01-09-2019 12:34 PM
Hi all,
I have model with 8x30Mb workstation size. The model has 240Mb size.
Saving / Opening the model takes 3..4 minutes.
Running 3D graphics it gets quite slow and I have more graphics to import.
I like to know what kind of hardware would be recommended, i.e. Processor, Memory, Graphics
Regards
Marco
Solved! Go to Solution.
01-09-2019 03:48 PM
Hi Marco,
I would guess the memory consumption of your model is much higher than 240MB.
Before looking for new hardware, you want to optimize the graphics. Open your object that contains the graphic in a new 3D window, rightclick on the graphic and ->optimize. Do that for all graphics and try different settings.
Mark that the structure of a graphic can get lost, so if you need to extract Animatable Objects for kinematics do that before optimizing.
If you have more than 1 object using the same graphics, load it only once in a class object and instance this into the model. It will only load and use memory once even if you have many instances. Anyway, calcuölating the visualization of many objects uses graphic processor performance, so you want at least not a shared graphics chip and memory on the main processor. Also in a laptop you need a separate graphic card like NVidia mobile.
Good Simulation!
Ralph
01-10-2019 05:35 AM
Hi Ralph,
sorry, these 240Mb is the disk size of the model.
My actual workstation is 3,4 Gb processors (8Cores), 16Gb memory, AMD Radeon HD7470 (1Gb) and 2Tb disk.
I have tried the 'Optimize' options (1. Prune tine graphics; 2. Visibility filter; 3. Flatten structure) for the graphics, but all of these just leave a few pieces of my 3D model, ... using V13.1
Lately I started using V14.2. Here this 'Optimize' options do not destroy the model.
Here I see a follows:
Step2. Visibility Filter (using default shapes) Nodes (CPU): 32.828 & Polygons (GPU): 3.869.521
... with all settings to default
...trying:
Step2. Visibility Filter (using default faces) Nodes (CPU): 71 & Polygons (GPU): 731.721
Step2. Visibility Filter (using default triangle stripe) Nodes (CPU): 71 & Polygons (GPU): 563.351
Step2. Visibility Filter (using default triangles) Nodes (CPU): 71 & Polygons (GPU): 149.953
...trying:
Step2. Visibility Filter (using default shapes) and 'Super' reduction level:
Nodes (CPU): 71 & Polygons (GPU): 149.953
...trying:
Step1. Only change Prune tiny graphics: 0.001 > 0.01: Nodes (CPU): 69 & Polygons (GPU): 1.932.906
...trying:
Step3. Flatten structure, set Keep Grouping: Nodes (CPU): 2390 & Polygons (GPU): 1.949.578
>> Any these selection still show correct graphics.
So I will redo my model 3D using the >> Step2. Visibility Filter (using default triangles)
I will also have the 3D graphics setup already in the class object library.
Kind Regards
Marco
01-14-2019 08:37 AM
I have updated my model to apply the JT file to the class and I also use the 'Optimize' option step2 (triangles). The disk size of the model is down from 240Gb t0 110Gb.
Still it takes quite some time to open and close it.
As the simulation runs, the moments in the 3D graphics are not really fluid.
I was wondering if the performance of my station (& graphics card) is sufficient.
Regards
Marco
01-14-2019 09:02 AM
Have a look here:
User Conference 2018 - Presentation about 3D Model Performance
3 weeks ago
I have checked my graphics card AMD Radeoan 7470HD whis is quite old (2012).
Below you can see a rating, can you give a recommendation ?
2 weeks ago
I think, regarding the graphic adapter performance, any reasonably modern NVIDIA card should be fine. In my experience, an up-to-date graphic driver often accomplishes more than a fancy new graphic adapter.
And in Plant Simulation, much more important is the reduction of the rendered graphic node count.
2 weeks ago
Using the 3D optimize, I have reduced:
- the amount of nodes from 11715 to 24
- the Polygones from 2376938 to 183674.
I applied Step1: 0.001 > 0.1; Step2 Shapes to Triangles
This model show 3D graphics (JT) of real world machinery.
Finally my model, ... in a general overview run < 1 FPS
I guess to need more powerful graphics card.
2 weeks ago
There are models with far higher numbers of nodes (your value of 24 is fantastic!) and numbers of polygons and they can run on machines without dedicated graphic adapters (i.e. Intel on-board graphics) and they run a lot faster.
I think we are still missing an important piece of the puzzle. Just to be certain: You are talking about the numbers for the entire frame, right?
Here are a couple of questions regarding your overall situation:
Do you have point clouds and/or shadows in your model?
Which OpenGL level do you use? (basically, for starters, it is only interesting whether it is >= 3.1 or something lower)
Can you post a screenshot of your model with frame rate output? (this should contain some useful information regarding these matters)
2 weeks ago
I have totally 11 of these machines with 24 nodes.
The tools to load, .. which are standing standby next to the machines have 10 nodes 175965 Polygons.
I although have walls which are partially transparent.
I added checkboxes on my dialog to show/hide these walls.
If I hide all walls the frame update goes to 7..8 FPS.