NeatBench 5
You have not yet have Neat Video installed on a particular machine and you just want to check how fast it would be on that hardware? Or you want to see how fast that hardware generally is for a heavy video render task like Neat Video? Or you want to select one of several hardware configurations, for example, in a shop where you have just found a new shiny box of the latest and greatest computer from your favorite manufacturer?
Then you can use NeatBench — a standalone command line tool that detects the available CPUs and GPUs and then runs the standard Neat Video benchmark on them. NeatBench supports both NVIDIA and AMD GPUs, so you can directly compare those cards head to head and also compare their performance with the CPU-based processing. See the current technical requirements to GPUs and drivers.
Here is how you can use NeatBench:
- Download and unzip NeatBench to Desktop:
- Start NeatBench from Desktop.
- Once NeatBench finishes, find its log file: a file with the name starting with NeatBenchLog in the home folder. You can also check the command line output, it contains the same data.
- Check the frame rates reported and use them as a measure of the hardware speed.
If you also have similar measurement from your older or newer machine or from a computer of your friend or colleague, you can directly compare the computing power of the corresponding hardware.
We hope NeatBench will help you in selecting the best hardware for your video processing tasks.
If you already use Neat Video you can always check its speed with the built-in Optimize tool (menu Tools > Preferences > Performance > Optimize ). That tool runs the filter with different combinations of the available CPUs and GPUs and measures the actually observed processing speeds. Then the best combination can be selected for actual rendering.
Recent NeatBench speed test results
If you are thinking about investing into a new CPU or GPU and wondering how well they are going to perform, check NeatBench test results below. These results were obtained through our direct testing or received from our community members.
NeatBench 5 ran with default settings (1920x1080, 32-bit, temporal radius 2) on systems with different CPU and GPU models.
CPU | Best speed (Frames/Sec) |
---|---|
AMD Ryzen Threadripper 3970X | 48.7 |
Intel Core i9-13900K | 46.1 |
Apple Silicon M3 Max (16 cores) | 46.0 |
Apple Silicon M2 Ultra | 42.1 |
Intel Core i9-12900K | 41.1 |
Apple Silicon M3 Max (14 cores) | 36.0 |
Intel Core i9-10980XE | 33.8 |
AMD Ryzen 9 5950X | 33.7 |
Intel Xeon W-3245 | 31.3 |
AMD Ryzen 9 3950X | 27.8 |
Apple Silicon M2 Pro | 26.9 |
Apple Silicon M3 Pro | 24.3 |
Apple Silicon M2 Max | 23.9 |
AMD Ryzen 9 3900X | 23.2 |
Apple Silicon M1 Max | 22.7 |
Intel i9-10900K | 19.8 |
Apple Silicon M2 | 14.9 |
GPU | Best speed (Frames/Sec) |
---|---|
NVIDIA GeForce RTX 4090 | 120.0 |
AMD Radeon RX 7900 XTX | 97.0 |
AMD Radeon RX 6950 XT | 78.2 |
NVIDIA GeForce RTX 3090 | 70.0 |
NVIDIA GeForce RTX 3080 | 59.7 |
Apple Silicon M2 Ultra (76 cores) | 56.0 |
NVIDIA GeForce RTX 2080 Ti | 42.9 |
Apple Silicon M2 Max | 42.5 |
Apple Silicon M1 Max | 36.5 |
Apple Silicon M3 Max (40 cores) | 35.4 |
AMD Radeon RX 5700 XT | 35.1 |
AMD Radeon Pro Vega II | 32.8 |
Apple Silicon M3 Max (30 cores) | 30.8 |
AMD Radeon Pro W5700X | 28.7 |
Apple Silicon M2 Pro | 24.2 |
Apple Silicon M3 Pro | 17.4 |
Apple Silicon M2 | 11.5 |
The lists of the CPUs and GPUs located above do not represent certified and/or recommended hardware. The lists do not include all the models available on the market and only consist of some CPU and GPU models we tested directly or received reports on from the community. The results stated in the table are approximate only. The same CPU or GPU can work at different speeds depending on other hardware / software components it works with. NeatLab does not receive any monetary or other material support from any of the hardware developing vendors.
Older Version
If you want to compare the speed of the most recent version of Neat Video 5 with the previous major version 4 on the same hardware, please do that by running the tests with the most similar settings as described in this post.