SW2006 Pre Release is now available.

Paul,

I don't know about you, but as far as I'm concerned. "an exponential increase in performance" is a very strong statement, but not very credible. Kinda like saying you won a gazillion dollars. How much is the "exactly"

According to one guy on the SW hosted group, feature creation, rebuilding, etc. (all the things that really matter) are the typical 10-15% slower than

2005. It'll be interesting to seperate the truth from the marketing spin

Mark

I made a comment previously about the term exponential. It takes at least three points to determine an exponential curve. It is kind of like the Ronco Rotisserie. If you watch the infomercial the phrase, "Just Set It and Forget It" is repeated over and over again. When the unit arrives it has a big yellow sticker on the front saying, "Don't take Set it and Forget it Literally". Since SW has traditionaly had about a 4-5% decrease in performance per release the best they could say is that they stopped the exponential decrease they had been seeing. Their will probably be a statement in the EULA saying, "Don't take exponential increase in performance literally."

If you can't measure it, is it really there?

If you can't measure it, how do you know it isn't?

Just being a smarty-pants. I'm also waiting to see what an "exponential increase" is.

I'd settle for an "exponential" decrease in crashes. Particularly during file save, you know, the thing your supposed to be able to do to make you feel warm, fuzzy, and safe.......

I used to think '04 (been a user since '99) was the worst release ever (stability wise), then '05 came along and I said "let's use it, it can't be any worse".... then it proceeded to crap all over me. If '06 continues the trend, I think I'm done with it, the search for an alternative will start in earnest. I'm already having a look at Visi next week.

I really hope that '06 is an improvement, but after some of the stupid things I've seen get broke in the last two releases, I'm bracing for another disappointment.

Rory

Tool & Die for Automotive stampings Some Molds Check Fixtures

We are looking at Catia v5 soon as well, I know its the same company, Dassault, but they have 5 times as many developement engineers working on it than Solidworks has. Maybe thats why they can listen and implement customer requests.We have had a request to implement a particular sheet metal function for aerospace since v2002, looks like v2006 still doesnt have it.

We went to the roll-out yesterday. As best as I could tell from the presentation, the major part of the "increase in performance" is that they've made lightweight a lot more useful. When the machine doesn't have to resolve a part, it's much quicker. They've also implemented background processes, so that you can keep dropping draft quality views on while it calculates the high quality views, for example. This could be a big deal as more processing can be moved on to other processors or other cores in a given processor.

Jerry Steiger Tripod Data Systems "take the garbage out, dear"

Drawing performance is a big deal and needed to be addressed, especially for large assemblies. I have assemblies that have to be worked on in high quality mode just to see the lines. So I would still have to wait it out.

Paul,

bench times

Ship in a bottle

2003 28.5 2004 30.25 2005 32.25 2006 31.4

Star 2.1

2003 36.0 2004 39.65 2005 43.7 2006 45.3

Patbench

2005 6.625 for 8 2006 6.625 for 8

Not as bad as I feared. The Star 2.1 might be the result of percieved sluggishness in the 2006 sketcher. This may be better (or worse) on SP0.

The 2006 Ship times are probably partially the result of optimized and re-coded graphics.

It looks like they may have, at least, stopped their downward slide.

I went to a Solidworks roll out this morning at the Anahiem convention center. It was hosted by Go-Engineer. The guest speaker was Chris Garcia VP of R&D at SW. He outlined his plan to overhaul SW with regards to performance an quality. There were a total of 25 major functional areas scheduled for optimization. 10 of them made it into 2006 first release, and deal largely with assembly and drawing performance. I have to say, the on screen real time demo "looked" impressive.

The most encouraging thing was that this guy "appeard" to have a plan, and it made sense to "me".

Regards

Mark

I appreciate it. It is no small thing to run a benchmark on four releases.

The word exponential doesn't quite pop out of the numbers.

If you look at these in a practical light, Ship in a Bottle and STAR give some sense of where rebuild times are going for parts. STAR especially is as free of graphics and user interface effects as possible. Probably not a big deal for most unless involved in the mold side where part rebuild times can be a big problem.

But Patbench effects a lot of people who will want to use patterns more often and they will still run into that wall. It seems to me two walls need to be surmounted with Patbench performance. First, the rate at which the times climb IS exponential with the number of levels. This of course means there will always be a limit. Second, Patbench will run SW out of memory at a high enough setting. This is another big wall both from the standpoint of making parts that can be opened on any system and on speed because when a single part feature can consume all a system's free RAM there is a serious limit.

It is nice that Mr. Garcia has a plan but it still looks like the tough stuff is being avoided for now. Just the term optimzation is a tipoff. You can optimize a Model T all you want and it still won't manage to pass an F1 car with the engine idling. SW really needs to get into some cutting edge stuff when it comes to performance and they had a start with Cosmic Blobs.

Let's face it. 3D CAD has got to be one of the most demanding tasks in the history of computing because ultimately more and more will be modeled at a single time. SRAC made a breakthrough 10 years ago with their FFE solver and it revolutionized FEA. SW revolutionized a clunky Pro/E interface about the same time. But precious little has seemed to happen in the CAD world since then except for and exponential increase in hype and a steady decrease in the truly revolutionary. SW doesn't even seem to be able to borrow from their parent company the technology that allowed an entire 777 to be modeled at once.

MM,

I recall my VAR saying SW overhauled the 2001 release. This was supposed to improve performance and of course added 150 plus improvements. SW just loves to blow smoke up everyone's posterior at new release time. Their past track record makes the VP's comments regarding performance and quality overhaul hard to believe. Sounds like this guy hijacked a few buzz words from the newsgroup or maybe a lean manufacturing seminar for programmers.

I'm sure you remember the three Amigos pilgrimage to SW and what was reported back after meeting with these same folks. Sound familiar

Kman

I try today with an average assembly with 2000 components. I open it in Lightweight, place the views. Placing views run pretty fast. In less then 10 sec, I can place 4 views. It displays first in draft quality while it's still computing the high quality. I can see in the Task Manager that my system is running at 100%. It still let me moving views around. After about 2 minutes, all the views start showing all details in high quality.

Paul,

The only way to increase SW performance "IS" to optimize. Many times this can, and often does, result in the complete rewrite of components. Starting over is about as radical as it gets. If making your code more efficient isn't the answer, than what is ?? How would you do it Paul ?? Give up and walk away ?????

I'm not taking their side in this, they still have alot to prove. This Garcia guy was all nuts and bolts. No fanfare or wiscracks or buzzword BS. That's the reason I paid attention to him, but I'm in the show-me mode.

Mark

"Kman" a écrit dans le message de news:ZbOpe.1509$ snipped-for-privacy@newsread1.news.atl.earthlink.net... SW just

On the other side, they are not going to say that their next release is a dog. At leas they keep trying. My solution is: drop subscription, stick with a release you like, and upgrade software and hardware simultaneously to compensate.

"what Andy gives me, Bill takes" (or something close)

Mark,

If you review the What's New Guide for 2006, the performance improvements were specifically targeted towards drawings and assemblies. In these areas, performance is dramatically improved.

For example, many assembly operations that previously required components to be resolved have been eliminated. These include mating, component insertion, interference detection and section view creation. Lightweight drawings now support the creation of section views, adding annotations and ballooning, without resolving components.

While I agree, work still needs to be done, these improvements are significant.

CG

Wait till you try plotting these drawing views that are loading in draft quality. Half the drawing views plot out in draft quality now. So much for improving speed of the drawings when I have to keep reloading my drawings to get them plotted out correct

was speed up the initial placing/moving/detailing of views. If you want high quality prints you will have to wait until it's fully loaded. It speeds up the initial part of making the drawing.

The demo I saw had a 6800 piece assembly of which the AE created 4 views and was dimensioning in around a minute. To me, that was leaps and bounds over where it has been over the last couple of years. He also inserted another instance of that assembly in a top level assembly (13600 parts) and was moving the sub assemblies around very smooth.

I'm hopeful, that's all I can say.

Later

j wrote:

With all their years of trying you would think SW would be close to perfection. What are they really trying to do?

That is fine, except if you decided to reinstate the subscription, say a couple of years later they demand a penalty fee. A gentler form of black balling and extortion for leaving the fold.

Kman

I don't know if I would call all this lightweight stuff optimizing. Are they going to optimize like gamers optimize and code inner loops in machine language and access the hardware directly. I doubt it. I don't disagree optimizing will increase performance but you have to put it in perspective. First ask the question where does performance have to be to get the job done? We are creating proposals in Fast Cad because SW can't put a proposal layout together in a timely fashion. That is true lightweight. Dump the 3D model until it is really useful. If optimizing won't provide the performance necessary then what? The code will be faster, but not fast enough. So how do you know what fast enough is? That is what the real focus should be on. It gets rid of the smoke and mirrors approach of judging one years speed increases on last years losses. If Garcia is really as nuts and bolts as you say, he should agree to this. We have to find measures of performance that can be relied on to reflect what we experience and we have to get SW to agree to meet certain measurable goals.

I'm not going to try and answer the how to do it part of your response. I know just a little bit about how SW works and how CAD works. I have to really struggle to write the simplest code. But there are those who do know these things. I mentioned Cosmic Blobs because SW is looking at alternatives that are radical departures from what they are now doing. All we can do as users is to determine how fast is fast enough and measure what we are given by that measure.

Paul,

You can't access the hardware directly in Windows. Every thing gets abstracted through the HAL

I doubt it. I don't

In this case it sounds like your using a screwdriver for a hammer.

Ya gotta start somewhere don't ya ??? Just the fact that we didn't lose much is a start. We'll have to see if it's an accident or part of the plan won't we. They still have time to screw it up before SP0.

In any event, we're limited to the maximum speed of Parasolid to solve things. I've seen UG solve complex parts much faster than SW. There are a whole shitload of possible variables, not the least of which is hardware.

I agree that a common metric should be used. Most benchmarks are for hardware, and are intended more to muddy the waters than to prove anything

I'm no progammer either, but I recognised some of the techniques he talked about from my UNIX days. They were the types of things that allowed you to design a whole aircraft on a 60mhz UNIX RISC box.

Solidworks is a very "serialzed" program. Not only in the data structure, but also in the way it functions. This is one of the reasons we watch the hourglass so much. Everything it does is a bottleneck for the next thing. It's a matter of architecture, it doesn't have to be that way.

So, besides being argumentative (which is out of character for you), what's your point ??

Regards

Mark