PugetBench for DaVinci Resolve
Written on August 26, 2019 by Matt BachAt Puget Systems, one of our primary goals is to make sure that our customers end up with a fast, reliable workstation that is perfectly tailored to their unique workflow. The main way we do this is by benchmarking a wide range of hardware in DaVinci Resolve (and numerous other software packages) that we then publish in our ongoing series of hardware articles.
However, while this ensures that we are selling the right hardware, it does not give our customers a great idea of how much faster a new workstation might be compared to their existing system. In addition, with the amount of hardware that is currently available, it is impossible for us to test everything on our own, so we want to provide tools for other hardware reviewers to perform real world benchmarking in professional applications like DaVinci Resolve
To address both of these issues, we are making our DaVinci Resolve benchmark available for download so that anyone can can perform the same testing we do in our hardware articles.
Commercial License
Includes optional uploading of results, email support, automation, and logging.
Be sure to check out the full list of benchmarks we have available!
The main benchmark download includes all the files necessary to run the "Standard" preset. If you wish to run the "Extended" preset, you will need to download the supplementary "Extended Assets" from the link below (note that this preset requires the Studio version of DaVinci Resolve and a GPU with at least 8GB of VRAM):
This benchmark is currently for Windows ONLY. We have plans for a MacOS and Linux version in the future, but there is no ETA for when it will be available.
System Requirements
DaVinci Resolve 16.2 or newer (beta versions not officially supported)
- 16GB of memory
- 4GB of GPU VRAM
- 50GB free space on your drive
- Compatible with both Windows 10 only (MacOS and Linux support is planned for the future)
- Must be run from a local drive (no network attached storage)
Note that this benchmark is still in BETA. Plug-ins and customized preferences in particular may prevent the benchmark from running properly. If you run into any issues, we recommend resetting your preferences to default.
A Note on DaVinci Resolve Free vs Studio
Blackmagic has two versions available for DaVinci Resolve - a Free version and a paid “Studio” version. In the context of benchmarking, the primary differences between these two versions is the following:
Free | Studio | |
---|---|---|
Maximum GPUs | 1 | 8 |
H.264/HEVC GPU Decoding | No | Yes |
H.264/HEVC GPU Encoding | No | Yes |
Maximum Timeline Resolution | 4K | 8K+ |
OpenFX | Watermarked | Full |
Noise Reduction | Watermarked | Full |
While the “Standard” and “R3D Mode” presets can be run on either version, the lack of GPU decoding/encoding means that you will always see lower performance with the Free version of DaVinci Resolve. In addition, if you want to test multiple GPUs or run the Extended preset, you will need the Studio version.
How to run the benchmark
- Download and unzip the benchmark from the link above.
- Install DaVinci Resolve (if not already)
- After installing Resolve, run it once to accept any default configuration options, firewall permissions, etc.
- Confirm Resolve configuration options. We recommend the following:
- DaVinci Resolve -> Preferences… -> Memory and GPU -> GPU Processing Mode: Auto
- DaVinci Resolve -> Preferences… -> Memory and GPU -> Use display GPU for compute (Studio edition only): enabled
- Launch the “Run PugetBench for DaVinci Resolve” program
- Configure any settings you want to change (especially if you have a commercial license).
- Run the benchmark!
Note: If you want to run any of the extended tests, also download the "Extended Assets" zip file and copy the "Extended" folder to the "Assets" folder. You should end up with both a "Standard" and "Extended" folder side by side in the Assets folder.
When the benchmark is complete, it will give you an "Overall Score" and upload the results to to our online database (unless told not to with the licensed version). From the uploaded results, you can view the sub-scores and individual test results.
Troubleshooting
If you have any issues with the benchmark not properly running, first verify that your system meets the System Requirements and that the version of DaVinci Resolve you are using is supported. If there are no issues with your system or software, there are a number of troubleshooting steps you can try:
- Make sure you are following each step listed in the How to run the benchmark section
- Do not attempt to use the system while the benchmark is running.
- Clear the application preferences and reset them to default. You can do this either in the DR preferences window, or by deleting the “DaVinci Resolve” folder in “%appdata%//Blackmagic Design”
- Disable Windows Defender or any real-time virus scanning programs when you run the benchmark.
- The benchmark does not work over an unmapped network drive (I.E. //192.168.1.2/Benchmark/). Mapped drives should work, but we highly recommend having the benchmark on a local drive as storage performance can affect the results.
- If Resolve or the system crashed mid-run, it can corrupt the project files. Re-download the benchmark (or unzip from the previous download)
- Re-download the benchmark and use a unzipping tool like 7zip to unzip the files rather than the tool that is built into your OS.
Lower than expected performance can be caused by a number of factors, but the most common are:
- Comparing results from the free version of Resolve to the paid Studio version.
- Outdated BIOS/drivers/software
- Running background processes (Antivirus, Windows Updates, etc.)
- Windows power profile (High Performance is recommended for the best results)
- DaVinci Resolve application preferences are not correct (GPU acceleration not enabled, etc.)
If you continue to have issues, we recommend posting in the comment section below - ideally with a screenshot of the problem and a copy of any log files that are generated by the benchmark.
What is tested?
Our DaVinci Resolve benchmark looks at rendering with a wide range of codecs at 4K and 8K (Extended preset only) resolutions, OpenFX, as well as performance in Fusion.
Our test media consists of the following resolutions and codecs:
Test Media (59.94 FPS) | |
---|---|
4K H.264 150mbps 8-bit | 4K ProRes 422 |
4K RED | 8K H.265 100mbps |
8K RED |
Special Thanks to:
Linus Media Group (who produces the popular Linus Tech Tips and TechLinked channels on Youtube) for the 4K and 8K RED clips which were also transcoded to create the H.265, and ProRes clips.
Neil Purcell (Lighting Cameraman / Camera Operator) for the 4K H.254 clip from his Panasonic GH5.
For the 4K and 8K media tests, we benchmark with a relatively basic grade in order to focus on the processing performance for those codecs. Exporting is done both to H.264 (a very standard delivery codec), as well as to DNxHR HQ in order to minimize the encoding step of the process.
The GPU Effects portion of the test focuses on OpenFX and noise reduction with the following effects:
- Temporal NR x3 - 2 Frames Better
- Temporal NR - 2 Frames Better
- Film Grain
- Spatial NR - Better
- Lens Blur x5
- Lens Flare
- Optical Flow - 50% Enhanced Better
- Face Refinement
Lastly, the Fusion tests look at the performance in the Fusion tab for a range of projects. These focus on:
- 3D Titles
- 3D Lower 3rd
- Phone Composite (tracking + green screen replacement)
- 3D Backlit Text (motion graphics)
- Turbulant Particles (motion graphics)
How does the scoring work?
The scoring system used in our benchmark is based on the performance relative to a reference workstation with an Intel Core i9 10900K and NVIDIA Titan RTX 24GB.
The results are all combined into the "Standard", and "Extended" Overall scores depending on the benchmark preset you chose.
Update Log
Version 0.92.1
- Fixed tests for AMD GPUs - a few tests had incorrect names that resulted in the overall scores not being generated.
- Improved DaVinci Resolve type and version parsing in order to work with the DR 17 beta.
Version 0.92 BETA (major update)
- Complete re-wrote benchmark from the ground up.
- Complete re-vamp for the tests - results are not comparable to previous versions.
- Results are now uploaded to our online database. This is required for the free version, but opt-in if you have a license.
- Improved method used to gather system specs. This should break less often on unusual or older system configurations.
- Licensed options added.
- Status logs and configuration settings moved to "~Documents⧵PugetBench⧵DaVinci Resolve⧵TIMESTAMP"
- General bug fixes and stability improvements.
Version 0.61 BETA
- Minor update to fix issue when running the benchmark from a non C: drive
Version 0.6 BETA
- Replaced H.264 media with 150mbps footage straight from a Panasonic GH5
Version 0.5 BETA
- First release.
Thanks! Small comment, you should add a line "C:" below the cd command since if you run the batch file from a drive different from C:, it will not cd onto the c: drive to find resolve.exe....
Good call, I didn't even consider that. I'll add that and update the benchmark. Thanks for the suggestion!
This is great to have a proper Resolve benchmark, but why no ARRIRAW or arx files in the benchmark? There are a lot of us shooting on Alexa that would love to see it included.
ARRIRAW (and BRAW for that matter) is definitely something we want to include. Right now, we simply don't have the time to actively go out and source test footage that meets all our requirements - primarily full permission to redistribute. If you or someone you know would be willing to help, however, we can add ARRIRAW into the benchmark pretty easily and start including it going forward
What we need are clips that are ~14 seconds long (we only need 12 seconds, but that gives us a bit of extra without making the file too large). A zooming/panning/moving shot of anything with a decent amount of color should work well - we mostly just need things constantly changing in frame and decent amount of different colors so that we can do more with it in terms of grading in Resolve.
We need UHD (or true 4K, either way) clips at both 29.97 and 59.94 FPS.This is an example of the RED footage we are using: https://www.youtube.com/wat... and the H.264 clip we are using: https://www.youtube.com/wat...
If you are interested in helping us out, feel free to reach out to me via email (just to avoid cluttering these comments). mattbach@pugetsystems.com
Hi Matt, thanks for the benchmark. I just wanted to add to this that you can download BRAW and ArriRAW directly from the websites
https://www.blackmagicdesig...
https://www.arri.com/en/lea...
I've been testing on a few different setups with high/low clocks, high/low core counts, Pascal vs. Turing, etc. A couple of observations:
1) The Fusion benchmark appears to be single threaded and thus very sensitive to clock speed. On a 24 core Xeon 8160, due to low clock speeds, I see no difference in score between a 2060 Super and a GTX 1060 6GB.
2) The 4K benchmark does not use more than one processor. Depending on the test it seems to use all cores/threads (24c/48t in my case), but only one of the two processors.
Have you seen these same behaviors?
That is not really the fault of our benchmark - that is just how Resolve behaves in general and is what we see over and over again in our testing: https://www.pugetsystems.co...[]=DaVinci%20Resolve
The only one that sounds a bit odd is Resolve only using one CPU. It should use both, although the overall load will likely be a lot lower on the second CPU.
I figured out what was going on. Windows can only handle 64 threads before it creates CPU affinity groups. In my case, I had Hyperthreading on, which made for 96 threads, so Windows split that into two groups of 48 threads (or 24 physical cores). Resolve cannot use more than one group, so it ended up using half the threads, which was half the physical processors. I turned off Hyperthreading, so that there were only 48 cores and 48 threads, under the 64 thread limit. Having done this, it used both processors just fine. You'll want to keep this in mind with the 3000 Threadrippers and Epyc where you can have 64 cores on one processor. The default will be to have SMT enabled, so you will exceed the 64 thread limit.
Interesting, I wasn't aware of the 64 thread limit. Thanks for sharing!
Is the RTX TItan worth for video editing? At least significant performance difference than the 2080 TI (in Davinci Resolve) ?
Both of those cards are in our recent DaVinci Resolve GPU roundup: https://www.pugetsystems.co... .
In short, performance is about on part, so the Titan RTX is really only useful if you are working with 6K or especially 8K timelines due to the higher VRAM it has.
deleted
I've a ryzen 5 3600 cpu and gtx 1660 6g gpu with 16gb ram ddr4 3200 is this able to editing 4k material or i need to proxy ??
Is there a newer version that works with Resolve 16.1 and Fusion 16.1 now that they have been released?
Your article is very interesting and meaningful.
www.caramembuatwebsiteku.co...
Excelent, one question, how realistic is a workflow for the Pros running others app in background? I know that setup these type of test have more work but the true potential of CPUs like the TRs is have a lot of things running a the same type without any lag, at the end you can do more for example, some applications dont scale well on so many cores so at least you can use the remain power for others tasks. I mean, this type of thing is not much enfatized on benchmarks and when AMD release the 64c TR monster this will be more evident.
Matt I think there's an incompatibility between DaVinci Resolve Studio 16.1.2 and the 0.61 Beta. With the new option to use GPU for debayer and decompression is selected I get stuck at 2% in the 4K RED,3x Temporal NR - Better 2 Frames portion of the benchmark. The only way I can pass it is to set it to debayer only.
There shouldn't be any issues, we used this version of the benchmark in our recent post about the new R3D GPU options. Maybe try resetting the preferences to default? There may be some conflict with residual preferences if you were using an older version of Resolve previously.
Deleting the preferences folder didn't work. If for sure reset the preferences but the benchmark is behaving the same, stopping at 2% in the same spot. It doesn't affect anything I work on so I'm just going to leave it and rerun it in the future when I upgrade my hardware.
Can this benchmark tool be scaled up to support the new 12K BM files? Lots of people on another forum claiming they can run 12K on their systems and we definitely need to see a standardized benchmark tool for 12k.
Yea, easily. The problem with 12K is that you are going to need a LOT of VRAM to work with 12K timelines. Even with 8K timelines, we recommend a GPU with 20GB+ of VRAM, so I worry that 12K timelines will need something like a Quadro RTX 8000 48GB which is an extremely expensive card.
What we might actually do is an article looking at what it takes to edit 12K BM media, but leave it out of the main benchmark simply because so few people are going to be able to actually do it. Or, maybe it will be more of a "add-on" to the base benchmark that people can selectively run (just like how we have a 8K preset).
One of the hardest things for us is often actually just getting our hands on test media that we have full permission to distribute. If you know of someone who has a URSA Mini Pro 12K and is willing to record a short clip, let me know at mattbach@pugetsystems.com !
Hi, Why do i get the error "Benchmark did not complete No Rendering timing .csv file was created"?
Same question here.
Have you solved this problem? I tried a lot of things and it still doesn't work...
I found a sollution! I moved folder from desktop to other drive. It's not place where I usually keep my files, but it's the only way I can run benchmark.
Do you guys have an easy chart to reference our scores to so we know how awesome or terrible our rigs are?
At the moment, you can compare to our DaVinci Resolve hardware articles. In the next version of the benchmark, however, we will be adding support for uploading results to our benchmark database:https://www.pugetsystems.co...
Can you add results checker for rendered files? On one of my machines Fusion benchmark produces corrupted files, it would have been nice if there was a warning about such situations. On the other hand, I don't know how difficult it is to implement such things.
Thank you for all effort. I guess is the newest version is 0.92 but the downloadable version is 0.61. Do you intend to share download link of version 0.92?
Probably in the next couple of weeks. It is a major improvement - which is why we have already started using it in our articles - but it still has a bit of polish needed before we can put it up for download.
Hello. Any improvements about the.0.92 version? I especially want to compare the RTX 3000 series with other old series myself, but I think the new version has not been released yet. Thank you
It had to get pushed back a bit because of all the new Adobe apps that require updates to the benchmarks. Since we have people with paid licenses for those benchmarks, we had to switch over to working on those for a bit. Hopefully that won't take long and we can get back to doing the last few things for the Resolve benchmark soon!
The description above for version 0.92 states that an 8 GB GPU is needed for the extended test. When running the actual test, it states that a 10 GB GPU is needed. Running it with my 2070, extended tests 1 through 6 work but I get a warning on the 7th test and then the test seems to stall.
An 8GB GPU should be enough - I'll update the text in the benchmark for the next update.
However, like all RAM/VRAM requirements, it is also going to depend on what else is using up the RAM/VRAM on your system. If you also have Premiere Pro open, or even a bunch of Chrome tabs, an 8GB GPU might not be enough. Heck, even a 24GB GPU could not be enough if you have enough other stuff using up VRAM. So while an 8GB GPU should be able to run the Extended test, that is really the bare minimum and can still have issues on some systems.
I did have Firefox open, so I closed it and reattempted. This time it got through the 7th extended test but failed on the 8th.
I've noticed in Task Manager that DR seems to hang on to whatever GPU memory is allocated, possibly some sort of memory leak. Is there a way to flush GPU memory?
Something goes wrong for me in the last test and causes a system reboot. Is that me or the Resolve 17 beta?
That could either be a sign of system instability, or it could just be the beta being a beta. Blackmagic's betas seem to be pretty true betas, so there are often a ton of bugs and performance issues. I don't recommend using beta versions of Resolve unless there is some feature you want to check out - I definitely wouldn't use them for actual projects.
I tried to run pugetbench with Davinci Resolve 16.2.3, Dual Xeon E5-2699v4(HT Disabled for certain purpose), Dual Radeon VII but it showed an error message, failed to generate score. how can it be solved? https://uploads.disquscdn.c...
Something is preventing that test from rendering. Unfortunately, the error reporting from Resolve isn't that robust, so the cause is going to be hard to pin down. Usually, it is something like there not being enough drive space free for the render to complete, the render is failing due to a software/hardware issue, or Resolve is crashing during that render.
I'm trying to run the benchmark but keep getting the error that Resolve quit unexpectedly. It happens as soon as the “loading resolve project” step is completed. I’m running Resolve Studio 16.2.8 and also tried it on the latest beta with the same result. I tried deleting my preferences file as well. I'm running with dual 1080 Ti cards with the latest Studio drivers from NVIDIA. Any thoughts to try?
I'm seeing the same issue. I have
Windows 10 Pro 20H2 (19042.789)
AMD Ryzen 9 5900X
MSI GeForce RTX 3060 Ti Ventus 3X
NVidia Studio Driver version 461.40
DaVinci Resolve Studio 17 Beta 8
PugetBench for DaVinci Resolve 0.92.1
If I run the benchmark, It will get to "Loading DaVinci Resolve project..." then there will be an error message from Resolve stating that it has quit unexpectedly. From there, PugetBench eventually times out saying Error: Resolve has not launced or the initial resultes file is not present within the 2 minute timeout period" https://uploads.disquscdn.c...
Same here, Studio 17, Beta 8
Beta versions of Resolve are not fully supported. It should work, but BlackMagic's betas are true betas, not early releases. They are full of bugs and issues that are still being worked out. I wouldn't recommend using the beta versions of Resolve unless there is a specific feature you want to try out - and I would revert back to the last stable release before working on any actual projects.
Matt, I was able to re-run the benchmark after installing 16.2.8. I also discovered that there are a couple of folders created that concerned me a bit. In my root drive, I had new folders for "install" and "vol0". It would be great if these could be either defined in settings or at least declared prior to running so users know.
What was more concerning was seeing a new user folder named mattb under c:\users\. I'm scrambling around searching my PC for points of intrusion only to realize that this is the same time that I ran the benchmark tool and it corresponds with your name. Is this folder necessary? If so, it should also be declared somewhere, perhaps when the user presses Run, that these folders will appear and that end users can safely delete them without freaking out.
Hmmm, those are odd and shouldn't be there. I wonder if Resolve is making folders to try to mimic the original system for things like capture clips, cache location, and media storage locations. It shouldn't need to since things like that are not used by the benchmark, but it may be a default behavior kind of thing.
I'll see if I can point all those to at least a single location that we can delete when the benchmark is done running.
Thank you. And thank you and team even more for the benchmark tool in the first place.
Hi Matt, I was trying to solve this again and have now downgraded to Studio 16.2.8 with a new database, and still get the same error. Any more thoughts? It's right after the step when it says "Loading DaVinci Resolve Project".