3 Replies Latest reply on Sep 14, 2015 10:59 AM by gstadter

    Rendering - Storage Device

    ti.s.cavalcanti Level 1
      1. I have a SSD with system Windows 10 installed, and After Effects CC 2015.


      • When I go to do a render, is more fast do this for the same Storage Device (SSD) in use, or use a external Storage Device (HD)?


      • In the case of rendering to the same disk (HD), becomes slower as it is obliged to read and write in the same time. And with SSD that does not happen? Or should I export the video to an external storage device (HD) for a better speed?




      Windows 10 Pro;

      After Effects CC 2015;


      GTX 770 2GB;

      12 GB RAM 1600 Mhz;

      CPU Quad Core 3.5 Ghz;

      HDD 1 TB 7200 Rpm;

      SSD 240 GB - Read: Maximum 450 MB/s | Write: Maximum 400 MB/s;

        • 1. Re: Rendering - Storage Device
          cc_merchant Level 4

          The performance difference will be negligent, since the system is already bottle-necked by the CPU and the amount of memory. I would not worry about it.

          • 2. Re: Rendering - Storage Device
            ti.s.cavalcanti Level 1

            Thanks for your answer!

            The cache works by writing lots of relatively little files -- one for each frame in the comp, plus one for each frame for each layer. For example, a 10-layer, 100-frame composition can generate 1,100 cache files (one per layer per frame, plus one per frame for the comp overall).


            In other words, the cache isn't just storing the composite of all layers together; it's saving that, plus the internal composite of each layer individually.


            That's a lot of I/O requests, so keeping the cache on an SSD is very helpful in minimizing latency, and keeping the cache separate from footage is helpful in reducing I/O requests on the same drive.

            Walter Soyka

            • 3. Re: Rendering - Storage Device

              Correct to say that the read/write going on with the cache is much greater than the IO to/from the footage?   are they somewhat close in IO demand when rendering?   or is it "mostly" just reading the footage during the process?