PC Hardware Optimization for Resolve 19

I've searched around without finding anything recent or specific, and maybe I'm using the wrong search terms, but I'm trying to find out which hardware components have the impact on performance with Resolve. In other words if I can spend more on something, what should it be? Faster GPUCPU, larger GPU/CPU cache, more PCI channels/buses, faster SSDs, more memory?

I'm going based off the hardware reqs I found here in the forum (below), but they don't give any direction on performance.

• 16 GB of system memory. 32 GB when using Fusion
• Integrated GPU or discrete GPU with at least 2 GB of VRAM
• GPU which supports OpenCL 1.2 or CUDA 12


Thanks so much!

What are you doing? Just simple edits? Longer videos? Lots of fusion? NR? Other intensive features? Resolution?

In general

GPU. At least 8GB of Vram for 4K. More is better. Higher resolutions even more.

CPU matters more for Fusion. The IGPU can also matter if you need the decoding functions. Think HEVC 4.2.2

You need fast enough storage for the files you're editing.

I wouldn't skimp on RAM. Even 32GB is a long enough number you'll soon enough want more.

On a very general level, you want "more" of these in your GPU.

CUDA Cores
Memory Bus
VRAM

Hi.

In Resolve the CPU is used to run the app, disk I/O, fusion, decode and encode of codecs.
Resolve does all its image processing in the GPU on the graphics card. Higher CUDA/OpenCL performance is better.

In the paid STUDIO Version of Resolve, can certain combination of Codec, Resolutions, Bit width and Chroma subsampling be hardware decoded/encoded on either a AMD/nVidea Graphics card or in some Intel non Xeon CPUs.

Resolve do all its Image processing in RGB fp32. 4K videos have 4 times the pixels of HD. And for 4K is the absolute minimum 6 GB of vRam on the Graphics Card,
but more are recommended.

Regards Carsten.

Higher CUDA/OpenCL performance is better.

I like that you can differentiate from shaders, compute units and cuda cores and the APIs CUDA and OpenCL which are different things.

I'm going based off the hardware reqs I found here in the forum (below), but they don't give any direction on performance.

In Resolve FP32 compute performance is what matters the most in >90% of cases. Get the highest FP32 compute GPU with the highest VRAM you can afford. Either AMD/Mac/nVidia. As a general guide.

The differences between them are in niche scenarios for which you're better to go look at some benchmarks from reputable sources like Puget which does a breakdown of benchmarks that are workflow specific.

For example if you use a 8K Red RAW and ProRes workflow then a card's 20% improvement in noise reduction or "neural engine" makes no sense if its 150% slower in Red RAW and ProRes. Or it doesn't even work properly with 8K+ material if its VRAM starved (like the 8GB nVidia cards regardless of cuda cores).

Note down your specific workflow and look at benchmarks that reflect it. Don't get blanket statements that this is better than that based on brand alone. It doesn't apply, it never applied.

I've searched around without finding anything recent or specific, and maybe I'm using the wrong search terms, but I'm trying to find out which hardware components have the impact on performance with Resolve....

As already stated, the GPU and GPU memory are very important for Resolve. However if you will be dealing with 4k H.264 or H.265 (HEVC), especially 10-bit 4:2:2 variants, the video hardware acceleration can also be very important.

Even IF you plan on transcoding all H.264/H.265 content to optimized media before editing, the RATE at which you can transcode that is affected by the hardware video acceleration on both input and output sides. I've worked on large documentary projects where it took a week on a powerful machine to transcode the material to ProRes proxies for editing. If anything happened and the proxies got out of sync or went permanently off line, it then took another week to regenerate them.

Many people have posted "I had no idea about which Resolve platform configurations had hardware acceleration, I just spent lots of money on a new machine and it turns out all my 4k material from all my cameras is not accelerated -- why didn't anybody tell me this?"

This table from Puget Systems shows the current status: https://www.pugetsystems.com/labs/artic ... -pro-2120/

This table from Puget Systems shows .....

Hi

And this link from nVidea shows hardware assisted
Encoding and Decoding for nVidea Graphics cards:

https://developer.nvidia.com/video-enco ... matrix-new

Regards Carsten.