1 Reply Latest reply on Sep 16, 2015 12:19 PM by cc_merchant

    Yes, another new build/what gpu card thread needs some input

    2drumstix Level 1

      I’ve been reading the many tech blogs and forum posts about Premiere and video cards, and so here are my questions.



      Is my current understanding for when your GPU enabled CUDA is used correct?:

      CUDA Video Rendering and Playback – for generating the previews as you playback your timeline in your Program Monitor. (I leave my clips yellow, and don’t render previews)

      CUDA Rendering/Export – only for certain effects, rescaling, frame rate conversation, HD to SD, and deinterlacing. (I like to deinterlace all my DVD renders)


      I’ve read the Tweakers Page, and it’s very helpful. But I’m a workflow combination of Category A & B. Cat A in that I want a snappy experience when scrolling and I use Magic Bullet Suite. And I’m also Category B in that I don’t render previews but I want to speed up export times. But I have still have a few questions of which Nvidia card and how many to install.


      Tweakers page talks about a balanced system, so I don't want to bottleneck myself from the start if possible. I need snappy playback, but need speedy export times. I think tweakers page is saying the 970, 980 and 980 ti are all considered warrior cards, but hoping for input on the following?

      For a snappy playback experience, is a GTX 970 enough? Or a GTX 980 is the go to? Or is the bigger bang the 980 ti?

      For rendering and export, same question as above?


      And lastly, I know I can install two video cards, recommended they be of the same family and perhaps the generation, any input on the following?

      If I Install two of either GTX 970, 980 or 980 ti – put both monitors on card one. But does the 2nd card get used at all with CUDA enabled playback when your playback preview is being generated for your Program Monitor?





        • 1. Re: Yes, another new build/what gpu card thread needs some input
          cc_merchant Level 4

          My perception on video cards is something along these lines:


          970 : great on a 5930 system, could be used on a 5820.

          980 : great on a 5960 system, could be used on a 5930.


          Both these cards may be overkill on lesser CPU's. The 960 is more suitable on these lesser systems.

          The 980 Ti or Titan would require at least an overclocked 5960 with 64 GB memory or dual Xeon 12+ core CPU's. In most circumstances these cards would be idle for a significant amount of time on lesser systems, thus making it overkill, budget-wise.


          Lumetri adds to the requirements for the video card, so if you are a heavy Lumetri user, you may need to step up one level. So, if you planned on a 970, get the 980 instead, if you planned the 980, step up to the 980 Ti.


          Dual video cards are hard to justify for PR use alone. If you use applications like Octane or other 'video card hungry' applications, dual video cards can make good sense. What happens with two video cards in PR is that video card 1 handles all odd frames and video card 2 handles all even frames. You can imagine that in order to benefit from this odd / even processing by the video cards, you must have a workflow that constantly uses far more than 50% of the GPU when using a single video card. This could be a good test to see if your workflow and system could potentially profit from two cards and justify the extra cost. If your typical timeline and your workflow show that the single GPU is usually used 50% or less, there is no justification for two video cards.


          Hope this helps.