Performance scaling for Nvidia GPUs by GPU Generation

Hardware

  • CPU: AMD Ryzen 9 5950X 16-Core 32-Threads.
  • Motherboard: Asus X570 Crosshair 7 DarkHero.
  • System Memory: 128 GB DDR4 3200 CL18.
  • GPUs: RTX 4080, RTX 3080, RTX 2080, GTX 1080.
  • Bandwidth: PCIe Gen 4.0, x8.
  • OS: Windows 11 Pro.
  • GPU driver: NVIDIA Studio Driver version 536.40.
  • Displays: No monitors attached to the GPUs.

 

How we ran the tests

We ran four scenes in Progressive mode, double-checked the results, and tuned each scene to reach 100% GPU utilization. Each test ran for a minimum of one full minute to provide comparable, repeatable results.

 

​Results

  • Pascal to Turing shows average 1.6X speedup in CUDA

  • Turing to Ampere shows average 2.4x speedup in CUDA​ and 2.53x in RTX

  • Ampere to Ada shows average 1.87x speedup in CUDA​ and 1.86x in RTX

  • We tested the same scenes between the RTX 3090 and RTX 4090, there is 2.15x speedup in CUDA and 2.2x speedup in RTX

 

Test 1:

IntCam002_Post.jpg
Test 1 scene preview
vgBAN6Yxkh.png
Test 1 performance results

 

Test 2:

Cabinet_01.jpg
Test 2 scene preview
Screenshot_39.png
Test 2 performance results

 

Test 3:

001.jpg
Test 3 scene preview
Screenshot_40.png
Test 3 performance results

 

Test 4:

Note: this scene includes thousands of leaves and grass objects, and has all shading baked to textures, hence showing a big boost for RTX mode. In most cases, CUDA mode should be close to RTX mode in performance.

004_GPU.jpg
Test 4 scene preview
Screenshot_42.png
Test 4 performance results

 

Performance scaling summary

You can expect V-Ray GPU to scale very well on recent NVIDIA GPU generations, often delivering nearly double the performance from one generation to the next.

 

Was this article helpful?
3 out of 3 found this helpful