Is it safer to have the gpu or gpu maxed out for an extended period of time?

[iemcj]

I have a 3 hour long FCPX wedding video I'm working on I am using the NEAT video plugin (the built in one in FCPX is fine for home videos but barely better than useless for real work). The downside is it's very time intensive when correcting for the noise on 4k video, I'm at 1.01 frames/second using just the gpu, or .948 frames a second using the cpu. (using my imac, specs in sig).

Since it's going to take literally a week to properly render, would it be better for the machine to be having the CPU or the GPU the one I am using for this task since render speed won't be very different? Thanks!

 

[jerwin]

can you run NeatBench on both machines?

https://www.neatvideo.com/download/neatbench

 

[iemcj]

can you run NeatBench on both machines?

https://www.neatvideo.com/download/neatbench

I could but that's not really the issue lol. The question is if it's better for the thermals or something else in the computer to run EITHER the gpu or the cpu hard for a week. It'll be on imac no matter what, that's what a desktop machine is for while I'm using the laptop on the go.

Besides that benchmark tool is largely useless, that's assuming a fairly low resolution and low complexity file is being worked on. I'm running very specific noise reduction profiles on 100mbps hevc 4k footage.


EDIT: actually just realized you may be asking because you'd like to know for your own reasons, I apologize.

On the maxed out 2017 imac

CPU only (1 core): 3.21 frames/sec
CPU only (2 cores): 5.68 frames/sec
CPU only (3 cores): 6.51 frames/sec
CPU only (4 cores): 7.1 frames/sec
CPU only (5 cores): 7.14 frames/sec
CPU only (6 cores): 6.33 frames/sec
CPU only (7 cores): 6.25 frames/sec
CPU only (8 cores): 5.97 frames/sec
GPU only (AMD Radeon Pro 580 Compute Engine): 13.1 frames/sec
CPU (2 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 9.48 frames/sec
CPU (3 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 9.29 frames/sec
CPU (4 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 10.3 frames/sec
CPU (5 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 10.7 frames/sec
CPU (6 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 10.4 frames/sec
CPU (7 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 8.96 frames/sec
CPU (8 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 9.51 frames/sec

Best combination: GPU only (AMD Radeon Pro 580 Compute Engine): 13.1 frames/sec




On the maxed out 16" macbook pro

CPU only (1 core): 3.21 frames/sec
CPU only (2 cores): 6.67 frames/sec
CPU only (3 cores): 9.31 frames/sec
CPU only (4 cores): 11.4 frames/sec
CPU only (5 cores): 12.9 frames/sec
CPU only (6 cores): 13.9 frames/sec
CPU only (7 cores): 14.1 frames/sec
CPU only (8 cores): 13.5 frames/sec
CPU only (9 cores): 12 frames/sec
CPU only (10 cores): 11.1 frames/sec
CPU only (11 cores): 10.5 frames/sec
CPU only (12 cores): 9.52 frames/sec
CPU only (13 cores): 9.12 frames/sec
CPU only (14 cores): 8.64 frames/sec
CPU only (15 cores): 8.15 frames/sec
CPU only (16 cores): 7.66 frames/sec
GPU only (AMD Radeon Pro 5500M Compute Engine): 12.8 frames/sec
CPU (2 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 9.94 frames/sec
CPU (3 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 10.6 frames/sec
CPU (4 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 11.8 frames/sec
CPU (5 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 15 frames/sec
CPU (6 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 15.4 frames/sec
CPU (7 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 15.7 frames/sec
CPU (8 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 13.4 frames/sec
CPU (9 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 15 frames/sec
CPU (10 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 14.3 frames/sec
CPU (11 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 13.9 frames/sec
CPU (12 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 12.9 frames/sec
CPU (13 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 12.2 frames/sec
CPU (14 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 11.5 frames/sec
CPU (15 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 11 frames/sec
CPU (16 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 10.1 frames/sec

Best combination: CPU (7 cores) and GPU (AMD Radeon Pro 5500M Compute Engine): 15.7 frames/sec


Again this differs radically from the *actual* framerates you'll get if you're doing higher complexity levels of noise reduction and are working with higher resolution and bitrate footage. Here's the actual settings I'm using on one of the clips and the real world results on the imac.

Frame Size: 3840x2160 progressive
Bitdepth: 32 bits per channel
Mix with Original: Disabled
Temporal Filter: Enabled
Quality Mode: High
Radius: 2 frames
Dust and Scratches: Disabled
Repeat Rate: 0% of repeated frames
Jitter Filtration: Normal
Spatial Filter: Enabled
Quality Mode: High
Frequencies: High, Mid, Low, Very Low
Artifact Removal: Enabled
Edge Smoothing: Disabled
Sharpening: Enabled

Neat Video 5.1.8 Pro plug-in for Final Cut

Detecting the best combination of performance settings:
running the test data set on up to 8 CPU cores and on up to 1 GPU
AMD Radeon Pro 580 Compute Engine: 8192 MB total, using up to 95%

CPU only (1 core): 0.161 frames/sec
CPU only (2 cores): 0.29 frames/sec
CPU only (3 cores): 0.369 frames/sec
CPU only (4 cores): 0.406 frames/sec
CPU only (5 cores): 0.449 frames/sec
CPU only (6 cores): 0.448 frames/sec
CPU only (7 cores): 0.476 frames/sec
CPU only (8 cores): 0.464 frames/sec
GPU only (AMD Radeon Pro 580 Compute Engine): 1.19 frames/sec
CPU (2 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 0.818 frames/sec
CPU (3 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 0.909 frames/sec
CPU (4 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 1.1 frames/sec
CPU (5 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 1.06 frames/sec
CPU (6 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 0.997 frames/sec
CPU (7 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 0.989 frames/sec
CPU (8 cores) and GPU (AMD Radeon Pro 580 Compute Engine): 1.01 frames/sec

Best combination: GPU only (AMD Radeon Pro 580 Compute Engine): 1.19 frames/sec

 

[cynics]

I've done both for years on end with no negative effects on either the CPU or GPU. My CPU would be maxed out at for 2x the time the GPU would due to the amount of transcoding I did (literally none stop for years) and both GPU and CPU are still operating fine.

Even after years at 90c the thermal paste was still "ok" when I replaced it. (I had already cleaned the paste off the GPU and left it on the VRAM and CPU in the below pic).

 

[jerwin]

If everything is working properly, the GPU and CPU will undervolt in order to save themselves.... And if they do undervolt, the rendering time will be extended.

If it was me, I'd choose the method that results in the least heat build up (measured with something like istat menus)-- or consumes the least energy (measured with a killawatt or other power meter.)