic detection for NDA/pre-launch hardware built into the Puget Bench application. That is not intended to catch everything but is an additional layer of security to help prevent NDA leaks. When NDA hardware is present, a banner is shown to the end user of:

When this occurs, the “Upload results to the public database” option in the UI is turned off and disabled regardless of if the user has a benchmark license or not. If you wish to test this functionality without having NDA hardware, you can create an empty “nda.json” in the com.puget.benchmark folder.

GitHub Access

By request, read access can be given to the GitHub repo for the benchmark packages to enable custom testing and package development. Please email your representative for more information.