Early results; early problems

January to April 7, 2005
Here's a summary of this period:

• There were problems, but I got work done.
• Here's a piece of my original Macintosh work . This is what the Alienware system was to help me do faster and better.
  
• Check this for one of the early (January) files
  
• I got better at this over time. Here's a sample of my later work.
• A "typical" lecture took 35 hours of preparation, taping, and editing. At two lectures per week, I was working 70 hours per week on videos for this course. Add grading, teaching other courses, writing grants, supervising student research, etc., and I was working 110 hours per typical week.
• Intermittent hardware and software failures increased production times by up to 40 hours. Some of that was just wasted computer processing time. Often, though, I had to babysit production to see when a failure would happen. I could get work done in parallel, but I stayed up with the computer overnight too many times.
• The results were encouraging. My students learned more (we're still doing cross-year comparisons of student work to provide publishable evidence of this, but I saw their work over the year and I know what I saw.) Other faculty at Florida Tech were so interested in this work that they wrote a grant proposal with me to extend the approach to their courses. Also, this is a Creative Commons course--anyone can use my stuff in their courses, and other faculty and other commercial instructors have started using it.
• I wasn't Alienware's happiest customer, but if things had continued like this, you would never have seen a blog like this from me.
  

A few more details:

After a basic checkout, I started trying to get the video system working. There were a lot of problems. The machine didn't work well out of the box. It crashed frequently. It corrupted videos. It gave error messages and halted video processing at random times, and would sometimes fail to process a file and other times process the same file with no problem.

Processing of these videos was very time-consuming. It often took 10 hours to render a video and export it to Windows Media format. Adobe Premiere often crashed 80% or 90% the way through--so I wouldn't realize I had a problem until the system had been working 8 or 9 hours. Then I would run some experiments with small files, make a guess about what to do with the course video, and try again.

I'm used to dealing with software and configuration problems, so I went to the Alienware and Adobe (Premiere Pro) and Matrox websites, Googled for additional information, picked the brains of some of my students, and spent a lot of time trying things to see what I could do to get this system working.

Several people reported issues with the Matrox card that sounded like the problems I was having. Troubleshooting suggestions included a variety of updating tasks--I did those. Some helped, but there were still frequent failures. Among the common suggestions was advice to turn off memory-resident programs -- and I did seem to have fewer failures if I turned off Norton and disconnected the network. But this was awkward. I needed Internet connections and I needed Norton supervising them. Even with those restrictions, there were still too many failures and I could never tell when the next failure would hit.

In February, I decided to see what would happen if I disabled the Matrox card. I pulled its cables and revised the system files so that it wouldn't load the Matrox drivers at boot time. Video editing became slightly less smooth, but much more predictable.

There were still intermittent problems. Some were mysteries. Premiere Pro would halt with an uninformative error (dropped frame, or unknown error). Sometimes it would get through the identical file if I rebooted the system or turned off the network. Of these, some were probably related to memory utilization, maybe to memory leaks in Premiere Pro. I was able to improve reliability by severely restricting my use of filters.

Some files reliably caused crashes. Troubleshooting in May, a student of mine was able to show that one of the video files that I had captured through the Matrox card was corrupted in ways that caused Premiere Pro to fail.

The problem was not entirely with the Matrox card. Even when the Matrox card was not involved in capture or editing, there were still some Premiere Pro failures. There were also disk I/O errors, especially with the USB drives I was using for backup. There were intermittent failures to copy files on the main drives, many more than I've come to expect on Windows 2000 (this was my first use of Windows XP), but I could always redo it or do something slightly different and save the file, so it was disturbing, but not fatal.

Some problems were more predictable. For example, running Windows Media Player, exiting, and then trying to work in Premiere Pro was moderately likely to cause a crash.

Overall, I still couldn't predict how long it would take to process a video file (it would depend on whether Premiere and the Alienware system felt like crashing, and for how many times) But, by and large, I could get most tasks done, close to when people were relying on me to get them done.

During this time, I didn't call Alienware for help once. I wasn't happy with the system, but I learned enough on the web to doubt that I was going to get much improvement by pestering tech support staff. These looked like multivendor problems, which are the hardest for technical support staff to solve. I figured that come May, after the end of my Spring term, I'd do some troubleshooting, take the Matrox card physically out of the system, and see what else I could do to improve its reliability.

The fact that I could get work done doesn't mean I was a satisfied Alienware customer.

• I had lost the Fall term -- If I'd known that I wouldn't get the computer until November, I'd have bought something else.
  
• The sales process had left a bad taste in my mouth.
  
• And the system didn't work well. The point of buying from a company that has expertise in video processing was to avoid the multivendor problems that you might reasonably expect when you push a system hard (big processing loads for long periods).
  
• I'd paid a lot for a video coprocessor that didn't seem to add much performance but did seem to trigger a lot of problems--problems that I didn't expect on a system that was supposed to have been fully checked out with the Matrox card--that's what I had paid the big bucks to get.
  

So, I wasn't likely to buy another Alienware machine, or recommend it to anyone else. But until April 7, 2005, I was able to get work out of it.