Neat Blog

4 Factors when choosing a GPU for faster digital noise reduction

You can check the latest article on GPUs for video editing here.

In the previous article we highlighted two important pieces of computer hardware that affect the speed of Neat Video on your system and discussed the first, which is the CPU (Central Processing Unit). The second main component is the GPU (Graphics Processing Unit), which handles the visual processing needs of your computer. That’s the topic for this article.

About GPUs

GPU’s are powerful processor cards that were originally added to computers to enable them to run video games better. However, software developers quickly realized their processing capabilities could be put to work for general computing. Neat Video is one of these GPU-enabled applications, which harnesses this GPU power to allow it to work much faster.

This is good news because the GPU card is one of the easiest components to replace. Often it’s a case of simply removing the old GPU card and slotting in a new, more powerful one. Making it the quickest and most cost-effective way to improve your experience of Neat Video. However, there are many GPUs to choose from and different GPUs perform very differently on different plugins and applications, so it’s critical to choose the best GPU for your purposes.

Neat Video works with many NVIDIA or AMD graphics cards. In either case, when it comes to digital noise reduction there are a number of other factors which must be considered together and cannot be viewed separately.

So, let’s discuss the important features that make a great graphics card for our purposes. Then, we’ll give you some recommendations on various GPU cards that we’ve tried out, so you can make a more informed decision on what GPU works best for your needs and your budget.

Factor 1: GPU processing power

Processing power (measured in GFLOPS) of a GPU is measured by the number of FMA (multiply-and-add) instructions that can be executed by all GPU’s cores per second. You can find information on the GPU processing power for AMD and NVIDIA GPU cards here:

https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units
https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units

Factor 2: GPU memory (vRAM)

GPU uses data to perform its calculations, which is kept in its random-access memory (vRAM, measured in GB). For optimal performance, Neat Video needs about 1GB vRAM for Full HD videos and 4GB for 4K. It will work with less memory too, but at a reduced speed. Remember, host applications may also use GPU memory so you must factor that in and allow sufficient vRAM to ensure stable performance of the whole system.

Factor 3: GPU memory bandwidth

More GPU vRAM results in better performance, but this is not the whole story.  Another thing to look at is memory bandwidth (measured in GB/sec). Essentially, memory bandwidth is the speed of the video RAM. GPUs with more memory bandwidth can transfer data and to and from GPU cores faster, allowing Neat Video to perform noise reduction quicker.

Factor 4: CPU connection speed

GPUs can be connected to the computer using the internal PCIe bus (preferred way) or using an external eGPU connection, for example, a Thunderbolt connection. In either case, it is important to make sure the fastest possible connection is used. PCIe should better support and actually work at x16 speed (not reduce its speed to x8 or x4 because of other cards also installed in the system). Both PCIe and Thunderbolt need to be of the latest generations to be able to transfer large amount of video data at high speed (for example each PCI Express 3 lane is twice as fast as a PCI Express 2 one). If that is not ensured, then the connection between GPU and main system may become a significant bottleneck and slow down the overall render speed.

Some great performing GPU cards

These video cards all have impressive processing power, vRAM and memory bandwidth. Compare these products to find one that offers a balance between performance and price for your needs and your budget.

BrandModelProcessing power
(GFLOPS)
Single precision
VRAM
(GB)
Memory
Bandwidth
(GB/s)
Neat Video Speed,
FPS (tested on Full HD
8-bit video with
default filter settings)
Price
(may vary from
store to store)
NVIDIAGeforce GTX 1080 Ti106081148420.40$699
NVIDIAGeforce GTX 10808228835215.20$499
NVIDIAGeforce GTX 1070 Ti78168256 $469
NVIDIAGeforce GTX 10705783825613.90$399
AMDRadeon Vega Frontier Edition1132116484 $999
AMDRadeon RX Vega 64 (Windows)10215848417.50$550
AMDRadeon RX Vega 64 (Mac)10215848411.60not sold separetely
AMDRadeon RX 580 (Mac)579282569.17not sold separately

Do you really need to upgrade?
Check your existing GPU’s performance

Keep in mind that many older GPU cards such as these still offer good performance. Compare your current GPU hardware carefully before deciding if it is worth upgrading.

BrandModelProcessing power
(GFLOPS)
Single precision
VRAM
(GB)
Memory Bandwidth
(GB/s)
Neat Video Speed,
FPS (tested on Full HD
8-bit video with
default filter settings)
NVIDIATitan X614412336 
NVIDIAGeForce GTX 980 Ti5632633614.70
AMDRadeon R9 Fury X8601.14512 
AMDRadeon R9 Fury7168451216.40
AMDRadeon R9 390X5913.68384 
AMDRadeon R9 39051208384 

Two tools that will help

If you want to know how well Neat Video will perform on your hardware, we have two tools that will help – Neat Bench and Neat Optimizer

Neat Bench is a small free standalone application that can be downloaded from the Neat Video website, here. It’s really quick and easy to run – you could even download and install it at the store before you purchase a new PC to check performance. That way, you could request changes to the system if necessary before you buy.

Neat Optimizer can be found within the tools menu in Neat Video. Both of these will give you a report on how Neat Video is performing on your system, which will let you know if you should consider upgrading one or both of these components.

What’s next?

  • Do you have the right CPU for faster video noise reduction? If you haven’t read the article on the 5 factors to choosing the best CPU – you can read it here.