How fast is your CPU at running the AMD Blender benchmark?

[cbn]

To run the benchmark...

1.) first install Blender:

https://www.blender.org

2.) Then download the AMD file:

http://download.amd.com/demo/RyzenGraphic_27.blend

3.) Launch blender.

4.) Left click "File".

5.) Left click "open" on the drop down menu.

6.) Select the AMD blender file "RyzenGraphic_27.blend"

7. Press F12 to start the benchmark.

 

[csbin]

AMD ThreadRipper 13.04 seconds（render=100）

http://www.tomshardware.com/news/amd-threadripper-vega-pcie-lanes,34581.html

 

[cbn]

My result with E5 1620:

1.08.81 (1 minute, 8.81 seconds)

 

[cbn]

My result with stock speed W3670:

1.09.27 (1 minute, 9.27 seconds)

So within half a second of my E5 1620. That is interesting (Higher single thread 4C/8T vs. lower single thread 6C/12T).

 

[RichUK]

Didn't we do this about 5 or so months ago?

 

[cbn]

Didn't we do this about 5 or so months ago?

Yes, I did find the following thread:

https://forums.anandtech.com/thread...-benchmark-confusion-run-150-samples.2494600/

.....And most of the time Blender 2.78a was being used (current version is 2.78c).

But notice the huge difference even between 2.78a and 2.78b in the following video:

https://www.youtube.com/watch?v=2XNMYV_1pAw

https://youtu.be/2XNMYV_1pAw?t=50 (38.57s, Blender 2.78a)

https://youtu.be/2XNMYV_1pAw?t=82 (27.04s. Blender 2.78b first run)

https://youtu.be/2XNMYV_1pAw?t=115 (25.66s, Blender 2.78b second run)

EDIT: Added times next to the links above.

 

[Carfax83]

20.26 seconds for my 6900K at 4.3ghz.

 

[Blockheadfan]

27.8s
Stock 1700 3200 CL14

 

[cbn]

27.8s
Stock 1700 3200 CL14

So even with the 3.0 GHz base clock it is 8.2 second faster than the Ryzen 7 @ 3.4 GHz reported this post.

Nice.

 

[cbn]

After watching this video mentioning Linux as being faster than Windows 10 when using CPU for Blender, I decided to re-run the benchmark using my stock speed W3670 (post #4) this time with Linux Mint 18.1 MATE (rather than Windows 10):

W3670 (stock speed) with Windows 10: 1.09.27 (1 minute, 9.27 seconds).

W3670 (stock speed) with Linux Mint 18.1 : 52.49 seconds

So definitely a non-trival improvement in rendering time for Linux compared to Windows 10.

 

[Justinus]

AMD ThreadRipper 13.04 seconds（render=100）

http://www.tomshardware.com/news/amd-threadripper-vega-pcie-lanes,34581.html

So, wait. The file is set to 150 samples automatically. Are you saying the ThreadRipper 13.04 second result was at 100 samples?

 

[csbin]

So, wait. The file is set to 150 samples automatically. Are you saying the ThreadRipper 13.04 second result was at 100 samples?

（render=150）1800X≈36 seconds
（render=100）1800X≈24 seconds

 

[Justinus]

（render=150）1800X≈36 seconds
（render=100）1800X≈24 seconds

Well, that seems plausible then.

My 5960x scored 13.73 seconds at 100 samples with the linked version of Blender and the benchmark. AVX ftw?

 

[cbn]

（render=150）1800X≈36 seconds
（render=100）1800X≈24 seconds

These look like render times from the old blender.

Are you sure ThreadRipper was running 100 samples?

 

[csbin]

Well, that seems plausible then.

My 5960x scored 13.73 seconds at 100 samples with the linked version of Blender and the benchmark. AVX ftw?

Benchmark Use 2.78a

 

[csbin]

These look like render times from the old blender.

Are you sure ThreadRipper was running 100 samples?

sure

ThreadRipper Use 2.78a，No SIMD optimization

https://www.pcper.com/image/view/82405?return=node/67835

 

[cbn]

sure

ThreadRipper Use 2.78a，No SIMD optimization

https://www.pcper.com/image/view/82405?return=node/67835

I'm using 2.78c right now and my screen looks the same.

P.S. In this thread they believe Threadripper was running 150 samples......and that makes sense based on the improvement Blender received since 2.78a was released. (See post #6 for a comparison of 2.78a and 2.78b. Also look at Blockheadfan's time of 27.8s for a stock speed Ryzen 7 1700 with DDR4 3200 CL14 RAM in post #8.....these are much lower than the times you posted in #12)

 

[cbn]

（render=150）1800X≈36 seconds
（render=100）1800X≈24 seconds

36 seconds is from the old blender:

https://forums.anandtech.com/thread...and-discussion.2499879/page-252#post-38916696

http://www.tomshardware.com/news/amd-threadripper-vega-pcie-lanes,34581.html


AMD demonstrated ThreadRipper's performance in a few scenarios, including the now-famous Blender rendering test of the Ryzen CPU and logo. The render required a mere 13.04 seconds, which is much faster than the 36 seconds AMD demonstrated with the Ryzen 7 1800X prior to its launch.

 

[Carfax83]

With 100 samples, I got 13.20 seconds compared to the 20.26 second I got with 150 samples.

 

[formulav8]

Has AMD said exactly which version they used? It could have a notable impact.

 

[formulav8]

I ran the latest version, opened the file and hit start. Got 36.2s. Stock 1600 and ram running @ 2400 (was 2933 until Asus 0613 bios messed things up).

Edit: Ended up with a 35.5 after 2 runs.

 

[cbn]

Has AMD said exactly which version they used? It could have a notable impact.

The c version of Blender has been available since Feb. 28th 2017:

https://www.blender.org/features/2-78/

An updated Blender 2.78a was released on October 26, 2016. That release has 69 important bug fixes. The update Blender 2.78b was released on February 8, 2017. Due to an error in the binary, a 'c' version was released on February 28, 2017. Check the 2.78c changelog.

So I can't imagine them using anything different....especially as it is so much faster than 2.78a.

 

[JimmiG]

Well it seems everyone is running different versions and render settings so hard to compare any results..
FWIW, running 2.78c and just opening the file and hitting F12, I got 23.86 seconds. 1800X @stock.

 

[moonbogg]

No one knows what version or sample rate to use apparently.

 

[cbn]

No one knows what version or sample rate to use apparently.

If a person follows the instructions in the opening post they will be running version 2.78c and the samples will be 150.