Announcement

Collapse
No announcement yet.

Which components of a RTX gpu have more influence on Chaos Vantage performance?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Which components of a RTX gpu have more influence on Chaos Vantage performance?

    Hey, guys! I hope you all are well.

    I'm planning to get a RTX, but first I would like to understand a little better how the GPU components would influence on performance, so I can make a smarter decision when it comes to the cost x benefit ratio.

    Cuda cores, memory, clock... Which of those specs have more impact on Chaos Vantage performance, and how should I read them when comparing the specs to make a decision?

    All the best,
    Gabriel Chahoud


  • #2
    Probably the most important factor is how big your scenes are. If your scene can't fit in video memory it will either fail to load (crash) or will render a lot slower. It's hard to guess how much memory you will need - you may have to find a way to test in advance (borrow a card, or use a VM). Rendering final images at very high resolution also uses up some of that memory. But don't be too intimidated by this, I think in most cases scenes fit in single digit gigabytes (<10gb).

    From there, higher clocks and more SMs and RT cores make it run faster of course. You could probably get an approximate idea by checking the V-Ray benchmark results: https://benchmark.chaos.com/v5/vray-...res-only=false
    It's not exactly the same render engine as Vantage but it should give you an idea of how things scale between the different models.
    Last edited by npg; 27-03-2023, 03:06 AM.
    Nikola Goranov
    Chaos Developer

    Comment


    • #3
      Thank you very much for your attention, Nikola! I have some additional questions, if you can help me:

      As we don't have a Chaos Vantage benchmark tool, recently I saw a video comparing two RTXs running the V-Ray Benchmark - the Rtx 2060 vs Rtx 3060 -, and surprisingly, even though they have similar clocks and the 3060 have about 1500 Cuda Cores more than the 2060, the 2060 had a better performance on the V-Ray benchmark (screenshot on the end of the post).

      Because of that too, I'd like to double check some things before buying one of them:
      • Do you think that benchmarks below would be similar with Chaos Vantage?
      • Would you say that the most important things to "predict" a gpu's performance on Vantage would be: memory, clock and cuda cores? Or besides those is there anything specific I should pay attention to when choosing the gpu model?
      • How much difference of performance I should expect between the RTXs 2060 and 3060 for Chaos Vantage?
      Click image for larger version

Name:	2060 vs 3060.png
Views:	297
Size:	747.7 KB
ID:	1176722



















      Comment


      • #4
        It seems there is something wrong with his test run but I am not sure what. As you can see on the screenshot from our benchmark page, the average result for 3060 is around twice that of 2060. We use desktop GPUs of every RTX generation: 20-series, 30-series, 40-series and each generation is like roughly twice faster than the previous one in Vantage. If you open the details for these V-Ray benchmark results on the webpage you will see that there are some low scores on the last pages. Especially with laptops there are many things that could affect it - other processes, power management, overheating, etc.

        To your three questions:
        - No, I don't think this youtube video is representative of 3060 performance on Vantage, there is something wrong on his laptop.
        - There is no formula. Statistics like core counts (also memory speed etc.) can be useful when comparing the same generation of cards like 3060 and 3070 but there are more complex differences between generations. Real-world testing is better. The V-Ray GPU RTX results from our benchmark website should give you some idea. The results are averaged there, you can look at details for specific test runs as well.
        - Roughly double

        Click image for larger version  Name:	benchmark.png Views:	0 Size:	28.8 KB ID:	1176777
        Nikola Goranov
        Chaos Developer

        Comment


        • #5
          Awesome, Nikola! Now it's a lot clearer for me.

          Thank you very much for your thorough answer, it was very helpful.

          Comment


          • #6
            I think it is a prime example of why not to easily believe in videos posted online!

            Comment

            Working...
            X