Vram utilization of GTX vs RTX gpus.

It's often said that  "to emulate rtx functions, more vram is utilized on gtx cards".

The first question i have, What are "RTX functions"?

Is this specific render settings that need to be enabled, or just iray in general?

 

Then, what is the actual difference?

A couple MB, or several GB?

 

Hard data(actual numbers) only please, no suposition or opinion.

 

I'm still going to do my own testing and i'll post results later, but i'd just like to know what others are seeing.

 

 

Comments

  • nonesuch00nonesuch00 Posts: 17,929

    DrunkMonkeyProductions said:

    It's often said that  "to emulate rtx functions, more vram is utilized on gtx cards".

    The first question i have, What are "RTX functions"?

    Is this specific render settings that need to be enabled, or just iray in general?

     

    Then, what is the actual difference?

    A couple MB, or several GB?

     

    Hard data(actual numbers) only please, no suposition or opinion.

     

    I'm still going to do my own testing and i'll post results later, but i'd just like to know what others are seeing.

     

     

    That's impossible to answer because the size of the hypothetical scene you want us to comment on is also unknown. It too could be a couple of MB or a couple of GB and that's not even factoring it bells and whistles that increase memory usage and processing power like shader materials and all the hidden geometry of normal maps, opacity maps, and so on.

    Hard facts need hard data and you've supplied no hard data. 

    To go a real test case that has relevance use a GTX GPU and a RTX GPU for the same scene, swapping cards and measuring usage for your test scene.

  • LeanaLeana Posts: 11,026
    edited November 2021

    "RTX functions" refers to real-time ray tracing, among other things. And yes, that would be used by iray in general.

    Those functions use the specialized RT cores of RTX cards to do the computing more efficiently. These cores don't exist in GTX cards so their features need to be emulated for GTX cards.

    The amount of extra resources required will depend on the scene and how it would use the RTX features.

    Post edited by Leana on
  • prixatprixat Posts: 1,585
    edited November 2021

    That's an interesting question. The nearest thing to a direct comparison would be between a GTX1660ti and a RTX2060. The 1660 is the same generation but with it's RTX cores disabled!

    Iray used to allow us to turn "Optix" off. (Optix is nVidia's graphics library to accelerate common graphic functions including Raytracing)
    That was stopped a few versions back, now Iray examines your hardware and decides for itself which graphics library to use for each rendering device.

    Since we no longer have a choice it's a bit academic. smiley

    see attached, for example. it decided Optix was best for the RTX3060 and Embree was best for the CPU. (Embree is Intel's Raytrace accelerator)

    Screenshot 2021-11-30 105614.png
    1172 x 344 - 50K
    Post edited by prixat on
  • jmtbankjmtbank Posts: 164

    DrunkMonkeyProductions said:

    Hard data(actual numbers) only please, no suposition or opinion.

     

    People can't give you hard exact numbers for the same reason you can't get a figure for the ram your scene is about to need out of the program before you hit the render button. 

  • kyoto kidkyoto kid Posts: 40,575

    ...however for GTX cards Optix also consumes VRAM as there are no RT cores.

Sign In or Register to comment.