Benchmarking the HP Z840 Workstation for Video, Part 1: Editing

Editing

I performed all editing tests with the 2015 release of Adobe Premiere Pro CC. Here’s a brief description of the tests.

• Run Premiere Pro: Time how long it took for Premiere Pro to load and become available on both systems.
• Motion Tracking: Time how long it took for Premiere Pro to track a woman walking in an AVCIntra 100 video file. I used sample files provided by Adobe here.
• Polly - Warp Stabilizer: Time how long it took for the Warp Stabilizer effect to analyze and stabilize 16 seconds of handheld AVCHD footage.
• Haunted - Dining Room: Time to output a 64-second DSLR promotional video to 720p H.264 output.
• 4K multilevel: Time to output a 96-second three-layer (picture-in-picture) 4K RED sequence to 4K H.264 output.
• 4K single level: Time to output a 96-second single-layer 4K RED project to 4K H.264 output.
• Udemy tutorial: Time to render a 5:52 (min:sec) two-layer AVCHD tutorial sequence (picture-in-picture) to 1080p H.264 output.
• Beth audition: Time to render a 5:00 (min:sec) single-layer AVCHD ballet audition sequence to 720p H.264 output.
• Josiah concert multicam: Time to render a 4:26 (min:sec) three-layer multi-cam concert sequence comprised of two HDV streams and one ProRes stream to 720p H.264 format.

Test Configuration

I performed all tests using the following settings.

• Scratch disks - same as project (default)
• Media cache database - C drive (default)
• For projects that involved rendering, I output to the C drive.

There are many views on the optimal configuration for preview and rendering performance. I asked Adobe for their recommendation, and in a three-drive system, my contact recommended using separate SSD drives for boot and caching, and a separate content drive. Unfortunately, since I was testing on multiple drives, this schema would have forced multiple error-prone changes to the system configuration every time I changed the content drive.

To test how much difference the various options might make, I setup the tests shown in Table 3 (below) on the Z840 using the 4K multi-level project. Test 1 is the best-case scenario, which followed Adobe’s recommendations. Test 2 is the worst case, with all cache and scratch drives set to the HDD, and the file output to HDD. The difference in rendering time as about 4%. Although I would definitely follow Adobe’s recommendations for my permanent editing station, I thought simpler was better for these tests.

Table 3. Best and worst case rendering times.

Otherwise, I performed all tests with HTT enabled and disabled, which you’ll hear more about in the discussion below. In terms of workflow, I exported all sequences from Premiere Pro to Adobe Media Encoder, and encoded from there with the system otherwise idle. With this as prologue, let’s jump into the results.

How Did the Systems Compare?

How did the systems compare, you ask? The Z840 was much faster, of course. What did you think was going to happen? All performance times shown in Table 4 (below) are in seconds, with a decrease in rendering time exceeding a totally arbitrary 35% highlighted in green.

Table 4. Total performance comparisons.

As you can see, there’s a very solid good performance boost across the board, with the least improvement shown with the multiple camera HDV/ProRes shoot. Not surprisingly, both RED-based projects showed very significant improvement, with the simplest project, the AVCHD audition video, showing the most benefit.