GTX970 - TITAN X speed test rendering for IRAY

My config i7 2600k + 32 Go + PCI express 2.0 motherboard. MSI GTX970 (4 Go, 1664 Cuda) main GPU, MSI TITAN X (12Go, 3072 Cuda) secondary GPU

Test 1: scene : Victoria 6 + Skyler hair : Iray material for both. Environment map + headlamp. The scene weight is 2 Go. Render Time 1 minute. OptiX prim ON

CPU : 85 iteration.

GTX 970 : 700 iterations

Titan X : 1067 Iterations

GTX970 + Titan X : 1460 iterations

CPU + GTX970 + Titan X : 1136 iterations.

Test 2 : Scene test : Pursuit on my gallerie. The scene weight is 5 Go, impossible to render with the GTX970. Firefly + Bloom + Noise filter, quality 2. Render time 3 minutes. OptiX prim ON

Titan X : 262 iterations

GTX970 + Titan X : 348 iterations but the memory usage is absurd : 4Go on the GTX, 8 GO on the Titan X.

Conclusion.

The CPU is useless. The best option is to use 2 GPU. the number of CUDA is important but the gain is not linear

 My intention was to use the GTX for the viewport and the Titan X for rendering.

 

Comments

  • IceCrMnIceCrMn Posts: 2,142

    Did you happen to save those renders so we can see them for comparison?

  • nicsttnicstt Posts: 11,715

    I almost always get better results with just my render card selected for rendering; if selecting anything else ever has made a difference, then it was very small positive but potentially more detrimental. It also increases heat and noise another negative for me. I like quiet. :)

    OptiX seems pointless at best.

    Anyone actually know what it is supposed to do?

  • JPiatJPiat Posts: 70

    None are finished... The idea is to compare the number of iterations after the same period of time using different configuration.

    The default setting 5 000 iterations is not that bad. I use 18 000 iterations on most of my render. 

Sign In or Register to comment.