HP six-core Z400 Test Drive, Part 1
Those users considering an HP workstation now have four units to choose from: the top-of-the-line dual-processor Z800 and Z600, which both sport the fancy new case that HP introduced last year, and the less-expensive Z400 and Z200, both single-CPU computers that feature updated innards, but the legacy case from the older xw workstation line. The Z400 starts at $999 for a dual-core model, while the very small-form-factor Z200 starts at $699. In contrast, the Z600 starts at $1,829 and the Z800 at $2,109both for single-processor systemsand scale much, much higher.
Because of disk requirements, for most serious producers, the purchase decision will come down to a comparison between the Z400 and either the Z600 or Z800. Last March and April, I reviewed the 3.33GHz dual-CPU 12/24-core version of the HP Z800. In this review, I'll compare a 3.33GHz single-CPU six/12-core Z400 to that system.
By six/12, I mean a system equipped with a 3.33GHz Intel Xeon W3680 CPU and six individual cores, each with Intel's Hyper-threading Technology (HTT), which adds much of the functionality of an extra core. When a multi-processor-aware program goes to work, it can use virtually all the power in those cores, as you can see in Figure 1, recorded during the Sorenson Squeeze testing that I detail below.
| Related Links |
||||||
|
Just to round out the testbed description, as with the Z800, the system was running 64-bit Windows 7 with 24GB of RAM, plus the same Nvidia Quadro FX 4800 as the Z800. In a hot but not excessive configuration, expect the Z400 to cost between $4,500 and $5,000. To compute this, I configured 12GB of RAM and substituted the Nvidia Quadro FX 3800, which should suffice for efficient operation. I used 24GB and the FX 3800 to match the Z800's configuration, not because I thought this was necessary for optimum performance.
In terms of agenda, in this segment, I'll describe the Z400 a bit, briefly discuss the multiple CPU performance conundrum, and then share some Z400 and Z800 comparisons in the streaming encoding space. Next time out, I'll focus on the much more difficult issue of comparing editing and rendering performance in Adobe CS5, and present comparative stats from Rhozet Carbon Coder. Let's start with a look at the Z400.
If you or I were the Z400, we'd have a serious case of sibling envy, since mom and daddy HP spent all the big bucks making the Z600/Z800 look fabulous, and nothing on us. All this beauty is on the inside stuff can only go so far. As HP explains, using the legacy case for the Z400 was a cost consideration; to meet the price points, the company had to sacrifice the new design. So there's no convenient handles for moving the unit, no cool baffles on the inside, no case that snaps shut like a Mercedes sedan, no fine Corinthian leather. Just kidding on the last partZ800 doesn't have it either, and I seriously date myself.
All that said, however, the unit is still quite serviceable. The side cover snaps off, exposing the bare-bones inner chassis that lets you swap out PC cards and hard drives without a screwdriver. Better run for your Phillips head if your power supply fails, though, since it doesn't slide out, unlike that of the Z600/Z800. Of course, how often does that happen?
Actually, while you get a sense of tactile satisfaction with the Z600/Z800 when swapping graphics cards, memory, and the like, getting to components in the Z400 is faster and simpler. Take of the side cover, and you have immediate access to most components, with only a simple cowling to remove to access memory.
The only significant day-to-day functional difference is that the Z400 is louder than the Z800, which uses the sophisticated inner cowling to cool more efficiently. It's not loud enough to be distracting, but if you're looking for a quiet computer for noise-sensitive applications, the Z400 isn't it.
The Z800 has 12/24 3.33GHz cores, and the Z400 has six/12 3.33GHz cores, so the Z800 should be twice as fast, right? Well, that depends on the efficiency of the software program itself, and how you're using it. For example, in this issue, we're looking at streaming encoders, including the Windows debut of Telestream Episode Engine 6.
By way of background, with version 6, Episode ($495), Episode Pro ($995) and Episode Engine ($3,995) all have a unified interface. Among other differences, Episode can only encode files serially, while Episode Pro can encode two files simultaneously. Episode Engine, which I tested, can encode up to 24 jobs simultaneously, as you can see in Figure 3. Obviously, on a multiple-core computer, Episode Engine will be the most efficient if you're encoding multiple files, or a single file to multiple outputs.
Figure 4. Telestream Episode Engine's Split and Stitch technology.
Episode Engine also features Split and Stitch technology, which allows the encoder to break a single file into multiple pieces, encode the pieces, then stitch them back into a single file. This control is shown in Figure 4. If you're encoding a long file that Episode Engine can split into multiple components, multiple cores also helps.
Let's see how these various capabilities translate into performance results, with Table 1 telling the tale. The first trial is a 6-minute DV file encoded into On2 Technologies' VP6 codec. With Split and Stitch technology enabled, the Z400 was 59 percent slower than the Z800. Without Split and Stitch, the difference dropped to less than 10 percent. If you're encoding single files and own Episode or Episode Pro, spending a lot more on the Z800 for large single-file encodes doesn't seem like a good investment.
Ditto for multiple-file encodes. Line three shows the comparative performance encoding 16 files to three formats each, for a total of 48 files. The Z400 was only 34 percent slower than the Z800, making the Z400 a very cost-effective encoding workstation if you have Episode Engine. With Episode and Episode Pro, you would expect the difference between the two systems to go way down, because Episode can only encode a single file at a time and Episode Pro two at a time. If you're buying a workstation to run either of these products, you might be better off with a much more affordable dual-core or quad-core solution.
If you do buy Episode Engine, or are considering it, rest assured that it will efficiently consume all the processor cores that you throw its way. You can see this in Figure 5, which shows Episode Engine flatlining all 24 cores of the HP Z800, which is very impressive. It may not be twice as fast as the Z400, but it's sufficiently faster that high-volume shops should consider Engine and a dual-CPU, 12/24-core system like the Z800.
The next program that I tested was Sorenson Media Squeeze. Operationally, Squeeze can encode Windows Media and VP6 files simultaneously, but not H.264. Like Engine, Squeeze lets you set the maximum number of simultaneous encodes, which I set at the maximum for each computer.
As with Episode, I tested encoding 16 files to the three aforementioned formats. Probably because H.264 encoding is so inefficient, the Z400 proved only 37 percent slower than the Z800.
I probably could have squeezed (no pun intended) more efficiency out of the Z800 by loading multiple versions of Squeeze and running the H.264 encodes simultaneously, and certainly that's what I would do if I had multiple large files to encode to H.264, or even a single file to encode to multiple flavors of H.264. But in my standard tests, the Z400 again proved a very efficient, low-cost alternative to the Z800 in low- to mid-volume encoding shops.
Where Episode Engine and Squeeze are relatively efficient with multiple-processor computers, not all encoding programs are. As an example, Microsoft's Expression Encoder 4 Pro encodes a single file at a time. I encoded a single 6-minute test file to VC-1 format on both computers. Table 3 shows that the Z400 was only 9 percent slower.
These tests demonstrate what I stated at the top: that application performance on multiple-processor systems depends primarily on the multi-processor efficiency of the programs that you're using. If you're running Episode Engine in a relatively high-volume operation, the Z800 makes the most sense. Otherwise, at least respecting these encoding tools, the Z400 struck a great balance between affordability and performance.
Next time out, I'll present comparative performance with Adobe Creative Suite 5 and Rhozet Carbon Coder. See you in two weeks.









