SWW2005 - Performance

Probably the most talked about item in sessions I attended and in
private conversations was performance. Noticable for it's absence on
the enhancment survey was performance.

There was a presentation on What's new in 2006. Part of the
presentation was a performance enhancement. Whether this enhancement
which had to do with large assembly drawings will have a real impact
remains to be seen. If it is like half the performance enhancements
listed in the short history presented prior to the unveiling it will
have minimal impact on most of us.

I also attended a session on Large Assembly Performance. The gist of
the presentation could be summarized:

1. Throw fast hardware at it.
2. Don't use part features that slow down SW.
3. Use configuration to defeature parts for large assemblies.

I would respond to this:
1. I already have a really fast computer. It would cost $1,500 to
upgrade and then I would get a 2-5% improvement.
2. I use the features to get the intent. Anyway, my parts are mostly
prismatic and I already use good practice in things like putting
fillets at the bottom of the feature tree.
3. If I have to start adding two extra configurations for every part
configuration I use then I will be using up a lot of time on the front
end to save some time on the back end. Anyway we want to render the
large assembly for our literature and having blocky looking parts won't
be allowed.

And the bottom line is:

Unless they can speed up the process in software from parts to
assemblies to drawings by a factor of ten "this dog ain't gonna hunt."

 

Your not alone on this issue P. If you would like to read some good
posting about SolidWorks great performance you can check out the post I
placed on 2-2-05 titled A SolidWorks Performance Story in this group.

See ya.

hunt."

 

Who could miss that story.

A little history. Back in 1999 at SWW in New Orleans I attended my
first Round Table and brought up performance. I was kind of a lone
voice then, but others did chime in. And a very useful Inside
SolidWorks session was done by a SW employee which opened my eyes to
which features and techniques were expensive and which weren't. As I
recall I sat next to Matt L. in that session, but I could be mistaken.
I have since been harping on the issue incessantly. And when the
SolidWorks Wish List existed, performance was again the number one
enhancement. Over the years I have posted results from benchmarks that
I do on real parts (that can't be shared) and from benchmarks on things
that can be shared. One thing is clear. While real, across the board
performance has steadily dropped when normalized to CPU performance,
the models the general user community creates have gotten steadily more
demanding. Yes there have been band aids like Lightweight assemblies
and Rapid Draft/Detached Drawings. But in terms of brute force, CTRL Q
regeneration performance things have gone downhill. When I went to
Concord with the 2 other Amigos performance was high on the list (along
with stability). What was provided in answer to that was mostly band
aids and hype. I don't think SW and the user community use the same
measures of performance. There isn't a benchmark out there that will
work equally well on SW98 and SW2006.

So now we are in a hiatus of processor performance and the next
generation of fast hardware appears to be going the 64 bit and
multiprocessing route, neither of which SW is prepared for. In 1996
when I started using SW a 160Mhz CPU with 128Mb of ram gave decent
performance on the parts and assemblies I was doing. Now days a
screaming AMD FX53 with 2GB is just adequate.