We are looking at new PC's and comparing the P2000 vs the P4000. The P4000 is only a few hundred more. We are on SW 2018SP5 and likely will not upgrade again until 2021. I heard that the Graphics card can improve Drawing performance, but I am wondering by how much. I haven't found anything definitive.
Our primary assemblies can have 3,000-5,000 parts and some process layouts can have 6-10 of these assemblies. Many of the parts were created long ago, and with the base part detail set high. There are also a number of highly detailed parts downloaded from suppliers, especially that large warehouse supplier. Some machiens have a number of parts with engraved text. Nothing was cleaned up or simplified.
THIS WILL NOT CHANGE any time soon. I know fixing these will speed EVERYTHING up.
Right now with M2000 cards, Windows 10 says they are not being pushed very hard. Assemblies rotate well. Drawings however can be a nightmare.
Does anyone know if better graphics cards, once you are in the midrange level will help the drawings much? Would the difference between a P2000 and P4000 be noticeable?
I have a few side questions too.
Is there a macro or any method to bump the part part detail down for a large group of parts?
I've heard configuration increase load times, but what about configurations for a simplified model, and a detailed one, or engraved text shown for individual drawings, but not shown for assembly drawings? I would think the overall time savings for a simplified configuration would be beneficial.
SolidworksGeneral
In short, no.
Graphic card don't do much with drawings.
CPU speed, amount of RAM and their speed, drive speed could help drawing.
SW2019 supposed to be faster with drawings and large assemblies.
#TASK has a macro for changing level of detail, Ive used it to great effect. I've also read somewhere that having assemblies AND parts set to the same level of detail (LOW) can help too.
Using #TASK you can also consider freezing the feature tree of parts
The opportunity came up to test the drawing updates today. The assembly has about 9,300 components in multiple machines including some modeled mesh guarding. There is a drawing with an iso view in color, and a top view in color. Switching to the top view on sheet two too the following times with the following processors:
8 minutes, Xeon E5-1630 v3, 4 cores, 8 threads, 3.7 - 3.8GHz Turbo, 64GB 2133MHz DDR4 memory
7 minutes, Xeon E5-2687W, 8 cores, 16 threads, 3.1 - 3.8GHz Turbo, 64GB 1600MHz DDR3 memory
All threads pegged when it first started. The 4 core dropped to about 1 core after about 90 seconds I think and the 8 core dropped after about 60 seconds. From then on both had small spikes but were only using 1 processor most of the time.
The load times were about the same on Both PC's. This has me looking more at 6 core i7's now vs. the 4 core Xeon W-2125. Although people still report benefits of ECC memory, and I have experienced some memory glitches, but it has been very rare.
It's a lot more than just threads. When doing some performance evaluations the fasteners are about the middle. Thank you for the thought. It is a good way to get things started at least. I have investigated a macro or program with a script to change the level of detail, but the level of detail appears to be a more difficult setting to adjust that way.
Interesting. Maybe I had misunderstood in the presentation. I will check this out on our simulation station. It's older and slower, but has multiple cores. It also has an older, but high end graphics card for the time.
If you run a macro to just suppress all Helix features from that large warehouse supplier's part models, all threads will suppress also, and your graphics triangle count will be tremendously reduced, improving performance in all contexts. I'd find it hard to believe if the threads were related with mates (but I don't know your workflow), so this automated simplification shouldn't break mates, particularly if you've used axis or origin as the mate attachment entity.
I do not have such a macro. I suppress those as procedure when importing the component. Sorry that I don't have the magic bullet to offer.
Admittedly, this won't change, as you state. It's your call and based on your priorities, but I'd guess that one day spent on this could save at least a week within a year. Fair ROI.
All cores, regardless of view count. You can see it swap between threads. In view of one graph per CPU, you can see it switch use of paired threads as one gets overworked, to share the load.
Does drawing use all cores?
I think its 1 core per view.
I was wrong. It is up to 224 cores (448 threads). Yowza.
RenderBOXX
It's also quite the overkill unless you're using it for Simulation or Rendering.
I disagree only on a minor point. OP's question still is addressed by no to graphics card in drawing performance.
The CPU speed affects assembly performance. The number of CPU cores affects drawing performance. I can watch this occur as Process Explorer shows my CPU cores get hammered in a drawing. Specifying CPU(s) to maximize core count will improve drawing performance.
Conditionally, I'd even consider one workstation spec for producing drawings, and another workstation spec to work on assemblies. This usually is not practical.
There is a hardware expansion you can use for extreme performance, called a RenderBOXX by BOXX. It provides additional expansion CPU's (up to 54 cores?). It is intended for Simulation or Rendering use but would boost drawings easily.
If you can afford, go with 8GB card (P4000) - you'd be pleased with the result while displaying 2D on your computer, especially with complicated parts including curves. You're also doing 3D model, aren't you? then better card will perform better job for you
whatever you add to the file will increase the loading time.However, it'd save you significant time during creating 2D dwg
I don't think you'll notice much even between a low end and top of the line.
Simplify the drawing is the only way.
Removing thousands nuts and bolts make huge different in drawing.
TY I've been hoping to find something more definitive on this. I had heard about graphics cards helping with drawings in a presentation, but there were no real metrics were given and I don't have any notes from the presentation. As I mentioned below, it could be low entry level vs midgrade.
The CPU, RAM, are near max, and the SSD's are not the choke point currently.
TY I've already been over all the "tips". Drivers and such are certified, but most things relating to models or modeling I have not control over. The Rx Diagnostic gives mixed results. Changing some settings withing SW can really skew the results.
Most of the stuff I have found has been ambiguous, and don't really talk about the card. I vaguely recall on one presentation there being talk about the graphics card calculating all the hidden stuff in drawing views. What had me wondering about the P4000 was the larger bus and therefore higher speed, but again I am not sure if that comes into any play with drawings. The drawing performance may also have been relative to an entry card vs mid range.
Well, I ran into this:
Tips to Improve Large Assembly & Drawing Performance - GoEngineer
This is sitting in that article:
However, it is very ambiguous as to whether or not they mean assembly or drawing performance....and by how much.