If you are an experienced overclocker who knows what performance you get from certain overclocks, this post might not be of interest to you. If you are running your system at stock speeds though and wondering what performance gains you could achieve with overclocking, you should definitely continue reading.
In this post, I will show how export times in Premiere Pro CC2015 are affected by overclocking the CPU, its cache and ram.
Test system setup
Before we go into the test itself, here are the related specs of the system to be tested.
CPU: Intel i7-5930K
RAM: 4x 16GB G.Skill Ripjaws V DDR4-2800 CL15
Mainboard: Asus X99-E WS 3.1
PSU: Be Quiet! Dark Power Pro 11 1000w
CPU cooler: Noctua NH-D15 with 2x Noctua 140mm industrial 2000rpm fans
Case fans: 5x 140mm Noctua industrial 200orpm,1x 120mm Noctua industrial 2000rpm
Case: Thermaltake Core x9
The baseclock (Bclk) for all tests is left at 100. The voltage settings are set to auto except the ram voltage which is set to 1.35V at the 2800MHz setting.
The Intel i7-5930K, a six-core x99 Haswell-E CPU, is tested at 3.7 GHz (stock turbo mode), 4.3 GHz and 4.4 GHz.
The CPU cache is run at 2.4 GHz (stock) and 3.5 GHz. For two overclock profiles, it is also run at 4.0GHz
The RAM is tested at 2133MHz (stock) and with the provided xmp-profile at 2800MHz.
I decided to use a 3 min UHD 25p timeline that should be encoded to h.264 UHD. The source footage is a 1-minute clip shot with the Sony A7Rii that I duplicated twice. The recording codec is XAVC-S 4K 25P 100M. The following screenshot shows the sequence settings. It is in German, but I hope you still understand:
I chose export settings that are in accordance with Vimeo’s export guidelines for UltraHD. So, here are the export settings:
For better monitoring, I decided to export via Adobe Media Encoder . Right before the export finished I took a screenshot of the export time (usually 1-2 seconds before end).
Here is an example:
Export time results
The following graph displays the different export times:As you can see, there is quite a significant decrease in export times due to overclocking. Between the stock settings and the highest tested overclock, the export time decreased by more than 2 minutes. That is a decrease of 17.7%. If you now consider doing the same test with a longer sequence, the time saved becomes quite noticable.
Further, you can see that cache and ram had quite an impact as well – more than I expected. A cache overclock from 2.4 to 3.5 GHz as well as some ram with overclocking xmp (or manual overclock) should be considered as the test suggests.
The time saved varied from setting to setting which also surprised me a little. It seems some combinations of settings work better than others. Alternatively, the auto settings did cause some variation.
The test showed again how much overclocking can affect your export times. If you want to keep it simple, you could save a lot of export time by increasing the multiplier of your CPU and raising the core voltage accordingly. If you have time to spare, you can work on your cache and ram frequencies. As shown in the test, these two decrease exporting times even further.
Don’t forget to make sure your overclocks are stable. There is no point of having a very fast system that crashes 20 seconds into the export. Software like OCCT, Realbench or Aida64 are stress tests you can try. For ram, memtest is a popular tool.
In the end, you also need to decide how much cooling you are willing to provide to run a certain overclock. In my case, even though my system can run @4.4GHz, I decided to stay at 4.3GHz because this way I could run a cooler system with less fan noise and still very good performance.
I hope this little test helped you decide on whether you want to overclock and if on how far you should go.