## Introduction
We have recently updated our Puget Bench for Creators suite to version 2.0 for both Premiere Pro and DaVinci Resolve. These updates don’t significantly alter the scores but address bugs and compatibility issues. Importantly, they change the testing approach and the elements assessed. This article discusses how these updates affect current hardware performance in these applications and their implications for professional creators.
We won’t delve into all changes in Puget Bench for Premiere Pro and DaVinci Resolve but will focus on key updates. Notably, none of the scores from version 1.0 are comparable to those from 2.0, including subscores and specific tests. Additions include NIKON N-RAW codec testing, 4x MultiStream testing, and a “Real World” test category that combines various tasks and effects to reflect actual timeline use.
For Premiere Pro, we removed software and HEVC encoding tests due to changes by Adobe and updated GPU Effects tests. In DaVinci Resolve, we added ProRes Encoding and updated AI tests, adjusting tests in “basic” and “standard” presets to better reflect user experience. With these adjustments, we hope to enhance how our benchmark guides creators and reviewers in evaluating PC hardware for professional workflows.
These benchmarks are free for personal use, and we offer both free press licenses and paid licenses for commercial use, providing access to advanced features and improved support.
### Test Setup
To determine hardware performance with these benchmark updates, we tested a range of platforms and GPUs, ensuring common features: up-to-date Windows installations, drivers, and BIOSes; JEDEC RAM at maximum frequencies; overclocking disabled; and VBS and ReBar enabled. All GPU tests used AMD Ryzen™ 9950X3D-based platforms.
We tested DaVinci Resolve with version 20.3 and benchmarks 1.2.2 and 2.0, focusing on the impact of the updated test suite. In Premiere Pro, we used versions 25.2 and 26 with benchmarks 1.1.1 and 2.0 respectively, due to compatibility issues with Adobe’s HEVC support changes.
## Puget Bench for Premiere Pro
Testing began with Premiere Pro, posing challenges due to version inconsistencies between benchmarks 1.x and 2.x. Script-based HEVC exporting was removed from post-25.2 versions, while new GPU effects requiring version 25.6 and later were added. Full NVIDIA Blackwell GPU support is included from version 25.3. These changes highlight differences in hardware support and performance as well as the updated blend of tests in Puget Bench for Premiere Pro 2.0.
The 2.0 benchmark showed better discrimination between configurations, especially favoring GPU impact over CPU in higher tiers. Despite changes, overall hardware recommendations remain valid, although there may be slight shifts.
In LongGOP tests, significant differences arose, with AMD Radeon™ RX 9070 XT notably underperforming compared to the RTX 5090. Improvements in NVIDIA media engines and elimination of CPU-based encoding tests contributed to this, although the AMD card remains competitive in specific areas like single-stream H.264/HEVC processing.
Single-stream tests revealed no advantage from multiple NVENC or NVDEC encoders on NVIDIA GPUs, indicating incomplete hardware support for parallel processing. Intel’s Quick Sync shows strength in HEVC, maintaining parity in H.264.
Intraframe tests saw minimal change, except for the RTX 5060’s poor performance due to insufficient VRAM, highlighting GPU preference for CPU power.
RAW codec tests highlighted the 9070 XT’s issues in CinemaRawLight and the 5060’s struggle with 4K RED 4x MultiStream, again due to VRAM limits. The 9970X also saw decreased performance due to GPU importance in added tests.
Updated GPU effects testing favored the 5090’s performance over the 5080, while the 5070 Ti and 5060 lagged. The 9070 XT showed better results with new GPU effects reflective of modern workflows.
## Puget Bench for DaVinci Resolve
Unlike Premiere Pro, both DaVinci Resolve benchmark versions are compatible with version 20.3, focusing on changes in tested features without the influence of application version differences.
Overall scores showed increased differentiation by GPU with minimal CPU impact, with minor changes in ranking between platforms.
LongGOP codec tests demonstrated clear NVIDIA GPU tier differentiation and relative decline in the 9070 XT and 5060, impacted by MultiStream test inclusion and VRAM constraints. The 9970X saw decreased performance due to more hardware-accelerated tests.
In Intraframe codecs, similar trends were observed, with the 5060 and 9070 XT facing challenges in Real World tests. The 9970X still leads in CPU performance but loses relative strength due to less CPU-dependent tests.
RAW codecs saw minimal changes, with new tests scaling similarly