Post your Ryzen Blender Demo Scores! (AMD clarifies Blender Benchmark Confusion, Run @ 150 Samples)

[mikk]

With samples 100 my i7-6700k running at 4.0 Ghz renders the image in 38 seconds. With 200 samples 1min15. I'm almost matching the score of 8 Core Zen, how is this possible?

 

[superstition]

With samples 100 my i7-6700k running at 4.0 Ghz renders the image in 38 seconds. With 200 samples 1min15. I'm almost matching the score of 8 Core Zen, how is this possible?

How is it possible that my 2013 47 watt TDP laptop i7 got your score with 100 samples?

That's some EDRAM! Either that or OS X is just a lot more efficient than Windows.

 

[mikk]

How is it possible that my 2013 47 watt TDP laptop i7 got your score with 100 samples?

That's some EDRAM!

It's odd. I'm using Windows 10. Either there are big differences between different operating systems or memory bandwidth/latency is very important for this bench.

 

[jhu]

How is it possible that my 2013 47 watt TDP laptop i7 got your score with 100 samples?

That's some EDRAM! Either that or OS X is just a lot more efficient than Windows.

Definitely not OS X. Most likely the EDRAM.

 

[superstition]

Definitely not OS X. Most likely the EDRAM.

That was my assumption as well since OS overhead shouldn't have all that dramatic an effect on a benchmark like this. Also, cleaning off my desktop, quitting everything, and turning off WiFi didn't make much of a difference.

If it is the EDRAM it's another example of how Peter Bright was right:

http://arstechnica.com/gadgets/2015...obbing-us-of-the-performance-king-we-deserve/

 

[Paul98]

With samples 100 my i7-6700k running at 4.0 Ghz renders the image in 38 seconds. With 200 samples 1min15. I'm almost matching the score of 8 Core Zen, how is this possible?

You're not, people seem to be missing that for that setting it took Ryzen 25 seconds.

Thanks for the change OP

 

[mikk]

You're not, people seem to be missing that for that setting it took Ryzen 25 seconds.

It took 34 seconds in the Demo they showcased.


Memory latency/bandwidth doesn't effect the render time on my i7-6700k system.


DDR4-2133= 38.67s
DDR4-3200= 38.43s

 

[Paul98]

It took 34 seconds in the Demo they showcased.


Memory latency/bandwidth doesn't effect the render time on my i7-6700k system.


DDR4-2133= 38.67s
DDR4-3200= 38.43s

..... well just keep on posting bad info....

The press event where that screenshot was from, has them taking 25 seconds. Which also reflects what people with a 6900k get in their testing

 

[IEC]

37.65 seconds. 6700K stock settings, DDR4-3466, Win 10 Pro. Nothing changed except samples set to 100.

 

[Face2Face]

5820K @ 4.25GHz
Uncore @ 4GHz
DDR4 @ 3000Mhz

200 Samples
54.01 sec - About equal to a stock 5960X

100 Samples
27.18 sec

The Press Blender demo they did was @ 100 samples(Pause video at 1:04) and RYZEN @ 3.4GHz completed it in 24.XX seconds. I can't make out the last two digits(could be a 32 or 82), but it's definitely 24 secs. So it looks correct to me, comparing my scores to the 5960X.

https://www.youtube.com/watch?v=7yxSFmEOkrA

Would've been a whole lot easier if they would've used Cinebench or the CPU-Z benchmark

 

[RichUK]

How do you set it to 100 samples?

I'm using 2.78a

Never mind, found it.

What version of blender is everyone else using?

 

[dave_the_nerd]

I don't control the turbo. It's set up in the default configuration. There is no way I know of to manually adjust Macbook Pro BIOS settings for clock rate and such. It's a "Late 2013" Macbook Pro running OS X 10.9.5.

I did find a utility to disable turbo mode - helps temps a lot if I'm gaming.

 

[superstition]

I did find a utility to disable turbo mode - helps temps a lot if I'm gaming.

Interesting. I gave up on gaming with the Macbook because of fan noise. My 2008 model was better because when I would place it atop a large cast iron skillet it would transmit its heat into that and the fan wouldn't get loud. The 2013 model doesn't work with that kludge.

 

[Phynaz]

Kinda fun to have had Ryzen's next-generation performance in my laptop since late 2013.

Are you saying AMD will be four years behind Intel?

 

[RichUK]

29.64 second (100 samples).

Here's mine:

 

[HilbertSpace]

That's a great result RichUK - almost matching 6-core Intel chips with 4-core Kaby Lake.

My results:
MacBook Pro (Retina, 15-inch, Early 2013) - Intel i7-3630QM (2.4-3.4 GHz) 4 cores/8 threads
53.00 Seconds - 100 samples - blender 2.78a - MacOS 10.12.2

74.13 seconds - 150 samples - blender 2.78a - MacOS 10.12.2

 

[LTC8K6]

RichUK can you run it with your 7700K at stock?

 

[thesmokingman]

6700k @ 4.7 = 32.91

@7700k@5.2 damn that's a lot of voltage!

 

[RichUK]

RichUK can you run it with your 7700K at stock?

100% stock settings - 34.17 Seconds

(note: speedstep is enabled, hence why the CPU / Uncore frequencies are varied):

 

[Hail The Brain Slug]

5960x at 4.5 GHz

100 samples 2.78a: 19.69
100 samples AVX256 optimized version: 13.80

edit: I had unity, multiple VS instances, and chrome with several tabs going so they weren't clean runs.

 

[VirtualLarry]

Hmm, how do you set "Samples"? I'm a little intimidated by this program's vast menus.

I got ~1:57 on a 4.455Ghz i5-6400 (BCLK 165.0), with DDR4-2400, with default "Samples" setting.

 

[William Gaatjes]

My A10-6700 Richland needs 2:53:32. to render the image. Both render and preview sampling presets are set at 100.
DDR3@1600MHz.

edit:
The boost clock goes up to 4199MHz and then varies between 3699MHz and 4199MHz.
Having only 2 FPU out of 4 "cores" sure does not help for rendering.
And the devastator gpu i have is not supported(Pre gcn, VLIW4=terascale3)

 

[LTC8K6]

Thanks RichUK.

 

[William Gaatjes]

This calculation means nothing, but the a10-6700 only has 2 FPUs. Pure hypothetically if it would have 8 FPUs i could take the score 2:53 = 173 seconds/4 = ~43 seconds.

 

[wyssn11]