Hello, I am building a workstation for the use of Solidworks 2020. What do you have to suggest me for running easily at least 20000 components? Thanks.
Easily running 20,000 components?
CPU: Intel i9-10900k
Graphics: NVIDIA Quadro RTX4000 (or greater)
Storage: 512GB+ M.2 PCIe NVMe SSD
I don't think there is a system in existence that will "Easily" work with a 20K assembly in Solidworks. Furthermore how easy an assembly is to deal with will be affected far more by the practices used while creating it rather than system performance. A single poorly modeled part can bring the best of systems to it's knees while an assembly with well designed parts and assembling with many parts can be "Easily" handled.
That being said there are a TON of threads in here on system specs. They largely end up boiling down to this.
CPU - Fastest Core Hrtz you can get.
RAM - As much as you can afford, min 16 GB prefer 32
HD - SSD, You'll get much debate on form, type etc.
Video Card - Argument over which to get, Professional (Certified) or Gaming. Most would agree that if you have the money get the best Professional card you can afford. If you're on a tight budget and feel comfortable messing with PC's you can get the best Gaming card you can find.
Please provide more info such as do you do any rendering or FEA or any stimulation - plastic or metal parts: how complicated the parts
I need my Solidworks(specificaly 2020) to be fluent for CAD designs and to be able to render these designs. I am not that interested into running simulations, FEA, CFD, explicit or anything else like these. I need to be able to design a CAD model and render it, containing not so complex geometric designs but with a lot of pipes, flanges, bolts etc. There may be some complex geometries from time to time but these complex parts or assemblies are less common. I currently run my projects containing from 20000 up to 150000 components on an intel i7 5h gen, gtx titan x, 64gb ram and 512 ssd and it can not handle such big projects. Thanks already everyone that replied and helped me so far, I would like to hear more suggestions and I'm happy to provide more info if needed.
The only real change to increase your performance hardware wise would be to get the best processor you can, currently this would be the Intel i9-10900k. I would also recommend that you use a professional grade graphics card instead of GTX branded cards so that your graphics will be fully supported.
Know that as mentioned above, a large part of performance is not the hardware you use, but the modeling practices you are using. I would recommend that you do some research into best practices for large assemblies if you are concerned about this.
Alex Ghost wrote: I currently run my projects containing from 20000 up to 150000 components on an intel i7 5h gen, gtx titan x, 64gb ram and 512 ssd and it can not handle such big projects.
Alex Ghost wrote:
I currently run my projects containing from 20000 up to 150000 components on an intel i7 5h gen, gtx titan x, 64gb ram and 512 ssd and it can not handle such big projects.
When you say "Can't Handle" does that mean can't run at all or is slow. At 150K components I can't see anything running it "Well". For the most part you're running into software issues rather than hardware issues. Since SW can only use one core you have a whole lot more "Horsepower" in your hardware, it's just that SW won't use it.
I just did a single part array that had 54 x 54 multibodies and then dropped that part into an assembly and arrayed it 54 more times for a total of 157464 1" cubes. My computer did not handle it "Well". I can't even imagine what it would do if I tried to mate 150K cubes together.
Moving from an i7 to an i9 and all the other upgrades will make it a little better but it's not going to turn a slow moving slug into a seamlessly working machine.
I'm running an older CPU but it's still almost 4Ghrtz with an SSD an a Quadro P4000.
Should have try 54 x 54 McMaster Carr bolt.
Frederick Law wrote: Should have try 54 x 54 McMaster Carr bolt.With thread.
Frederick Law wrote:
And then if I turned the detail to as high as it will go I could watch my monitor melt :-)
150K parts is something I can't imagine attempting to work with in SW. I would think you would have to break it down into manageable S/A and then place them in the main assembly as greatly simplified configurations.
I've had assemblies where I've had to pattern 20K rivets with their counter-rivets and to make it workable I created a promoted assembly with two display states, one being a display state that hid both so that I would not see them in my assembly but could still have them in my BOM.
Solution worked great, I showed the first, patterned, changed display state after pattern and voila.
Saying "can't handle" I mean that it works but moves like a slug. For example when I try to open the assembly it may take about 20 minutes just to open it. When I open it moves very slowly, it may take 5 minutes just to zoom on a specific point to see something and if I try right clicking on a subassembly to open and work on it, it takes long enough for the program to do this, and because of these two or three large assemblies opened at the time there's a great possibility in general for solidworks to crash. I generally close the previous assembly when I open a subassembly trying to keep the program occupied at minimum.
as usually, the term "big assembly with xxx parts" mean nothing,
and the usual questions (already talked in many others threads) are the same :
how % of library-part will be in your assembly ?
are library-part "imported" or built with SW feature ? the best is PRT with config (excel)...
are library-part optimized for "geometric and level of detail" in your final assembly ?
will you use (over-use) the "heavy (forbidden) features" in sw ?
(Ext.Ref, geometric ellipse, fully detailed mesh, real 3d threads, etc...)
which workflow will you use ?
simple : asm (and you charge the donkey with heavy load for full part loading)
simplified with one level-step
simplified with two level-step
simplified with hidden details
the DOT graphic setting may impact total performance (because of triangles)
example of 20 000 parts project with same hardware :
may run good in you case, workable
but if DOT graphic is set to high will be slowly, with same files, same hardware...
i prefer use DOT-graphic, to 50-60%, i'm not a fan of 0%, also it depend the different scale and details you need, what will be the bigger distance in assembly 1m, 100m, 1000m,
and if you have cylinder or curve, and you don't what to see them with "not enough triangle / facetized)
another example with all i talked :
with the same hardware, a project with 20 000 parts will be "OK"
with the same hardware, a project with 1 000 parts will be "Awful".
Retrieving data ...