Although I am also a bit biased by profession, I would like to name a few dimensions which makes a comparison between programs running under VirtualPC difficult:
1. the user interface capabilities of some of those programs differ significantly. Cross application drag & drop, visualization of data structures and properties, etc., none of such are without demand on computer power.
2. the "business logic" that is, the underlying data model, its representation and the functional level working upon these data structures also differ extremely. Segmentation on character, frame, millisecond level, hyperlinks and semantic networks are a few functions which need their processing power and memory.
3. the range of data types accepted for analysis: text, rich text, rich text with embedded objects, images, audio and video streams also accounts for differences in processing resources needed. Coding of a 2 GByte video, a large MP3 file or a rich text file with embedded Powerpoint and Excel sheets with hundreds or thousands of virtual segments is of course a different business than managing plain text segments.
Trivial, but just like a 4-wheel drive car needs a few more drops of gasoline than a limousine there are different resource requirements for software products.
By the way, ATLAS.ti also accesses "documents on-the-fly" while using an intelligent caching procedure at the same time to keep a good balance between memory usage, disk access and user interface responsiveness. Disk access seems to be one of the bottlenecks of VPC.
The problem with VirtualPC is
a) having the correct settings including allocation of memory and disk space,
b) that Windows XP does not seem to be the most perfect OS to run on VPC (W2000 seems to be the better choice)
c) that a complex software like VPC (compared to software we are concerned about) does not only add a processing layer with the mentioned slow down and its own memory requirements, but as all complex software also reduces robustness of the overall system. This may not be dominant if using only the simplest Windows features but complex software also uses more sophisticated Windows technologies (like multimedia, COM, Structured Storage, etc).
When searching Google for best practices with VPC you get quite a few hints on optimizing this software. But of course, unless you need close integration into your Mac world a cheap PC might be the better choice. Maybe there is already a PC hardware module to be connected via USB or firewire just like an external drive, which - by the way - is also recommended to host the VPC disk image to speed it up.
Some links delivered by Google are:
http://support.microsoft.com/default.aspx?scid=kb;en-us;824355
http://www4.tpgi.com.au/users/aoaug/mac_osx.html#vpc
Another good article you may want to read is offered by the vendors of maxQDA:
http://www.maxqda.de/maxqda-eng/download/MQonApple.pdf
We will also investigate the VPC performance topic further.
No doubt, the best thing for the users would be native versions of the software in MacOS, Linux, etc.
We currently fathom reimplementing ATLAS.ti in Java, which is quite an enterprise and nothing to wait for with the next release, but being locked into the Windows world is sub-optimal for both users (including Windows users!-) and developers.
- Thomas
At 03:07 08.10.2005, you wrote:
>Last August, we have done a 2 days training in a lab of 15 researchers, all equipped with Mac computers and Virtual PC, all of them running QDA Miner. I did not witness it with my own eyes, since I was not the one who went there for the training, but my employee who went there just told me that the program was running very smoothly and was quite responsive.
>
>I can think of several factors that may explain why your experience was so different with Atlas.ti. First, the hardware and OS versions may be different. What was the processor speed? how much memory was available for Windows?
>
>Second, there is probably some differences in how each program uses the system resources and memory. According to the CADQAS Comparison performed at the University of Loughborough:
>
> http://www.lboro.ac.uk/research/mmethods/research/software/caqdas_comparison.html
>
>there are important differences in CPU usage between programs, with Nvivo and Atlas.ti requiring high levels of CPU usage, while QDA Miner requiring low CPU resources when running the same project (like MaxQDA and Qualitan). Part of this difference may be related to how projects are handled in memory, which brings me to the last point: the size of the project. My guess is that some software require a large portion or even the entire project to be held in memory, while others (like QDA Miner and if I'm right, MaxQDA, access the documents on the fly, only when needed, reducing the memory requirement. If a project is too large to fit the computer memory, Windows have to constantly swap part of the memory to disk, making the computer running very very slowly. In such a situation, one may think the computer has frozen, and a simple text search that would normally take a few seconds may take several minutes if not hours.
>
>My experience with Windows emulators on a Mac is that it is at least 4 times slower than running the same application on a PC running at the same speed. A fast application on a PC will run more slowing on the Mac and a slow application on the PC may be sluggish to the point of being very frustrating, so if processing time becomes important and if the application does not seems to run fast enough, the best solution will be as you suggested, to buy a cheap PC.
>
>I guess the best recommendation I would give to Mac users who are considering running a Windows QDA software with Virtual PC is to download trial versions and test them on a sample project about the same size as what the one they are expecting to have.
>
>Normand Peladeau
>Provalis Research
>www.provalisresearch.com
>
>
>>At 10/7/2005 02:22 PM, you wrote:
>>Normand wrote:
>>
>>"Actually, every QDA Software running on Windows should work on a mac with
>>PC emulator. We didn't test the version 5.0 of WordStat but we do have
>>clients running WordStat 4.0, Simstat and QDA Miner on a Mac with a PC
>>emulator. And I see no reason why this would not be the same for Atlas-ti
>>or Nvivo or any other QDA software."
>>
>>For a long time I tried running Atlas.ti on a Windows emulator (Virtual PC) and it was a very
>>unpleasant, frustrating experience. I eventually gave up. In general, I find Virtual PC to be incredibly
>>frustrating to use. The money spent for Virtual PC is probably better spent on a cheap PC.
_______________________________________________________________________
„Computers, like every technology, are a vehicle for the transformation
of tradition“ (Winograd & Flores, 1987)
ATLAS.ti Scientific Software Development GmbH - Berlin - www.atlasti.com
|