• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
0

Hardware Settings for Premiere Pro

Community Beginner ,
Dec 02, 2018 Dec 02, 2018

Copy link to clipboard

Copied

I was tired of being hardware bound by my 4 core, 12GB machine.  So I built a new one using an i9-7960X CPU running 32 cores at 3.5GHz, 128GB RAM, NVIDIA Quadro P5000, two M.2 and three "regular" SSD drives.  This should last me a lifetime , and be more than PPCC uses for a while.  But neither PPCC nor Media Encoder seem to use much of the resources.

Now with PPCC, my CPU utilization typically runs well below 20%, and on export it typically runs about 20% - 30%, although sometimes on an export it does peg at 100%.  Memory utilization is typically below 10% no matter what.  GPU usage, with CUDA enabled, runs below 3%, always.

My goal was real time playback with effects, and reduced export speed.  Between my old and new PC, the export time for a two minute video dropped from 10 hours to 55 minutes, but even at 55 minutes, the PC's resources utilization was low.  The same clip today with a LUT is running at 2 hours, witn 23% CPU, 12% RAM and 1% GPU.

On real time processing, my performance is no better than 4 cores with 12GB, although PPCC doesn't run away (stop responding to user input).

I'm hoping there are hardware or PPCC / ME settings I'm missing and that someone can guide me on them.

Thanks.

Views

1.1K

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Dec 02, 2018 Dec 02, 2018

Copy link to clipboard

Copied

That's quite a beefy machine, you have there.

I did something similar a year ago. I have learned a lot by watching videos from Puget Systems on YouTube:

Puget Systems - YouTube

Hope this will help you as well.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
LEGEND ,
Dec 02, 2018 Dec 02, 2018

Copy link to clipboard

Copied

If you read through the very useful information of Puget Systems and SafeHarbor computing, one common thread is that many mobos and even high-end CPUs are not good performers with video processing, there are only a few each of mobo's and CPUs they recommend and use in their own builds. So ... that's part of the potential issue.

The second is knowing where to park your GPU and other bits in connections to avoid conflicts in the 'lanes' of the mobo.

Neil

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Guide ,
Dec 03, 2018 Dec 03, 2018

Copy link to clipboard

Copied

Hi

Your rig has plenty of grunt for graphics...FOR GAMING

The 32 cores you have and the huge GPU's dont add up to much if the SOFTWARE cannot address multicores.

I keep saying this to people who rush out and build monster machines.

Most of todays software rely on processing in a liner manner i.e instructions are sent one after the other to the CPU.

Multi threaded parallel stream software are a handful, Intel and Microsoft made a huge marketing hype around "hyper threading" and multiple CPU cores years ago and when the poor guy on the street bought these they performed almost the same as their old machines, Why? Yet again its the software holding the hardware back.

We are getting close to software that will address each core individually and then you will render in record times. This is why l love my render farm and why large companies like ILM etc create massive render farms. They are really a group of pc linked together which the software then sends instructions to in a straight line. But when the 1st instruction hits node 1 (PC1) on a farm the following data goes to Node 2 to be processed and the next data set goes to Node 3 etc. Some farms have thousands of nodes. Pixar has a massive farm. In this way the bottleneck of software not being able to instruct cores in parallel is circumvented.

You are buying good gear but make not mistake you are still at consumer level, pretty near prosumer level but eons away from a proper pro setup.

Trust this explains everything

Mo

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Dec 03, 2018 Dec 03, 2018

Copy link to clipboard

Copied

LATEST

Moved to the Hardware forum.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines