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

How Much of the Difference Related To SSD vs. HDD?

Again, since the Z840 had a Turbo SSD G2 boot/program drive, the tests below reveal how much you can improve performance by storing your content files on an SSD drive over an HDD drive. As you can see in Table 5 (below), not much.

Table 5. Performance differences between SSD and HDD drives in rendering trials on the Z840. Click the image to see it at full size.

To explain, the two major groups show the results from tests performed with HTT enabled (on the left) and without HTT on the right. The green boxes in each section show which drive produced the fastest result with and without HTT, while the yellow boxes identify the slowest alternative. For example, with HTT enabled, the HDD drive produced the lowest time in the 4K multilevel rendering, at 401 seconds, which I can’t explain, though the HDD also had the lowest rendering time with HTT disabled and I ran both tests twice.

The Delta column shows the difference between the HDD drive and the fastest SSD drive. As you can see, while the HDD drive was the slowest in most instances, the difference maxed out at 8%, and that was for the Warp Stabilization effect, not for rendering. The largest difference in rendering speed was 2%.

Given the performance differences between the drives as shown in Table 2, why aren’t the differences more stark? The obvious answer is that when rendering to H.264, retrieving media from the drive isn’t a performance bottleneck. This isn’t to say that there aren’t any projects where drive speed wouldn’t be a bottleneck, it just didn’t seem to be a bottleneck in any of these projects.

As an aside, most authorities that I checked didn’t recommend SSD drives for media. Rather, most pointed to RAID configurations for security, cost-savings, or both. Given the results shown in Table 5, that seems like good advice, at least for editing/rendering. You’ll see much more significant differences in the encoding and particularly the analysis-related tests.

HTT or Not HTT

What about whether to process with HTT enabled or disabled? That differed by project and computer, as you can see in Table 6 (below). Specifically, the table shows the times for all tests with HTT enabled (on the right) and disabled (on the left). The faster scores between the two are in green, the slower scores in yellow. As an example, on the Z840, all the RED-based multilevel 4K projects processed faster without HTT, which is why those boxes are green in the HTT section on the left. In contrast, on both computers, rendering the DSLR-based Haunted-Dining Room sequence was faster with HTT, which is why all those rows are green.

Table 6. Performance differences with HTT disabled (on the left) and enabled (on the right). Click the image to see it at full size.

I hear you thinking, why are there any green boxes in the No HTT section? Shouldn’t HTT, which doubles the compute resources, always be faster? Unfortunately, no.

That’s because computer programs don’t automatically use all available cores in a system; they have to be designed to do so. While Premiere Pro is undoubtedly designed and written to leverage all resources, it’s comprised of multiple processes, like motion tracking and the warp stabilizer, and separate drivers that enable importing formats like RED, AVCHD, and others. Some of these are supplied by third parties, some written internally years ago when multiple-core computers were uncommon. If a function uses multiple cores inefficiently, it’s not unusual for that operation to be slower with HTT enabled.

In fact, when computers with HTT first hit the market, most programs weren’t ready, and slower performance with HTT enabled was very common. For this reason, HP makes it simple to enable and disable HTT in the system setup panel, though not all manufacturers (like Apple) do so.

As an example, Figure 2 (below) shows Windows Performance Monitor tracking CPU utilization while the Z840 renders the two 4K projects from the Turbo SSD G2 and Turbo SSD drives with HTT enabled. As you can see, CPU utilization is hovering between 30% and 40%. Without HTT, CPU utilization ranged from 40%−70%. As a result, overall performance is actually slightly slower with HTT enabled than without it.

Figure 2. CPU utilization on the Z840 while rendering various projects with HTT enabled.

Where might you consider running without HTT? This is shown in Table 7 (below). Negative numbers show instances where HTT slowed encoding, and the largest decrease was only -7%, which is hardly worth chasing. Positive numbers, of course, show instances where HTT improved overall performance, with the Z800 showing the most consistent improvement with five test cases where the benefits exceeded 20%, which triggered the green background. If you have an older computer, it makes sense to make sure that it’s running with HTT enabled.

Table 7. Comparisons of tests with and without HTT enabled. Negative numbers mean that performance was faster with HTT disabled.

So, here are the key takeaways:

• A newer system like the Z840 should significantly accelerate rendering performance across a range of project types and formats.
• Using an SSD drive for content doesn’t appear to deliver substantial reductions in rendering times. Less expensive HDD or RAID drives are probably a better option.
• Enabling HTT doesn’t always deliver the fastest possible performance, but is generally a good idea when rendering from Premiere Pro.

This ends the editing section; in the next segment we’ll look at encoding.