Question on HQ Viewport

[xsi_fanatic]

Hi Folks,

What video card is recommended for fluidity with the High Quality viewport ?

My monitor is connected to a Geforce 690, you'd think that might be good enough, but apparently it's not. So my guess is to go for a Quadro, but I haven't tapped that area yet and don't know what specs to look for. I also render with Redshift, so I might as well hit two birds with one stone and get something to serve both purposes.

Any insights or recommendations would be appreciated.


Cheers,
XF

[homam]

The HQ speed is acceptable if you are not forcing it to recompile the shaders. Of course having faster gpu helps, but you'll still notice a hang every time it recompiles shaders.

Regarding the gpu, it doesn't matter a lot between quadro or GeForce. What really matters is the gpu clock speed.

Hope this helps,

[xsi_fanatic]

The HQ speed is acceptable if you are not forcing it to recompile the shaders. Of course having faster gpu helps, but you'll still notice a hang every time it recompiles shaders.

Regarding the gpu, it doesn't matter a lot between quadro or GeForce. What really matters is the gpu clock speed.

Hope this helps,

Thanks for the feedback Homam.

Can you elaborate more as to what you mean by "recompile the shaders" ?

Also, tell me more about the GPU clock. What are the specs that I should be looking for ?

[homam]

From the documentation:

High Quality View works by compiling MetaSL code into CgFX instructions on-the-fly at runtime.

this is basically converting the render tree graph to more gpu friendly language. Soft triggers this compilation process everytime the render tree is changed.

clock speed is a term that referances the processor, memory or bus speed. Usually measured in Mhz. In case of gpu, higher this number, higher frame rate in your viewport.

For more reading:
https://en.wikipedia.org/wiki/Clock_rate

[FXDude]

Recompilation of all is invoked every time entering the view.

Otherwise since it's first incarnation, it recompiles only what is changed when making changes.

Good practice for larger scenes can be to isolate elements of interest prior to entering the view.

[mirkoj]

honestly hq viewport is poorly done not much of use feature that looks more to be something that they made to fill in what;s new list while waiting to pull the plug completely on Softimage.
avoid...

[FXDude]

Sure it can be gimmicky,

But it can still be good for certain things, I personally use the HQv to isolate objects i'm working on while lookdeving, to vew layered texture tiling and alpha blending/levels (also accurately represents all blending modes and color adjustements), or make a 'no material' pass (with no recompilation) to see where shadows fall with lights, or also to make snazzy looking playblasts.

I also normally have a 'viewing pass' where I apply a display property to all objects in that pass to display all as 'Shaded' or 'Object View', then set display settings to 'Mixed Viewing Mode', and then turn ON HQV just for antialiasing with 'regular' viewing modes with AA for working like that. (which also does not involve any recompliation when switching, and which can also be excellent for playblasts.)

[homam]

It is true that Autodesk like to fill the "What's New" list with immature features. However compared to the first iteration of the "Viewport 2.0" in Maya, HQV was very promising and could be the default viewports rendering mode once the perofrmance issues were resolved. Unfortunately it is not there

[FXDude]

Indeed the Maya VP2 has its own share of issues, enough to be no exaggeration to say that people typically revert to regular view and reseve VP2 for playblasting. (still a couple of years out)

So maybe Vp3

[Mathaeus]

I'm playing with Maya version having Viewport 2 as only option, that's LT, have zero problems with that. When we already are about viewports, while Maya is a bit less snappy than XSI 7.01, it's way more stable, I have one or two crashes weekly, same number was daily with XSI (operation related to mesh edges traditionally holds 1/4 of crashes )- anyway even that was acceptable during years.

[mirkoj]

point is that they already had killing SI in plan, it was time wasted on feature that wasn't possible to be implemented fully and properly to be usefull.
it is slow and tied to mental ray as I'm aware of it. big majority moved from mental ray a while ago so with no support for other materials...
anyway that time would be better used on bug fixes and leaving as stable as possible SI instead implementing another half done feature.

[FXDude]

I agree it's too bad it considers factory nodes only, but mentioned uses can still apply to Redshift (or 3delight) that translate factory nodes.

I also agree that dev time could have been better spent on other things, but that's easily part of the least of all things Softimage / AD related.


For VP2 no doubt is fine for 'regular usage', but 3rd party shaders seems to be one of the places it can also have trouble, although there's a way to bypass shaders and have a set of nodes just for viewing representations.

But that's among things like selection, uv issues and other things as scenes shaders get more intricate, but no doubt it's implementation is probably more complete, and more issues are to be expected proportionally how 'advanced' certain features may be (like motion blur)

But whichever RT engine, it still remains largely gimmicky or non critically essential eye candy.