I was doing a test beween TurboCAD 2016 Plat. x64 and Shark LT v9 x64. in loading an Chain Diff. STEP file (293 MB.) that I down loaded from GrabCAD. Believe the file was made with Solidworks. Anyway, with TurboCAD, it took 1 min. and 16 sec. to load and 16 sec. to render in Adv. Render (RedSDK). Loading and rendering are two separate functions. With Shark LT it took 3 min. and 24 sec. to load and render. Another file "mini Galley STEP file (1.57 MB.) load and rendered time TurboCAD and Shark LT were the same. With v10 coming, will there be any improvements with loading and rendering. Also in the forum, I have read that there is a problem with memory not clearing correctly. I could not get the Chain Diff file to load with TurboCAD x32, which could be from the memory limitations of x32. Both TurboCAD and Shark perform well in moving around and rotating 3D renders. TurboCAD has alot of tools, but isn't a NURB CAD program, but does have MESH. Hope the developers work on developing a strong x64 core, then work on the tools improvements. Seen to many software programs rush to upgrade to x64, then have problems down to road. TurboCAd and Shark seemed to handle sharing files (STEP's, .stp, exc)

>>With v10 coming, will there be any improvements with loading and rendering.

Yes. I spent quite a bit of time benchmarking loading and rendering/facetting in V10. It was pretty interesting.

I found that this was dependent on:

1. disk access speeds
2. processing NURB data from the file
3. converting NURB data into facets for display/rendering.

Regarding #1 Disk Access, I was amazed how much a SSD disk would make. You can immediately speed up any file read/write/import/export just by placing the file on a SSD disk.

For #2, we added support for multi-threaded reads. If you have 4 cores, I benched 2x performance for large files. My test file was 1 Gigabyte. There is overhead for multithreading where the larger the file the better the performance.

For #3, we added support for multi-threading facetting. Again you can get 2x performance with four cores.

With V10, we also updated the way we pass facets to OpenGL. We now dump as much data as we can to the GPU instead of the CPU.

>>I could not get the Chain Diff file to load with TurboCAD x32,
I have become very biased on 32 vs 64. I would not recommend 32 for large CAD files. If you have a large file, plus add app resources, windows(os) resources, cached data/display lists, it's easy to run out of 32 bit space.

I would recommend download v10 beta/trial and benchmarking your typical files to check performance. The results are very dependent on the size/type of data you are working with.

Section on Multithreaded Translators

Section on Faster Displays

Repaint Compare


Tim

Edited by user Thursday, December 8, 2016 11:15:10 AM(UTC)  | Reason: Not specified