Slightly different "User Preferences" "System" options in v2.78?

Here’s a screenshot of the “System” page of the User Preferences I’ve got:


The “Cycles Compute Device” is a little different from the tutorial as there’s an “OpenSubdiv” option below it. Also, the GPU option is listed as “GeForce GT 740M (Display)”. I’m guessing that this is because I’m on a laptop with a built-in GPU, but thought I’d flag it in case it’s significant.

Here’s a screenshot of my first render test (and quickest!):


The CPU/GPU selection on the Render option panel is greyed out, although I can still flip between the two.

The render times I achieved with differing settings are:

CPU 128x128: 5:08.74 (5:06.17 when repeated next day)
CPU 64x64: 5:10.49
GPU 128x128: 5:12.36
GPU 256x256: 5:51.20
CPU 256x256: 6:02.45

I also tried swapping over to “Blender Render” just to see what would happen, and got a solid-grey scene. But at least I got it 4.2 seconds, which is nice :grin:

1 Like

I have the same extra option as you, except im on a desktop and i believe my tower has its GPU built in as well. It’s a slim Asus GOR G20AJ Rogue. I think that most slims have the built in GPU as well as laptops for space. Im not a hardware expert so could be wrong. I left my Opensubdiv compute to “none” and it works fine.

@Poulet_Noir Did you play around with the Opensubdiv option? if so does it effect anything?

1 Like

Hi @CookieEater, no I didn’t touch it. I’m a “Don’t poke that unknown thing” kinda guy!

1 Like

Cool, thanks for the link Marc!

1 Like

@Poulet_Noir im a kind of mess it up and hurt my brain to try to figure it out guy haha

3 Likes

Here are my Times. I am thinking about getting the new AMD Ryzen 1700x once it gets a little cheaper since the I7 is beyond my budget, and maybe the gtx 1060 6GB one or the other.


Blender 2.7 Cycles Benchmark rev4
“Now with twice the orange!”

Set Up
CPU: AMD FX 6300 6 core 3.5 ghz
GPU: Nvidia GTX 750
OS: Windows 10 64 bit
Res: X 1920 px Y 1080 px

GPU CUDA
Time : 3mins 26 sec
Tile size
Center X 240, Y 180 (auto)

Tile Size
Center X 950, Y 540
Time 5 Mins 29 sec

CPU
Tile Size
Center X 950, Y 540
TIme 50 mins Estimated

CPU
Tile Size
Center X 32, Y 32 (Auto)
Times 7 mins Estimated

Hello all! Here is what my computer could do:

My best time: CPU: 3 minutes 23.63 seconds
Best GPU time: 9 minutes 54.63 seconds

My computer set-up:
Open CL
Intel HD Graphics 630 1536 MB
Dedicated GPU: Radeon Pro 560 4GB
OS: macOS High Sierra, MacBook Pro

So here are my results and findings:

I have played with the OpenSubdiv compute modes available under “System” in Blender 2.79 and found interesting differences.

CPU: Intel i7 3.6Ghz
GPU: Nvidia GTX 1060 3GB
OS: Windows 10 64bit

Results:
GPU - OpenMP

GPU: 1060 (128x128)
Time: 1 min 45 Seconds

GPU: 1060 (256x256)
Time: 1 min 28 seconds

GPU: 1060 (512x512)
Time: 1 min 27 seconds

GPU: 1060 (960x540)
Time: 1 min 44 seconds

GPU - GLSL Transform Feedback

GPU: 1060 (960x540)
Time: 1 min 36 seconds

GPU: 1060 (512x512)
Time: 1 min 37 seconds

GPU - GLSL Compute

GPU: 1060 (512x512)
Time: 1 min 32 seconds

GPU - CPU

GPU: 1060 (512x512)
Time: 1 min 27 seconds

GPU - None

GPU: 1060 (512x512)
Time: 1 min 25 seconds

[ CPU ]

CPU: i7 (16x16)
Time: 3 min 20 seconds

Here are my times only did 4 because my computer is very slow, my best time was 9min ;(

My GPU does is about twice as fast as my CPU and I got a GTX 760. I did run into CUDA error called illegal addres, but I restarted Blender and changed the start resolution in the performance tab to match the tiles and it worked okay


With the defaults settings for rendering in the lesson:
GPU Compute on a GeForce GTX 675MX
Tiles - X: 128 and Y: 128
The render time was 13:05.57

Now switching back to CPU I got a better rendering time of 05:40.31

Overall the image came out the same but the rendering times were different. So I guessing that some more testing will have to be done. But it is looking like the CPU might be my best bet when it comes to rendering.

Privacy & Terms