specAPC was originally supposed to be a day in the life type benchmark. The problems I have run into with it have been due to some kind of incompatibilities with my machines. This seems to be a Windows issue. And then, as a benchmark, it moves the mark every time it comes out. If it would run on every version of SolidWorks since it's inception it would be a good benchmark of not only hardware, but of the software and operating system. And with it's heavy emphasis on graphics it's results can be skewed by putting a really fast graphics card into an otherwise mediocre system. One of it's advantages has been the five iteration test which seems to hammer hardware more than any other program I have seen. But the later versions take so long to run that even this facility is too time consuming to use in setting up a machine. specAPC also controls a fair number of SW settings for consistency. I tried my hand at writing a very simple benchmark that controls as many settings as I deemed necessary while insulating the test from the effects of graphics cards and ongoing releases of SolidWorks. It is simple and therefore only tests a limited area in SW. However, those areas are fundamental to what SW does (sketcher, extrude, intersect, parent/child relations). I would like to find time to add more tests to it for drawings, assemblies and such. Ideally a benchmark would work over a much broader portion of SW. Ship in a Bottle is good and exercises graphics more. I put up a procedure for testing with it, but as you mentioned, not many follow it or even Mike Wilson's procedure which is a subset of mine. Bottom line still is that on benchmarks that run across releases like Ship in a Bottle and STAR 2006 is still slower than 2004. However, I have also found that in areas like assemblies 2006 seems to be faster.
