SolidWorks Benchmarking

"If you can't measure it, is it really there?"
Ship in a Bottle
 Download Mike Wilson's Ship in a Bottle from the link to the left.
Make four runs with the following settings:
1. TOOLS/OPTIONS/DOCUMENT PROPERTIES/IMAGE QUALITY -- Shaded to the far right
    Have VIEW/DISPLAY/SHADED WITH HLR checked.
2. TOOLS/OPTIONS/DOCUMENT PROPERTIES/IMAGE QUALITY -- Shaded to the far right
    Have VIEW/DISPLAY/SHADED WITH HLR unchecked.
3. TOOLS/OPTIONS/DOCUMENT PROPERTIES/IMAGE QUALITY -- Shaded to the far left
    Have VIEW/DISPLAY/SHADED WITH HLR checked.
4. TOOLS/OPTIONS/DOCUMENT PROPERTIES/IMAGE QUALITY -- Shaded to the far left
    Have VIEW/DISPLAY/SHADED WITH HLR unchecked.

When reporting results for 50 iterations give the following information:
CPU make and model
Memory make and quantity
Graphics card make, model and driver
The results of the test as:
time1/time2/time3/time4

Note:
The different settings in 1 thru 4 above put less load on the graphics card as they progress. By comparing the different settings you can get some idea of the effect your graphics processor has on the whole process.  Never save the part, it is not necessary and may give differing results from run to run.

Discussion:

Ship in a Bottle is a single part in 2001+ format IIRC so it will run across releases of SolidWorks and on different service packs. As such it can evaluate not only hardware performance but also software performance from release to release and service pack to service pack. It is made up of a single part using some fairly complex SW features and is driven by  equations that have built in logic to change certain dimensions each time the part is rebuilts. A macro makes it possible to run a set number of  rebuilds. Results are typically reported for 50 iterations. Ship in a Bottle will not provide much usefull information about assembly or drawing performance.

SPECapc SW
 Dowload from the SPEC.org website. Be sure to select the SolidWorks version of SPECapc. It is a large file and typically only works on current releases of SolidWorks. Run time is rather long and the results are biased towards graphics performance.

When reporting results be sure to report both the individual run results and the averages as well as a description of the hardware it was run on.

Discussion:

This benchmark consists of a variety of parts and assemblies that go through various gyrations that suppossedly represent a typical workday. Because the subject matter of the benchmark keep changing from time to time and because it is rewritten for major releases of  SolidWorks it is difficult to use it to evaluate software performance. However, it does give hardware a good workout and this is especially true of the graphics side. A perusal of the source code suggests that the results are heavily biased towards graphics performance which means that a high score with this benchmark will not necessarily translate into a good performance with complex parts or resolving assemblies with high parts counts.

SPEC also provides a comparison of benchmark results on various hardware platforms.  This may make it usefull in determining which system meets a particular need if the results are there.
STAR 2.1
 Download from this site. This benchmark consists of a single macro that creates it's own part during the run. Because it is designed to run as efficiently as possible it may be necessary to manually change some settings in the way macros run before it will work.

Load the macro into the macro editor. From the menu TOOLS/REFERENCES check:

SldWorks200x Type Library   
SolidWorks 200x Constants Type Library

 where x is the release of SW you are running.

Then save the macro and run it the usual way. Typically level 5 is all you need. Make at least three runs and average the results. Report the total time, rebuild time, version of SolidWorks, service pack and hardware description.

Discussion:

This benchmark performs a few very basic actions that all users use daily. It basically just draws a sketch, extrudes it and then intersects that feature with surrounding features. When you think about it that is all you are doing during a part modeling session. The benchmark decides which release of SolidWorks you are on and uses API calls appropriate for that release. It has been tested on releases from 2003 to the upcoming release of SolidWorks. As such it can be used to compare the performance of different releases of SolidWorks. The benchmark creates the part off screen so that graphics perfromance has very little to do with the results. In this way it is a good indication of rebuld performance and CPU/motherboard performance.

Bug reports and suggestions would be appreciated.

PATBENCH
 Download from this site. This benchmark consists of a small square tile with an array of holes in it. With each iteration the number of holes is increased. Load the part and run the macro. Typically results for 8 iterations will be reported along with the hardware it is run on.  This benchmark is compatible with SolidWorks 2001 through the current release.

Discussion:

It was found, quite by accident, that the pattern feature is quite memory intensive. In fact it has so far been impossible to run more than 12 iterations before consuming all memory on a 32 bit operating system with the 3 gigabyte switch enabled. Because it does so much memory I/O it is believed to be a good test of the I/O capabilities of a CPU and motherboard combination.  Since there is an upper limit to the number of instances of a feature in a single pattern this benchmark makes a good illustration of that limitation.


Rebuild
 Download from this site. This is not really a benchmark, it is a macro that simply turns on Verification on Rebuild, does a forced rebuild, and then turns off Verification on Rebuild. It then reports rebuild time in a small messagebox on screen. This macro is included with the benchmarking macros and programs because it can be used to run parts and assemblies known to be cpu intensive and check the effects of different service packs, releases and modeling techniques.

Discussion:

A message box was included to make elapsed time effects apparent for any SW file. Comment out the MsgBox call and you have a nifty macro for rebuilding during complex geometry creation without having to dig down and turn Verification on Rebuild on and off.
TSToolbox
 Download from this site. This utility consists of three tabs. The first tab is of interest to benchmarkers as it reports both rebuild time and time to refresh the display. It also reports file loading times. TSTool box contains a utility on the third tab to tell the version history of a SolidWorks document.

Discussion:

This program was originally posted on the SolidWorks website in the gallery. AFAIK there is no authorship or copyright contained in the archive and it is posted here in the belief that copying is not restricted. Please notify the author of this web site if you have reason to believe that is not the case and it will be taken down. Use at your own risk.

This page Copyright 2005 by Engineering Transport, LLC.
The STAR, PATBENCH and Rebuild  benchmarks are Copyright 2004, 2001 and 2005 respectively by Engineering Transport, LLC and may be freely used for non-commercial purposes.