New computer justification

I'm not a boss put when I discuss this with my boss the question always is
"will I get drawings faster" and the answer is usually "not really". What
usually happens is that we do more and more sophisticated things in our
models (ie we grow up to our hardware capabilites) once again reaching the
limits of the machines, that allows us to get a better product out the door
with fewer errors. Also that little bit of processor lag allows for a bit of
thinking about what you are doing and your next step.

Don't get me wrong I am all for the fastest latest and greatest I don't
think the issue is as simple as stated and any good boss will know that.


Steve R

 

This has been rolling around in the back of my head a bit. Just timing the
rebuilds will underestimate the impact of processor speed, just like task
manager overestimates. What I mean is, timing rebuilds only does not
capture the friction in the UI, like waiting 10 seconds to place a
dimension on a drawing of a large assembly.

So, I think the best we can hope for is an estimate, with task manager
reporting the upper limit, and rebuilds giving the lower limit.

I wonder if an event is raised at the beginning and end of view
manipulation. Subtracting that from task manager's report would get us
where we want to be, assuming that we wouldn't spend less time trying to
get the model positioned with a faster CPU.

I may be thinking too hard. Waiting for rebuilds should show a significant
return on a new workstation even though it underestimates the issue. I
suspect that a full time user on medium to large assemblies should be
replacing a CPU at least yearly. FEA users probably even more.

 

In the absence of precise tools to measure, I propose we find an acceptable
percentage or range of percentages to take from task manager / performance
log to use as an estimate for this sort of thing. I'm thinking 65-85% of
reported CPU utilization is reasonable, depending on the user. Bias toward
the high end for making drawings.

 

jimsym,

I disagree with your statement about everything being 3x faster.

I won't be able to hide parts 3x faster.
I won't be able to rotate my parts 3x faster.
I won't be able to draw sketches 3x faster.
I won't be able to move my mouse 3x faster.
I won't be able to navigate menus 3x faster.
I won't be able to read my email, web pages, or this group 3x faster.
I won't be able to create office documents 3x faster.
etc.
etc.

Don't get me wrong. I want, and need a new computer. And I know that
other aspects will be faster. But I have got to be realistic in what
is actually being our bottleneck, so I know how to show the
justifications for it. It would seem that any activity that taxes the
processor is our major bottleneck. That would include anything from
rebuilds, to creating animations, to doing geometry compares, to
rendering scenes, etc., etc., etc.

With that being said, out of everything we do that really puts a load
on our system, rebuilds are by far the most commonly done. This is
more than likely true for most people.

That is the reason for wanting a utility for measuring rebuild times
ONLY.

And by the way, the computers that they did look at for us at one time
were extremely, I mean extremely, high end system. I don't remember
exactly what they were (been a few months). But I read the specs, and
it was all like the best of the best hardware. I believe the systems
were in the $5000-6000. They will do that so that they won't have to
replace them as often. That's IT's choice. But I'm not going to
argue with them.... If it were me, I would buy something like you
described, more often. It would definately be easier to justify.

 

Not quite what you need, but there must be other people in your company that
have obsolete machines. So upgrading the engineering computers is like
upgrading a large number of machines as your machines will replace other
machines in your company.
That's how I justify changing ours every other year.

 

Seth,

Maybe you are going at this all wrong. Engineers understand numbers
like this, management doesn't. You might take the tack that IS
doesn't have the qualifications to manage CAD. Of course I would
present this a little more tactfully or you might be running SW on a
TI55.

Over the years SW and in particular Greg Jankowski have done a lot of
work trying to support the management side of SWX. I know at the last
SWW the CAD Managers Boot Camp addressed a lot of issues that come up
in this area. Either your IS department gets qualified to manage SW or
Engineering should be given the latitude to do it themselves. Your
boss might be able to get a CAD Manager easier than a new computer.
The CAD Manager will then have the ear of management and be able to
secure IS cooperation.

TOP

 

And by the way, the computers that they did look at for us at one time

You might want to have them look into that to help with cost justification.
A $6k system is going to be wated cash, because there will be a lot of
money spent on things that won't help SW. Consequently, they will still be
outdated in the same timeframe as a less well equipped machine.

Perhaps when you figure out how to measure rebuild time over a day (or
maybe a project), you could submit that information with a suggested
hardware configuration. The hottest possible SW workstation shouldn't be
more than $3-3.5k, unless you build an 8 or more core monster for large
analysis projects. More money that that and they'll be wasting it on things
that don't help, like SCSI RAID 5 systems and Quad SLI and other silliness.

 

Dell has a very fast machine for about 10k. But to do CAD you have to
replace the video card. For what Dell is doing I could hand build a
machine for a lot less.

TOP

 

Steve, what you just hit on is "How Good is Good Enough?" That really
applies to all jobs.

I do really feel that training can help eliminate unneccessary redraw
times and that a monitor to track them would at least quantify what %
of a days work is waiting for redraws as they are likely to be a
similar % of time regardless of whether the user is doing simpler or
more complicated constructions.

Bo

 

Drawings generally won't get faster, they will hold as SW continues to
require more and more resources.

TOP

 

I am thinking of continuing my practice of generating most of my work
in earlier releases in SolidWorks as those releases work fast for my
designs, without problems, & on slower processors.

Then just importing them into the current release on an as needed
basis is something that is easy for me to do. Obviously an
engineering group might not want to do this, but a few people might
choose to work that way, and simply take the best of what they can
that suits their jobs.

Bo

 

Seeing that my computer took 540 seconds convinced me that I REALLY needed
to get a new computer.

Have any of you tried it with Verification On Rebuild turned on? I did. It
took 30,388 seconds! That convinced me that I needed to turn VOR off and
remember to check manually after every "significant" change.

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

 

We replaced one of our computers with a Dell 690 that has the FX3500
graphics card. It's fast (about 5X as fast as my old machine) and cost about
that much. I'm pretty sure that I can get nearly the same usable performance
with a Dell 390 and the FX550 that costs a little over half as much. We'll
do some benchmarks when it comes in.

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

 

Jerry,

I have been looking at benchmarks that run across releases of SW. You
might find you can get the same performance gain by going back a few
releases in SW if that is an option. Some people buy new computers so
they can do bigger assemblies, bigger drawings or process more parts.
Some people buy new computers just to be able to do what they did last
year with the same speed.

I recently went on a one week crusade to fix an assembly that had a
reputation as a benchmark because it ran so slow. A week later people
were asking what I did because it ran so fast it couldn't be timed
accurately. Of course I doubt you are messing up the the way these
guys had done because you hang out here and listen to what has been
said by the many contributors.

As far as the specAPC benchmark, we may have found some kind of bug in
it that misreports results.

TOP

 

With the kinds of shapes we do, we usually have to be running fairly close
to the latest release. We can blame that on Mark Biasotti. We managed to
skip 2003 and 2005, because the pain of switching was more than the gain
from new functionality, but since 2006 we've had to keep up with all of the
nice new stuff that makes generating shapes easier.

Yes, that's been a very interesting thread. Fortunately, we don't do too
many "large" assembly drawings, so it's mostly of academic interest to us.

You would be amazed at the some of the screwball mistakes I make!

Bummer! It may not be much of a benchmark, but it is the only one that gets
very widely reported.


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