iray benchmark


Hi all.
I want to compare some cuda gpu cards, so I prepared very simple scene.
I hope you are interested in this.
Rule is very simple.

  1. Download the scene here or here and render it. Post the rendered image only.(10min render)

  2. Change the render option to ‘Iterations’. I set the passes to 100. Just render it.
    Post the render time only.

  3. Post your CPU + GPU specs.

Here is some examples;

1.10min render

2. 100passes rendertime : 6m 4s
3. spec - CPU : xeon x3210 2.13Ghz quadcore, GPU : radeon

  1. 10min render
  2. 100passes rendertime : 33sec
  3. spec - CPU : i7 860 (3.4oc), GPU : GTX460

Thank you :slight_smile:


sorry for confusing, I edited the sample post so you can easily understand :slight_smile:


your link isn’t working


Sorry, I changed the link.


Workstation specs:
EVGA x58 FTW3 motherboard
Core i7 960 CPU
12gb DDR RAM
Video card #1: Tesla 2050
Video card #2: GTX 470
Win7 64

10 minute render:

NOTE: The 100 iterations render pass for this scene was reached in 19 seconds with this hardware.

For what it’s worth, here’s an additional benchmark scene:


Of course, the attached images are all 10min renders.

You don’t need to upload 100passes image because there are all the same result. :slight_smile:

My purpose is that;

  1. the rendered image quality comparison for the same rendertime with different hardwares.

  2. the render time comparison for the same passes(image quality)

The attached image is for 1, and rendertime is for 2. :slight_smile:


Sorry for confusing, I modified the sample post.


Ah, well the way you have it listed is extremely confusing then (at least for me). It looks like the text below the your renders would be the render time for those images. But now I understand the post.

At any rate I’ve adjusted my original post accordingly.


Ten minute result on a Core i7 920 with no CUDA:

Rendering 100 passes on the same machine took 2:11.

10 minute render on a Core i5 650 cpu with a GTX 460 with 768 MB of RAM:

Rendering 100 passes on this machine took :49.


28 sec for 100 passes.

i7920@4Ghz Gtx470


10min result on my Toshiba notebook Core i5 M430 (2.27Ghz) with GeForce GT 330M with 1GB RAM.

100 render passes: 2min 28s

Seems my GPU won’t offer me much power :blush: But anyway, iray does seems impressive and promising to me. So much less parameters to change and mess with to get result :thumbsup:


100 Interations in 37 seconds

10 minutes result on my Quadcore 6850 Extrem @ 3.0 GHz with 2 x GTX285

Seems that Iray is ignoring the two GTX285 or what? :cry:

At least Vray RT GPU likes them! :slight_smile:


something wrong with you or Intars5d

you get a better render time for 100 iterations but he gets less noise after 10 min.


Thats what I thought, but what do I know? :wavey:
Is there a need to install special CUDA Drivers for Iray?


I am such a numb nut!!! Uploaded the wrong picture! :banghead:
Now its the correct one. I am so sorry …


i7 920
GTX 285 1GB
12GB Ram

58 Seconds for 100 iterations

10 min render:


oh you had posted your 100 passes render I guess


Yeeeees, I did… :blush:

  1. 10 minute render

  1. 100 passes = 44 seconds

  2. CPU: E6850 @ 3GHz GPU: 2 x Quadro FX 5800

Render on i7 with 4 x GTX480 to follow shortly.

  1. 10 minute render

  1. 100 passes = 11 seconds

  2. CPU: i7 930 @ 2.8GHz GPU: 4 x GTX 480



Hp xw6600, Bi-Xeon E5420 2.5Hgz, 16 Gb Ram, GPU 8800GTS

100 passes : 1 min 22 sec

10 min:[u]

should be interesting to compare it with mental ray, vray, vray rt and other solution (octane… smallLux) ?
Let see beside computer which is the best renderer for that particular image.


I think it’s not fair to compare other renderers because they have different material structures.
But nevertheless, if you wanna test, you can do it with this obj file.

And thank you for the posts, friends. Let’s keep going. :slight_smile:


you are right but today there is so many solutions that anyone can choose, it’s quite disturbing.

It is an interesting opportunity to make an almost clear gallery of the different renderer :

  • rendering time for simple but close to production case scene (like your bedroom), but will need different scene settings according to the rendering engine.
  • best settings
    • for most ‘realistic’ rendering,
    • or fastest render
      this should be multiplyed by a factor based on the simple case time (like iray 100 samples for gpu accelerated and an MR version for CPU only)
  • for a single pciture and a smal animation (camera + character)
  • maybe a chart with the feature, and pros and cons with polls and rating
  • we could even compare software with simple scene

I suppose this to be a difficult work but it can be very usefull for all of us.

Let s consider that we are currently trying to define the configuration of a few new computer, and it was so hard to finally find proof that quadro is far even powerfull than standard Geforce.

I think that the cg talk forum is the best place i know to handle and gather that kind of clear and real and solid information.

I may be dreaming, but Who is with me ?