So there I was, testing the single CPU, six-core HP Z400, trying to figure out the cost/benefits of the system versus the dual-CPU, 12-core HP Z800. When I started testing Adobe Production Premium CS5, the Z400 proved faster in every test. While I love these man-bites-dog stories, you don’t run them without lots of testing, and ultimately, the situation got sorted out. In the process, I learned something about CS5 that might save you some rendering time down the road, which I’ll share herein. I’ll also quickly detail when it makes sense to consider a dual-CPU system for CS5, and when a single CPU will do just as well.
The Z400 review should have been simple; the programs and test files were already installed on both computers, so all I had to do was run the tests and write it up. Instead, it turned into the perfect wave of reviewing inefficiency. The pain is all the more intense because most of the problems were self-inflicted.
The problems started when I unplugged the Z800 to install the same graphics cards as the Z400. When the computer rebooted, I noticed that the on-board battery had died and that the date needed to be reset. I filed these into my mental to-do list, which got pushed into the background when the Z800 wouldn’t boot.
Just to set the table, earlier in the week, my trusty dual Power Mac G5, which was on the same power strip, refused to boot. It still wasn’t working when I started my testing, so I started thinking dirty power. Unlike the Mac, however, the Z800 would try to run, presenting the various preboot text messages. It could even see the C: drive, though I presumed that it had been damaged.
After several hours of frantic fiddling and rebooting, it struck me that some critical setting had been reset when I pulled the plug to swap the graphics card, and the dead battery didn’t preserve the settings. I nosed around the system configuration, saw that the boot order had been rejiggered, and pointed it back to the C: drive. The computer started fine. So if your system won’t boot when the motherboard battery is dead and the power is disconnected, think system reset, not crashed hard drive. But that’s not the key lesson that I want you to learn. Then, I tried to run Premiere Pro, and it kept crashing. An hour later, I was ready to uninstall/reinstall, I and noticed that the time was way off in the system tray. I clicked the clock to open the Time and Date settings and saw that the date was set to 2099, again due to the dead motherboard battery and disconnected power. I corrected the date, and CS5 ran like a charm, but that’s not the key lesson I want you to learn either.
So I ran my tests, which showed the Z800 trailing the Z400, sometimes by 400% to 500%. It’s definitely a scoop, but is it right? I looked back at tests I ran on the Z800 when I reviewed CS5 for EventDV, and I saw that the times back then were much, much faster. However, Adobe had updated the system from 5.0 to 5.02, so it wasn’t an apples-to-apples comparison. I briefly considered using the older numbers for the Z800 and newer numbers for the Z400. But again, the Z400 was running 5.02, not 5.0, which I had used for my initial EventDV tests.
So I thought, I’ll uninstall CS5 on the Z400 and reinstall version 5.0 without the updates, which will give me an apples-to-apples test bed comparison. There was just one problem: CS5 errored out during the reinstall. I ran the Adobe clean script, with the same result. I system restored back to Day 1 and tried to reinstall, but I had same result.
I system restored back to Day 1, ran the clean script, manually deleted every Adobe folder on the system, and—2 hours later—got CS5 to reinstall. But that procedure isn’t the key lesson that I want you to learn.
So here it is. About the same time I got the Z400 up and running, I was on the phone with Adobe, and the technical rep asked me if I had checked the Use Maximum Render Quality check box in Adobe Media Encoder, in the bottom right. I said I had, and—because the Z400 wasn’t up and running yet—I asked what the default setting was. He said disabled, which meant that the Z400 tests had likely been run without this checked. I unchecked the check box, and the Z800 was back, faster than ever.
Here’s the kicker. You might not mind a 400% increase in rendering time if there was a noticeable quality difference, but I compared the files produced with the check box enabled and disabled, and there was no distinguishable difference. I even tested anew at extremely aggressive data rates, and I found no difference there either. The bottom line is that if you’ve checked the Use Maximum Render Quality because it seems like a good idea, you could be costing yourself significant processing time with no qualitative benefit. Run your own comparative tests to be sure, but that check box has already cost me 2 days of my life; I’m not giving it any more.
What about the Z400 versus the Z800? If you’re editing DV or DVCPRO HD, go with the single-core system; with AVCHD, Digital SLR, or Red, go dual core. HDV is right in the middle, depending upon project and output type. You can see the tests results here.