Benchmarking the HP Z840 Workstation for Video, Part 3: Analysis

Test Description

Let’s start with a brief description of my tests that I ran.

• Convert to YUV: Time required to convert a 96-second 4K file from H.264 to YUV format.
• MSU VQMT: Time required to compare that file with the original and produce a VQM rating.

As before, I tested with HTT enabled and disabled, and rendered from all drives on the Z840. Table 1 (below) shows the overall fastest times for each computer, with numbers in green on the right where the difference exceeds 35%.

Table 1. Total performance Z840 to the Z800. All times in seconds.

In this test, the benefit of the faster system was profound, particularly when converting from HEVC to YUV, a relatively low CPU-intensive operation where the bottleneck is mostly write speed. When actually analyzing the file, a more balanced operation, processing time still decreased by 49%. If you’re manipulating and analyzing large data sets for a living, a computer like the Z840 could dramatically improve file throughput and efficiency.

How Much of the Difference Related To SSD vs HDD?

The second question is how much of the 840’s performance related to the faster disks. Unlike our other two test groups, in these analysis-related trials, the faster disks delivered a tremendous performance boost, which you can see in Table 2 (below).

Table 2. Rendering time differences between SSD and HDD drives in trials on the Z840. Click the image to see it at full size.

To explain, the two major groups show tests results performed with HTT enabled (on the right) and no HTT on the left. The green boxes in each section represent the fastest score, the yellow the slowest. The Delta column shows the difference between the fastest SSD drive and the HDD, and the time savings were dramatic.

In fact, if you compare the performance of the Z800 with the Z840 when testing from its HDD drive, as I’ve done in Table 3 (below), the difference between the two systems is minimal; in fact, for the FFMPEG conversion, the Z800 was actually faster. This indelibly underscores what you knew all along: If you’re buying a Z840 for data analysis, for gosh sakes get the fastest SSDs available.

Table 3. Comparing the Z840 testing from the hard drive with the Z800.

It also begs the question: How much faster could the Z800 perform if equipped with an SSD drive? With my deadline looming, I couldn’t buy a new SSD drive or request one from a vendor, so I had to make do with gear I had in-house, which was an old 64GB SSD I had never installed because, let’s face it, when you’re a video editor, a 64GB drive is a waste of a drive bay.

In terms of disk benchmarks, this older SSD drive was competent in writing tests, but nowhere close to either HP SSD drive at peak writing capabilities. You can see this in Table 4 (below), where the first two data columns show disk benchmark results from the two types of SSD drives on the Z840, while the third data column shows the results from the older SSD installed in the Z800. On the other hand, this older SSD drive was around twice as fast in reading and writing than the HDD on the Z800, shown in the last data column in Table 4.

Table 4. Drive specs for various drives on the Z840 and Z800.

In the first test, converting the MP4 file to YUV, conversion time with HTT enabled dropped from 367 seconds to 200 seconds, a tidy decrease of 45% about commensurate with the faster write speed of the SSD. Unfortunately, I couldn’t replicate the second test on the SSD because it didn’t have enough capacity; it had 64 GB, and I needed 66 GB.

So I performed the MSU analysis test with 5-second long 4K files located on the Z800’s HDD and SSD. The SSD took 59 seconds, while the HDD took only 56. Why so close given that the SDD could read data more than twice as fast as the HDD?

Our best clue is the Performance Monitor grab from the Z800 shown in Figure 2 (below), which captured CPU utilization when analyzing the file from the SSD and the HDD. As you can see, CPU utilization is very similar, despite the SSD’s superior read speed. This tells me that the process was limited by the speed of the CPU, not by retrieval from the hard disk. In other words, for truly peak performance, you need a balance of disk speed and processing speed.

Figure 2. Z800 CPU utilization when performing the MSU VQM analysis from an SSD and HDD.

Obviously, if you have an older computer you’re using for analysis-type tasks, substituting SSD drives for HDDs is a no-brainer. However, while it should improve performance for low-CPU usage tasks that are primarily limited by disk-access speed, if the task is CPU-limited, you may see very little benefit at all. For analysis tools like the VQMT, a balanced system like the Z840, which provides exceptional disk speed and plenty of CPU, is the best possible test bed.