Performance behavior with CPU / GPU / CPU + GPU as expected?

uezi

Hi,

I've just migrated from a Mac to a Windows PC and for the first time, I've got a nVidia GPU. Now I got some questions on the observed performance...

Specs: Intel Xeon with 4 cores, 8 threads, 32 GB memory, Quadro M4000 (8 GB)

Scene: 1080p with 5 G3F/G3M figures, clothes, hairs, Props, Tavern, Background, lights, iray
ray Memory Assistant reports: 18 GB SysMemory, 4 GB VRAM

 

Rendering method	Duration
CPU only	6h 01min 01s
GPU only	1h 48min 41s
GPU + CPU	1h 26min 23s
GPU + CPU + Optix enabled	1h 21min 35s

Is this somehow expected ? 
Adding a GPU "only" gives a performance boost of ~4.5x ?
 

Thanks & best regards,
​    Peter

Toonces

Hmm, that card only has 1664 cudas. Also, I tend to run a config you don't have listed, i.e., GPU + Optix (no CPU).

To see if you're getting decent performance, you should run the scene in this thread:

https://direct.daz3d.com/forums/discussion/53771/iray-starter-scene-post-your-benchmarks/p1

Then simply peruse a few of the posts to see how yours compares with others.

Tobor

uezi said:

Is this somehow expected ? 
Adding a GPU "only" gives a performance boost of ~4.5x ?

These numbers are in line, though it really depends on the specific hardware.  The M4000 uses an earlier generation of CUDA cores, so you'd probably get a little better performance if you used one of the Pascal-based cards. Xeons run the gamut between pretty fast to dog slow; my current dual Xeon system is only slightly faster than my older single Xeon system, though it has four more cores.

You will note that there's only a 15-20 minute advantage between GPU and GPU/CPU, so you may elect to just use the GPU, freeing your computer during your render. I burned out my previous Xeon workstation because of very long renders. Even Xeons aren't made to be full throttle at 100%.