Depends on the media you are working with. 1080 Media normally wont require much more than a 980GTX card unless the timelines are complex with allot of GPU accelerated effects. A 960GTX or 970GTX card are normally more than enough for 1080 media and most workflows. When you get to 4K plus media then you want the 980Ti card ideally. The 970GTX and 980GTX will handle 4K media ok but not as well as the 980Ti. If the editor uses other apps such as Davinci Resolve as well then you definitely want the 980Ti and maybe the Titan X card. The Titan X card though is not required for Adobe since none of the Adobe applications so far have used even the 6GB of vram on the 980Ti card.
Thank you Eric!
Is there any site that has a good showdown of different cards going up against each other for Premiere use only? I've looked around and not found many... especially ones where there is a definitive answer about the difference between a card with 1000 CUDA cores vs 2000 etc...
Anyway, thanks again!
Bill Gehrke, from this forum, maintains a website which has a benchmark test for Premiere Pro. I believe you can find it at : PPBM7.com
You can test your own machine and compare it to others. However, Eric gave you very good and specific advice. Assuming the rest of your machine is up to snuff, what he said is valuable to know.
Choppy playback can be caused by a variety of things like : insufficient CPU cores and clockspeed, not enough system memory, memory which is not fast enough when using certain intense 4K codecs, insufficient amount of DDR5 video memory on a gpu, and MOST COMMON....a storage configuration which is TOO SLOW to feed media files and other operations to the editing program.
So, it's not just the GPU.......your whole machine needs to be properly balanced to achieve the best performance.
Thank you JF!
I found that on a beefy dual-Xeon, dual GTX Titan (not X) Superclocked system that playback of 4 layers of RED 4.5k was excellent when one GPU was assigned to Premiere Pro CC 2015 and the other was connected to a 4K monitor. It was dropping some frames when both cards were "enabled" for CUDA processing (using the NVIDIA tool).
Can you explain this?
GPU accelerated renders were much faster with both GTX Titans enabled, but I found this playback optimization quite interesting.
Well depends on the slot configuration. On the Dual Xeon boards 1 CPU handles half the slots and the other CPU handles the other half. Same with the ram. The Application scheduling based on which GPU was ready for the requests may be increasing the latency of processing causing the Playback to stutter. Either that or I suspect the data may be taking to long to transit the bus from CPU 0 to CPU 1 for the processing to occur on the other GPU. Slower clocked speed Xeons have far less margin for latency increase while maintaining realtime playback and leaving both GPU's available for the application or driver to select the available one relies on the application programming to efficiently handles that.