AWS GPU Rendering

An update to my AWS tests. This time Using GPU - G2.2xlarge instance.

Stop the clock! Setting up the nvidia driver to use the GPU was not pleasant and I nearly gave up. I'm not going into it here but it was very messy

My original cpu only test is here

This test was done using the same scene file in blender 2.76a - a 2000x2000 resolution basic "Aiko with hair and clothes" render..


AWS G2.2xlarge Nvidia Grid (4Gb) 1,536 cuda, Xeon E5-2670 CPU, 15Gb Ram - 69 seconds

MyPc GTX760 (4gb) Core2Quad, 8GB Ram - 2 minutes

So the G2.2xlarge instance was quite a bit slower at rendering the scene than the CPU C4.8xlarge instance (25 SECONDS TO RENDER) used previously and faster, as expected, than my home pc.

As I now have windows server set up with the Nvidia Driver, my upcoming G2.8xlarge test should be better (I can now also make an AMI copy of the instance and use Spot Price on new Instances)

Iray To 3Delight Conversion Script

Another useful script for Daz Studio was posted by Esemwy on the Daz Forums that converts Iray materials to 3Delight

-you can download it here

AWS 3Delight Studio Pro Rendering

3Delight Studio Pro (8 Core version)

This is the last test until GPU is enabled on my AWS setup. I'm using the free 3Delight Studio Pro installed on an Amazon Web Services C4.8xlarge instance running windows 2012 server, which is a tad overkill as this free license version of 3Delight is limited to using 8 cores.

The scene I'm using is Dreamlight's Cloud City which is quite demanding on my little Core2quad with 8GB Ram so I kept the image quite small. The Render below is the output (right click to view full image) Test Render 679 x 702 resolution

Computer Specs/Results...

Core2Quad @ 2.66Ghz 8Gb Ram - 1 hour 15 Minutes

AWS G4.8xlarge Intel Xeon E5-2666 v3 (8 cores utilized) 60Gb Ram - 15 Minutes

(you could also add 5 minutes for the ftp upload of the .rib file)

I will update this post with the results of a lesser, more suitable instance, such as C4.2xlarge (8 Cores only 15Gb Ram) at a later date

One thing to note is that, at least for me, the 3Delight render window will not load on a windows instance ( I think it requires opengl and instances without gpu use windows basic display-not sure on that one) It throws an error, but if you just leave it alone the render does complete and out pops a .TIFF file, you can then close the error pop-up.

I guess the 3Delight unlimited multi-core version would be very snappy with the full 36 cores of a C4.8xlarge instance

Daz Studio 4.9.3 Pro Released

Daz Studio 4.9.3 General release is available, still free and now able to network render to iray server

Roll-up Roll-up! - Get your Daz Studio free copy - erm  .......for free!

  DAZ Studio 4 - Create 3D Art for Free



Cosmos Laundromat Benchmark Cloud Rendering

Here's my next test using Amazon Web Services windows machine for Blender rendering.

This test using the Cosmos Laundromat Benchmark file is rather intensive with the grass,hair, fur and lighting. I actually did try to render on my home Core2Quad pc to compare but a hand popped out of the viewport, slapped me a few times and then gave me the silent treatment until I rebooted.

I ran the render on the same C4.8xlarge large instance as in my previous post which is CPU only. I should mention that I am using Blender 2.76a on the instance as later versions wont start (opengl missing error) note This was a windows server instance - I think linux would be faster. Also, I enabled "use Open Shading Language" in Blender

Here is the completed render (open in a new tab to view full sized)

AWS Windows Server 2012 C4.8xlarge 1 instance

1 Frame 2048 x 858 - Render Time 35.15 minutes

AWS C4.8xlarge windows instance on demand currently costs $3.091 per hour (expensive!) You can change the instance type at any time to a different cost/powered instance though -I installed Blender etc using a m3.xlarge instance and then changed it to C4.8xlarge for rendering. I will definitely need to be setting up a linux instance soon or incur the wrath of missus digidotz!

You can get a feel for aws and learn how to get things working by using a low powered EC2 t2 instance which is free for the first 12 months. Once comfortable using the service you might then move to a paid instance for rendering. You just need a fast internet connection for smooth remote desktop experience.

Thankfully there are cheaper AWS options available such as Renderbot which is very interesting.

Renderbot lets you build your own full scale-able renderfarm and can save you money by using amazon spot price on instances which are quite a bit cheaper than "on demand" eg C4.8xlarge on linux is currently $0.2625 per Hour spot price

My next tests will be using GPU enabled instances and then I will get to Renderbot.

Cosmos Laundromat material used - (CC) Blender Foundation |