SW Performance 2009 v 2012

[kellnerp]

I am working on a performance presentation and I ran across this interesting conundrum. I ran these benchmarks in 2009 and am doing so again in 2012 on a much faster machine. The rebuild times don't improve as much as the Punch Holder benchmark would suggest they should. 600s vs 48s

Does anybody have any thoughts on what might be going on? I have a hypothesis but would like to get some feedback.

Simple Cube rebuild
The part is a simple unit cube constructed various ways (square sketch using lines, square sketch using 2pt splines, loft, sweep, etc. )A macro is used to run and time 100 rebuilds and also SW Feature Statistics is used and then multiplied by 100.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

Phenom IIx6 1100T = 8GB = FX1400 = XP64SP2 = SW2009SP3
"Node news is good news."

 

[jefrado]

Not surprising.

We made the mistake of upgrading from 2008 to 2009. It ran our macros at least 20% slower and recursives no longer worked so I had to rewrite a lot of stuff to get around it.

Eventually, I know, we are going to have to upgrade, but we will put it off for as long as possible.

Its really a joke. I asked about just getting support for 2009 and they wouldn't do it. I am not sure why I even wanted it, since their support for vba issues is virtually non-existant and there are lots of issues (commands that flat out don't work and etc.). I don't think the dealers have ANY expertise or knowledge in this area and no one ever would put me in touch with someone at SW who had a clue.

As everyone should know, upgrading is VERY risky. Once you are into it, there is no going back since too many files by then will have been rendered unreadable by earlier versions.

I keep getting calls from SW people and I keep telling them the problem and they have no solutions that would make upgrading anything execpt a very large risk and expense for us.

Not a Happy Camper.

John

 

[kellnerp]

I am not sure about recursive calls not working. My STAR benchmark runs on 2009 and it is all about recursion. It will run on 2004 on up.

I have written elsewhere about the real cost of upgrading and ROI.

I did my presentation on 4/17. One of my findings was that things like graphics cards cannot explain the scatter in rebuild times. The biggest influence on rebuild time is the processor core, then clock speed, then SW version. SW appears to get slightly faster each release after 2009. That was a welcome change. But there is a lot of scatter in rebuild times that can't be explained by CPU, clock speed or release. There is no consistent pattern for graphics cards.

The Punch Holder database does not record northbridge and southbridge chips and this might be a big influence on the remaining scatter.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

Phenom IIx6 1100T = 8GB = FX1400 = XP64SP2 = SW2009SP3
"Node news is good news."

 

[jefrado]

What would happen, and it took a while to find out what was happening, was that the recursive would work for the first few passes then starting passing random data or data from previous passes. I was never able to determine.

John

 

[kellnerp]

Sounds like you need to start a new thread for this.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

Phenom IIx6 1100T = 8GB = FX1400 = XP64SP2 = SW2009SP3
"Node news is good news."