Face Count, VRAM and Model viewport display.

[wojtek]

Hey,

Is what is possible to display in the viewport at one time dictated by the VRAM or some other internal limit as well? I'm on a 3090 and considering getting an A6000 for the extra memory.

I can currently display about ~300M face meshes with the 24Gb of VRAM 3090 has. Bigger meshes (400, 500M) display in fragmented way (ie only 10 is visible for example).


Alexey, can you advise?

[Alexey Pasumansky]

Hello wojtek,

For high-poly models display you can enable VBO option in the Advanced Preferences tab and assign considerable amount of VRAM to it. On RTX 3090 you can try to set 6 GB in VRAM field and check, if it allows to navigate ultra-high poly models smoothly. Also please check, that NVIDIA driver is up-to-date.

[wojtek]

Hello wojtek,

For high-poly models display you can enable VBO option in the Advanced Preferences tab and assign considerable amount of VRAM to it. On RTX 3090 you can try to set 6 GB in VRAM field and check, if it allows to navigate ultra-high poly models smoothly. Also please check, that NVIDIA driver is up-to-date.

Hey Alexey,

Thanks, I did have VBO setup already. This is what happens when I try to load a 550M mesh (see attachment).

Doesn't seem to be a VRAM issue as task manager does not show the utilization anywhere near 100%.

edit: attached a second image showing the other result that I'm getting (430M mesh).

[Bzuco]

@wojtek
I tried import 24M model from FBX file(binary ver. 2014) and I am also getting not correct model ..missing polygons. When I imported same model but in OBJ format, it displayed correctly. I tried VBO on/off...difference was only in VRAM usage/allocation ( VBO-on 465MB, VBO-off 104MB ).
 I think A6000 is not necessary . With 24GB vram you should be able teoretically load model with ~ 5500M polygons(just vertexes and triangles, no normals, no smooth groups, no UVs, VBO-off).
You can check in gpuz util how much vram increases on your model(with all attributes UVs, smoothgroups, ...vertex colors, .. etc.)

[PolarNick]

For me it looks like a GPU driver problem. Maybe Metashape will be able to workaround it via subdivision of the mesh rendering into multiple draw calls (while the whole geometry data is already in VRAM)?

[wojtek]

For me it looks like a GPU driver problem. Maybe Metashape will be able to workaround it via subdivision of the mesh rendering into multiple draw calls (while the whole geometry data is already in VRAM)?

The cut-off seems to be 350M faces in my case. Has anyone been able to display more complex meshes?

[PolarNick]

+1 for cut-off around 350M

[wojtek]

I tested this with "Show Aligned Chunks" and I could easily display 2 chunks with 300M meshes each for 600M total.

VRAM was loaded up quite heavily but still not 100%.

Looks like the 350M limit is a Metashape problem.

[wojtek]

Hello wojtek,

For high-poly models display you can enable VBO option in the Advanced Preferences tab and assign considerable amount of VRAM to it. On RTX 3090 you can try to set 6 GB in VRAM field and check, if it allows to navigate ultra-high poly models smoothly. Also please check, that NVIDIA driver is up-to-date.

Hey Alexey,

Have you guys been able to reproduce what we've reported in this thread?

[Alexey Pasumansky]

Hello wojtek,

Yes, we were able to reproduce similar problem on ultra-high polygon count mesh models.

[Alexey Pasumansky]

Hello wojtek,

If you have time, please check pre-release of 1.8.0 build 13257 for high poly models display (with Enable VBO support option ticked on in the Advanced Preferences dialog).

[wojtek]

Hello wojtek,

If you have time, please check pre-release of 1.8.0 build 13257 for high poly models display (with Enable VBO support option ticked on in the Advanced Preferences dialog).

Hey, I just checked and looked for this thread to provide feedback. It does look like you have fixed the issue I was getting. Thanks!

EDIT: Also it seems that moving the viewport when the model has selected faces does not slow things down by as much as before. Great stuff.

[CheeseAndJamSandwich]

Hello wojtek,

For high-poly models display you can enable VBO option in the Advanced Preferences tab and assign considerable amount of VRAM to it. On RTX 3090 you can try to set 6 GB in VRAM field and check, if it allows to navigate ultra-high poly models smoothly. Also please check, that NVIDIA driver is up-to-date.

Alexey, can you please give us some guidance with respect to VBO?:

• When we should consider enabling it?
• What value in relation to how much VRAM we have?
• Or how big our models are?
• And what implications does this have to other parts of Metashape, alignment, meshing processing, etc?
• Should it be disabled during processing, and only enabled for the final model?
• And it obviously works wonders, but what is VBO, what does it actually do in MS???

Thanks.

[wojtek]

Hello wojtek,

For high-poly models display you can enable VBO option in the Advanced Preferences tab and assign considerable amount of VRAM to it. On RTX 3090 you can try to set 6 GB in VRAM field and check, if it allows to navigate ultra-high poly models smoothly. Also please check, that NVIDIA driver is up-to-date.

Alexey, can you please give us some guidance with respect to VBO?:

• When we should consider enabling it?
• What value in relation to how much VRAM we have?
• Or how big our models are?
• And what implications does this have to other parts of Metashape, alignment, meshing processing, etc?
• Should it be disabled during processing, and only enabled for the final model?
• And it obviously works wonders, but what is VBO, what does it actually do in MS???

Thanks.

VBO is explained here: https://en.wikipedia.org/wiki/Vertex_buffer_object

Keep it enabled, it should have no effect on anything but model display. in 1.8.0 the only setting is the on/off switch.

[CheeseAndJamSandwich]

VBO is explained here: https://en.wikipedia.org/wiki/Vertex_buffer_object

I found and read this (the first part, lol) when Alexey first told me to enable it... As i'd never heard the term...  It's a fascinating read!
I was hoping we might get a more layman's terms, Metashape specific explanation.

Keep it enabled, it should have no effect on anything but model display. in 1.8.0 the only setting is the on/off switch.

Well, this will make it much easier! 

But my engineer brain would still love to know some technical* details about it with respect to Metashape. 

* Something more palatable than the wikipedia page!